system/docpack: move the FASM manual out of kolibri.img, fix keys layout in docs (#739)

- Move the FASM manual out of DOCPACK to `/kolibrios/develop/fasm/fasm.txt` on ISO, next to the Oberon-07 and TCC docs
- Update the document letters in the Russian `guide.htm`, they were one letter off before
- Default document for an unknown parameter stays KFar keys

| DOCPACK, kpacked, sectors | en_US | ru_RU | es_ES |
|---|---|---|---|
| before | 218 + 150 | 232 + 62 | 218 + 199 |
| after | 95 + 468 | 105 + 466 | 95 + 493 |

Reviewed-on: #739
Reviewed-by: Kiril Lipatov <lipatov.kiril@gmail.com>
Reviewed-by: Mikhail Frolov <mixa.frolov2003@gmail.com>

.gitea/CODEOWNERS

@@ -0,0 +1,54 @@
+# =============================================================================
+# Apps
+# User-facing applications, games, multimedia, UI and desktop resources
+# =============================================================================
+
+^programs/(cmm|demos|emulator|games|media|other)/.*       @KolibriOS/apps
+^contrib/(games|media|other)/.*                           @KolibriOS/apps
+^skins/.*                                                 @KolibriOS/apps
+
+
+# =============================================================================
+# System
+# Kernel, drivers, filesystems, SDK, libraries, toolchain and system utilities
+# =============================================================================
+
+^kernel/.*                                                @KolibriOS/system
+^drivers/.*                                               @KolibriOS/system
+
+^programs/(bcc32|develop|fs|hd_load|system|testing)/.*    @KolibriOS/system
+^contrib/(C_Layer|sdk|toolchain)/.*                       @KolibriOS/system
+
+# Shared system/API includes in programs/
+^programs/[^/]+\\.inc$                                    @KolibriOS/system
+
+
+# =============================================================================
+# Network
+# Network stack, network drivers, protocols, libraries and network applications
+# =============================================================================
+
+^kernel/trunk/network/.*                                  @KolibriOS/network
+
+^drivers/ethernet/.*                                      @KolibriOS/network
+^drivers/(mii|netdrv)\\.inc$                              @KolibriOS/network
+
+^programs/network/.*                                      @KolibriOS/network
+^programs/network\\.inc$                                  @KolibriOS/network
+
+^programs/develop/libraries/(http|network|kos_mbedtls)/.* @KolibriOS/network
+
+^contrib/network/.*                                       @KolibriOS/network
+
+
+# =============================================================================
+# DevOps
+# CI/CD, repository infrastructure, build system and autobuild configuration
+# =============================================================================
+
+^\\.gitea/.*                                              @KolibriOS/devops
+^_tools/.*                                                @KolibriOS/devops
+
+^\\.gitmodules$                                           @KolibriOS/devops
+^build\\.txt$                                             @KolibriOS/devops
+^tup\\.config\\.template$                                 @KolibriOS/devops

.gitea/ansible/playbooks/deploy.yml

@@ -19,3 +19,62 @@
         owner: "{{ storage_user }}"
         group: "{{ storage_user }}"
         mode: preserve
+
+    # per-program binaries browsable at <version>/<lang>/data/, like on the old builds server
+    - name: Creating build tree folders
+      ansible.builtin.file:
+        path: "{{ storage_path }}/{{ version }}/{{ item | basename | regex_replace('\\.tar\\.gz$', '') }}/data"
+        state: directory
+        owner: "{{ storage_user }}"
+        group: "{{ storage_user }}"
+        mode: '0755'
+      with_fileglob: "{{ tree_artifact_path }}/*.tar.gz"
+
+    - name: Unpacking build trees
+      ansible.builtin.unarchive:
+        src: "{{ item }}"
+        dest: "{{ storage_path }}/{{ version }}/{{ item | basename | regex_replace('\\.tar\\.gz$', '') }}/data"
+        owner: "{{ storage_user }}"
+        group: "{{ storage_user }}"
+      with_fileglob: "{{ tree_artifact_path }}/*.tar.gz"
+
+    # stable URL for the newest build: <storage_path>/latest/<lang>/...
+    - name: Updating latest symlink
+      ansible.builtin.file:
+        src: "{{ version }}"
+        dest: "{{ storage_path }}/latest"
+        state: link
+        force: true
+        owner: "{{ storage_user }}"
+        group: "{{ storage_user }}"
+
+    # emulators hardcode <site root>/<lang>/data/data/kolibri.img; storage_path is the ci folder, one level under the site root
+    - name: Sending bare images to stable emulator paths
+      ansible.builtin.copy:
+        src: "{{ stable_artifact_path }}/"
+        dest: "{{ storage_path }}/../"
+        mode: preserve
+
+    # older builds shrink to compressed images and lose their unpacked tree; the newest few keep both
+    - name: Looking for zstd on the storage server
+      ansible.builtin.shell: command -v zstd
+      register: zstd_present
+      changed_when: false
+      failed_when: false
+
+    - name: Requiring zstd for compaction
+      ansible.builtin.assert:
+        that: zstd_present.rc == 0
+        fail_msg: >-
+          zstd is not installed on the storage server, so older builds cannot
+          be compacted. Install it and re-run this build.
+
+    - name: Compacting older builds
+      ansible.builtin.script:
+        cmd: >-
+          ../../utils/compact-builds.sh
+          {{ storage_path | quote }}
+          {{ compact_keep | default(5) | int }}
+          {{ compact_max | default(10) | int }}
+      register: compaction
+      changed_when: "'compacted 0 of' not in compaction.stdout"

.gitea/utils/compact-builds.sh

@@ -0,0 +1,74 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0-only
+# SPDX-FileCopyrightText: 2026 KolibriOS team
+
+# Shrink older builds on the storage server: compress their images and drop
+# their unpacked tree, which is rebuildable from the commit. sha256sums.txt
+# keeps naming the bare images, so a compacted build has to be decompressed
+# before it can be verified.
+#
+# Layout:  <root>/<version>/<lang>/kolibrios-<descr>-<lang>.{img,iso,raw} -> .zst
+#          <root>/<version>/<lang>/data/                                  -> removed
+#
+# Usage: compact-builds.sh <storage-root> [keep] [max-per-run]
+#        DRY_RUN=1 walks the same selection and changes nothing
+
+set -euo pipefail
+
+root=${1:?storage root required}
+# the build just published is always among these, so an upload in flight is safe
+keep=${2:-5}
+# a build is hundreds of MiB to re-compress: cap the first pass over the backlog
+max=${3:-10}
+dry=${DRY_RUN:-0}
+verb=compacted
+[ "$dry" = 1 ] && verb="would compact"
+
+# --rm drops the source only once the archive is written; the box serves the
+# site over NFS, so compaction yields to it
+compress_image() {
+    nice -n 10 ionice -c3 zstd -12 -T1 -q --rm "$1"
+}
+export -f compress_image
+
+leftovers() {
+    local build=$1
+    find "$build" -mindepth 2 -maxdepth 2 \
+        \( -type d -name data \
+           -o -type f \( -name '*.img' -o -name '*.iso' -o -name '*.raw' \) \) \
+        -print -quit
+}
+
+compacted=0
+pending=0
+
+while IFS= read -r version; do
+    dir="$root/$version"
+    [ -n "$(leftovers "$dir")" ] || continue
+
+    pending=$((pending + 1))
+    [ "$compacted" -lt "$max" ] || continue
+
+    if [ "$dry" = 1 ]; then
+        compacted=$((compacted + 1))
+        echo "$verb $version"
+        continue
+    fi
+
+    # a build holds nine images: one per core, each zstd kept to one thread so
+    # the two levels of parallelism do not fight. Archive before dropping the
+    # tree, so a failure here leaves the build merely uncompacted and the next
+    # run picks it up again
+    find "$dir" -mindepth 2 -maxdepth 2 -type f \
+         \( -name '*.img' -o -name '*.iso' -o -name '*.raw' \) -print0 \
+        | xargs -0 -r -P "$(nproc)" -I{} bash -c 'compress_image "$@"' _ {}
+
+    find "$dir" -mindepth 2 -maxdepth 2 -type d -name data -exec rm -rf {} +
+
+    compacted=$((compacted + 1))
+    echo "$verb $version"
+# <tag>-<commit count>-g<sha>: order by the count, never by mtime
+done < <(find "$root" -mindepth 1 -maxdepth 1 -type d -name '*-*-g*' -printf '%f\n' \
+             | sort -t- -k2,2n | head -n "-$keep")
+
+echo "$verb $compacted of $pending older build(s), $((pending - compacted)) left for the next run"

.gitea/workflows/build.yaml

@@ -128,7 +128,8 @@ jobs:
           fi
 
       # cache restore resets ctimes; realign tup DB to skip a full rebuild,
-      # but keep files changed since the cache's commit dirty so tup rebuilds them
+      # but keep files changed since the cache's commit dirty so tup rebuilds them.
+      # A submodule bump diffs as a bare gitlink, so expand it to its whole subtree
       - name: Sync tup DB with filesystem ctime
         if: steps.check-tup-cache.outputs.cache_restored == 'true'
         run: |
@@ -136,6 +137,10 @@ jobs:
           BASE=$(cat .tup/build-commit 2>/dev/null || true)
           if [ -n "$BASE" ] && git cat-file -e "${BASE}^{commit}" 2>/dev/null; then
             git -c core.quotePath=false diff --name-only "$BASE" HEAD > "$DIRTY"
+            git diff --raw "$BASE" HEAD | awk '$2 == "160000" { print $NF }' |
+            while read -r sub; do
+              git -C "$sub" -c core.quotePath=false ls-files | sed "s|^|$sub/|"
+            done >> "$DIRTY"
             echo "Cache built at $BASE; $(wc -l < "$DIRTY") changed files kept dirty"
             python3 .gitea/utils/sync-tup-ctime.py .tup/db "$DIRTY"
           else
@@ -162,12 +167,13 @@ jobs:
           tup init
           tup variant ${{ matrix.lang }}.config
 
-      # build the image for language variant
+      # build the image for language variant; keep the tup log for publishing
       - name: Build KolibriOS
         run: |
+          set -o pipefail
           export PATH=/home/autobuild/tools/win32/bin:$PATH
           source kos32-export-env-vars ${{ gitea.workspace }}
-          tup build-${{ matrix.lang }}
+          tup build-${{ matrix.lang }} 2>&1 | tee build.log
 
       # stamp the cache with its commit so later builds diff against it
       - name: Record cache base commit
@@ -181,15 +187,26 @@ jobs:
         id: vars
         run: echo "descr=$(git describe --tags)" >> $GITEA_OUTPUT
 
-      # rename build outputs to their final names so artifacts are bare files
-      - name: Rename images
+      # stage build outputs under their final names, plus distribution kit and checksums
+      - name: Prepare artifacts
         run: |
+          set -e
+          BASE="kolibrios-${{ steps.vars.outputs.descr }}-${{ matrix.lang }}"
           for ext in img iso raw; do
-            cp "build-${{ matrix.lang }}/data/kolibri.$ext" \
-               "kolibrios-${{ steps.vars.outputs.descr }}-${{ matrix.lang }}.$ext"
+            cp "build-${{ matrix.lang }}/data/kolibri.$ext" "$BASE.$ext"
           done
+          # zip: distr targets Windows/DOS users, Explorer opens it natively
+          (cd "build-${{ matrix.lang }}/data" && zip -r -9 -q "${{ gitea.workspace }}/$BASE.distr.zip" distribution_kit)
+          # whole build tree for per-program downloads; the big images ship separately
+          tar -C "build-${{ matrix.lang }}" \
+              --exclude './data/kolibri.iso' \
+              --exclude './data/kolibri.raw' \
+              --exclude './data/distribution_kit' \
+              -czf "$BASE.tree.tar.gz" .
+          sha256sum "$BASE".img "$BASE".iso "$BASE".raw "$BASE".distr.zip > "$BASE.sha256sums.txt"
+          mv build.log "$BASE.build.log"
 
-      #publish the three images; longer retention for the floppy image.
+      # publish the images; longer retention for the floppy image
       - name: Upload floppy image
         uses: actions/upload-artifact@v3
         with:
@@ -211,6 +228,34 @@ jobs:
           path: kolibrios-${{ steps.vars.outputs.descr }}-${{ matrix.lang }}.raw
           retention-days: 15
 
+      - name: Upload distribution kit
+        uses: actions/upload-artifact@v3
+        with:
+          name: kolibrios-${{ steps.vars.outputs.descr }}-${{ matrix.lang }}.distr.zip
+          path: kolibrios-${{ steps.vars.outputs.descr }}-${{ matrix.lang }}.distr.zip
+          retention-days: 30
+
+      - name: Upload checksums
+        uses: actions/upload-artifact@v3
+        with:
+          name: kolibrios-${{ steps.vars.outputs.descr }}-${{ matrix.lang }}.sha256sums.txt
+          path: kolibrios-${{ steps.vars.outputs.descr }}-${{ matrix.lang }}.sha256sums.txt
+          retention-days: 90
+
+      - name: Upload build log
+        uses: actions/upload-artifact@v3
+        with:
+          name: kolibrios-${{ steps.vars.outputs.descr }}-${{ matrix.lang }}.build.log
+          path: kolibrios-${{ steps.vars.outputs.descr }}-${{ matrix.lang }}.build.log
+          retention-days: 90
+
+      - name: Upload build tree
+        uses: actions/upload-artifact@v3
+        with:
+          name: kolibrios-${{ steps.vars.outputs.descr }}-${{ matrix.lang }}.tree.tar.gz
+          path: kolibrios-${{ steps.vars.outputs.descr }}-${{ matrix.lang }}.tree.tar.gz
+          retention-days: 15
+
   deploy:
     name: "Publish Images"
     runs-on: ubuntu-latest
@@ -235,31 +280,50 @@ jobs:
         run: |
           set -eu
 
-          rm -rf ./norm-artifacts
-          mkdir -p ./norm-artifacts
+          rm -rf ./norm-artifacts ./stable-artifacts ./tree-artifacts
+          mkdir -p ./norm-artifacts ./stable-artifacts ./tree-artifacts
 
-          find ./artifacts -type f \( \
-            -name 'kolibrios-*.img' -o \
-            -name 'kolibrios-*.iso' -o \
-            -name 'kolibrios-*.raw' \
-          \) | while read -r file; do
-            fname="$(basename "$file")"        # kolibrios-<descr>-<lang>.<ext>
-            name_without_ext="${fname%.*}"
-            lang="${name_without_ext##*-}"
+          find ./artifacts -type f -name 'kolibrios-*' | while read -r file; do
+            fname="$(basename "$file")"        # kolibrios-<descr>-<lang>.<suffix>
+            case "$fname" in
+              *.tree.tar.gz)    stem="${fname%.tree.tar.gz}";;
+              *.distr.zip)      stem="${fname%.distr.zip}";;
+              *.sha256sums.txt) stem="${fname%.sha256sums.txt}";;
+              *.build.log)      stem="${fname%.build.log}";;
+              *.img)            stem="${fname%.img}";;
+              *.iso)            stem="${fname%.iso}";;
+              *.raw)            stem="${fname%.raw}";;
+              *) continue;;
+            esac
+            lang="${stem##*-}"                 # <lang> has no dashes, <descr> may
 
             mkdir -p "./norm-artifacts/$lang"
-            cp "$file" "./norm-artifacts/$lang/$fname"
+            case "$fname" in
+              *.tree.tar.gz)
+                # unpacked on the server into <version>/<lang>/data/ for per-program downloads
+                cp "$file" "./tree-artifacts/$lang.tar.gz";;
+              *.sha256sums.txt)
+                cp "$file" "./norm-artifacts/$lang/sha256sums.txt";;
+              *.build.log)
+                cp "$file" "./norm-artifacts/$lang/build.log";;
+              *.img)
+                cp "$file" "./norm-artifacts/$lang/$fname"
+                # emulators (e.g. v86) hardcode <lang>/data/data/kolibri.img at the site root
+                mkdir -p "./stable-artifacts/$lang/data/data"
+                cp "$file" "./stable-artifacts/$lang/data/data/kolibri.img";;
+              *)
+                cp "$file" "./norm-artifacts/$lang/$fname";;
+            esac
           done
 
-          # derive the version (git describe) from an artifact name:
-          # kolibrios-<descr>-<lang>.<ext> -> <descr>
-          sample="$(find ./norm-artifacts -type f -name 'kolibrios-*' | head -n1)"
+          # derive the version (git describe) from an artifact name: kolibrios-<descr>-<lang>.iso -> <descr>
+          sample="$(find ./norm-artifacts -type f -name 'kolibrios-*.iso' | head -n1)"
           base="$(basename "${sample%.*}")"   # kolibrios-<descr>-<lang>
           version="${base%-*}"                # kolibrios-<descr>
           version="${version#kolibrios-}"     # <descr>
           echo "version=$version" >> $GITEA_OUTPUT
 
-          find ./norm-artifacts -type f -print
+          find ./norm-artifacts ./stable-artifacts ./tree-artifacts -type f -print
 
       - name: Setup SSH
         run: |
@@ -285,4 +349,6 @@ jobs:
             -e "storage_path=${{ secrets.STORAGE_PATH }}" \
             -e "version=${{ steps.normalize.outputs.version }}" \
             -e "artifact_path=${{ gitea.workspace }}/norm-artifacts" \
+            -e "stable_artifact_path=${{ gitea.workspace }}/stable-artifacts" \
+            -e "tree_artifact_path=${{ gitea.workspace }}/tree-artifacts" \
             -v

.gitmodules

@@ -4,7 +4,7 @@
 	branch = main
 [submodule "programs/develop/oberon07"]
 	path = programs/develop/oberon07
-	url = https://github.com/AntKrotov/oberon-07-compiler.git
+	url = https://git.kolibrios.org/KolibriOS/oberon-07-compiler.git
 	branch = master
 [submodule "programs/cmm"]
 	path = programs/cmm

data/Tupfile.lua

@@ -41,10 +41,6 @@ img_files = {
  {"LIB/KMENU.OBJ", "common/lib/kmenu.obj"},
  {"LIB/PIXLIB.OBJ", "common/lib/pixlib.obj"},
  {"MEDIA/AC97SND", "common/media/ac97snd"},
- {"MEDIA/IMGF/IMGF", "common/media/ImgF/ImgF"},
- {"MEDIA/IMGF/CEDG.OBJ", "common/media/ImgF/cEdg.obj"},
- {"MEDIA/IMGF/DITHER.OBJ", "common/media/ImgF/dither.obj"},
- {"MEDIA/IMGF/INVSOL.OBJ", "common/media/ImgF/invSol.obj"},
  {"NETWORK/FTPC.INI", SRC_PROGS .. "/network/ftpc/ftpc.ini"},
  {"NETWORK/FTPD.INI", "common/network/ftpd.ini"},
  {"NETWORK/KNMAP", "common/network/knmap"},
@@ -58,8 +54,7 @@ img_files = {
  {"SETTINGS/FB2READ.INI", "common/settings/fb2read.ini"},
  {"SETTINGS/HA.CFG", SRC_PROGS .. "/other/ha/SETTINGS/HA.CFG"},
  {"SETTINGS/ICON.INI", build_type .. "/settings/icon.ini"},
- {"SETTINGS/KEYMAP.KEY", SRC_PROGS .. "/system/taskbar/KEYMAP.KEY"},
- {"SETTINGS/KOLIBRI.LBL", build_type .. "/settings/kolibri.lbl"},
+ {"SETTINGS/KEYMAP.KEY", VAR_PROGS .. "/system/taskbar/keymap.key.kpack"},
  {"SETTINGS/LANG.INI", build_type .. "/settings/lang.ini"},
  {"SETTINGS/MENU.DAT", build_type .. "/settings/menu.dat"},
  {"SETTINGS/NETWORK.INI", "common/settings/network.ini"},
@@ -171,6 +166,7 @@ extra_files = {
  {"kolibrios/demos/rpento.lif", "common/demos/rpento.lif"},
  {"kolibrios/develop/fpc/", "common/develop/fpc/*"},
  {"kolibrios/develop/fpc/examples/", "../programs/develop/fp/examples/src/*"},
+ {"kolibrios/develop/fasm/fasm.txt", SRC_PROGS .. "/develop/fasm/1.73/fasm.txt"},
  {"kolibrios/develop/oberon07/compiler.kex", SRC_PROGS .. "/develop/oberon07/Compiler.kex"},
  {"kolibrios/develop/oberon07/LICENSE", SRC_PROGS .. "/develop/oberon07/LICENSE"},
  {"kolibrios/develop/oberon07/doc/CC.txt", SRC_PROGS .. "/develop/oberon07/doc/CC.txt"},
@@ -273,6 +269,10 @@ extra_files = {
  {"kolibrios/lib/i915_dri.drv", "common/lib/i915_dri.drv"},
  {"kolibrios/media/fplay", "common/media/fplay"},
  {"kolibrios/media/fplay_run", "common/media/fplay_run"},
+ {"kolibrios/media/ImgF/ImgF", "common/media/ImgF/ImgF"},
+ {"kolibrios/media/ImgF/cEdg.obj", "common/media/ImgF/cEdg.obj"},
+ {"kolibrios/media/ImgF/dither.obj", "common/media/ImgF/dither.obj"},
+ {"kolibrios/media/ImgF/invSol.obj", "common/media/ImgF/invSol.obj"},
  {"kolibrios/media/minimp3", "common/media/minimp3"},
  {"kolibrios/media/updf", SRC .. "/contrib/media/updf/build/uPDF"},
  {"kolibrios/media/vttf", "common/media/vttf"},
@@ -487,11 +487,6 @@ tup.append_table(img_files, {
  {"DEVELOP/MTDBG", VAR_PROGS .. "/develop/mtdbg/mtdbg"},
  {"DEVELOP/MSTATE", VAR_PROGS .. "/develop/mstate/mstate"},
  {"DEVELOP/SCANCODE", VAR_PROGS .. "/develop/scancode/scancode"},
- {"DEVELOP/EXAMPLES/CIRCLE", VAR_PROGS .. "/develop/examples/circle/circle"},
- {"DEVELOP/EXAMPLES/COLORREF", VAR_PROGS .. "/demos/colorref/colorref"},
- {"DEVELOP/EXAMPLES/CONGET", VAR_PROGS .. "/develop/libraries/console_coff/examples/test_gets"},
- {"DEVELOP/EXAMPLES/CSLIDE", VAR_PROGS .. "/demos/cslide/cslide"},
- {"DEVELOP/EXAMPLES/THREAD", VAR_PROGS .. "/develop/examples/thread/thread"},
  {"File Managers/KFAR", VAR_PROGS .. "/fs/kfar/kfar"},
  {"File Managers/OPENDIAL", VAR_PROGS .. "/fs/opendial/opendial"},
  {"LOD", VAR_PROGS .. "/fs/lod/lod"},
@@ -573,6 +568,7 @@ tup.append_table(img_files, {
  {"DRIVERS/UHCI.SYS", VAR_DRVS .. "/usb/uhci.sys"},
  {"DRIVERS/OHCI.SYS", VAR_DRVS .. "/usb/ohci.sys"},
  {"DRIVERS/EHCI.SYS", VAR_DRVS .. "/usb/ehci.sys"},
+ {"DRIVERS/XHCI.SYS", VAR_DRVS .. "/usb/xhci/xhci.sys"},
  {"DRIVERS/USBHID.SYS", VAR_DRVS .. "/usb/usbhid/usbhid.sys"},
  {"DRIVERS/USBFTDI.SYS", VAR_DRVS .. "/usb/usbftdi/usbftdi.sys"},
  {"DRIVERS/USBOTHER.SYS",VAR_DRVS .. "/usb/usbother/usbother.sys"},
@@ -597,6 +593,7 @@ tup.append_table(img_files, {
  {"DRIVERS/VT823X.SYS", VAR_DRVS .. "/audio/vt823x.sys"},
  {"DRIVERS/SIS.SYS", VAR_DRVS .. "/audio/sis.sys"},
  {"DRIVERS/SDHCI.SYS", "../drivers/sdhci/sdhci.sys"},
+ {"DRIVERS/I2CHID.SYS", VAR_DRVS .. "/i2chid/i2chid.sys"},
 })
 tup.append_table(extra_files, {
  {"HD_Load/9x2klbr/", VAR_PROGS .. "/hd_load/9x2klbr/9x2klbr.exe"},
@@ -650,20 +647,28 @@ tup.append_table(extra_files, {
  {"kolibrios/3D/blocks/block.bin", VAR_PROGS .. "/bcc32/games/blocks/block.bin"},
  {"kolibrios/develop/tcc/lib/crt0.o", VAR_PROGS ..  "/develop/ktcc/bin/lib/crt0.o", group = "../programs/develop/ktcc/<crt0.o>"},
  {"kolibrios/develop/tcc/lib/tiny.o", VAR_PROGS ..  "/develop/ktcc/bin/lib/tiny.o", group = "../programs/develop/ktcc/<tiny.o>"},
- {"kolibrios/develop/tcc/lib/libsound.a", VAR_PROGS ..  "/develop/ktcc/bin/lib/libsound.a", group = "../programs/develop/ktcc/<libsound.a>"}
+ {"kolibrios/develop/tcc/lib/libsound.a", VAR_PROGS ..  "/develop/ktcc/bin/lib/libsound.a", group = "../programs/develop/ktcc/<libsound.a>"},
+ {"kolibrios/develop/examples/circle", VAR_PROGS .. "/develop/examples/circle/circle"},
+ {"kolibrios/develop/examples/colorref", VAR_PROGS .. "/demos/colorref/colorref"},
+ {"kolibrios/develop/examples/conget", VAR_PROGS .. "/develop/libraries/console_coff/examples/test_gets"},
+ {"kolibrios/develop/examples/cslide", VAR_PROGS .. "/demos/cslide/cslide"},
+ {"kolibrios/develop/examples/thread", VAR_PROGS .. "/develop/examples/thread/thread"}
 })
--- For russian build, add russian-only programs.
-if build_type == "ru_RU" then tup.append_table(img_files, {
- {"DEVELOP/EXAMPLES/TESTCON2", VAR_PROGS .. "/develop/libraries/console_coff/examples/testcon2_rus"},
-}) else tup.append_table(img_files, {
- {"DEVELOP/EXAMPLES/TESTCON2", VAR_PROGS .. "/develop/libraries/console_coff/examples/testcon2_eng"},
-}) end
+-- -- For russian build, add russian-only programs.
+-- if build_type == "ru_RU" then tup.append_table(img_files, {
+-- {"DEVELOP/EXAMPLES/TESTCON2", VAR_PROGS .. "/develop/libraries/console_coff/examples/testcon2_rus"},
+-- }) else tup.append_table(img_files, {
+-- {"DEVELOP/EXAMPLES/TESTCON2", VAR_PROGS .. "/develop/libraries/console_coff/examples/testcon2_eng"},
+-- }) end
 
 if build_type == "ru_RU" then tup.append_table(extra_files, {
+ {"kolibrios/develop/examples/testcon2", VAR_PROGS .. "/develop/libraries/console_coff/examples/testcon2_rus"},
  {"kolibrios/utils/period", VAR_PROGS .. "/other/period/period"},
  {"kolibrios/games/Dungeons/Dungeons", VAR_PROGS .. "/games/Dungeons/Dungeons"},
  {"kolibrios/games/klavisha/klavisha", VAR_PROGS .. "/games/klavisha/klavisha"},
  {"kolibrios/games/klavisha/basekurs.kla", "ru_RU/games/basekurs.kla"},
+}) else tup.append_table(extra_files, {
+ {"kolibrios/develop/examples/testcon2", VAR_PROGS .. "/develop/libraries/console_coff/examples/testcon2_eng"},
 }) end
 
 end -- tup.getconfig('NO_FASM') ~= 'full'

data/common/settings/AUTORUN.DAT

@@ -6,8 +6,9 @@
 #/SYS/CROPFLAT          "XS800 YS480"	1	# set limits of screen
 #/SYS/LOADDRV           COMMOUSE	1	# serial mouse driver
 #/SYS/DEVELOP/BOARD     ""   0	# Load DEBUG board
+#/SYS/LOADDRV            RDC  0	# videodriver for RDC M2010/M2012
 /SYS/SETUP              BOOT 0	# Load device settings
-/SYS/LOADDRV            RDC  0	# videodriver for RDC M2010/M2012
+/SYS/LOADDRV            I2CHID 0	# Start I2C touchpad driver
 /SYS/@TASKBAR           ""   0	# Start taskbar
 /SYS/ESKIN              ""   0	# Set style
 /SYS/@ICON              ""   0	# Multithread desktop icons

data/common/settings/system.ini

@@ -5,6 +5,22 @@ font smoothing=sbp
 font height=9
 speaker mute=off
 
+[touchpad]
+speed=100
+; how far a finger must move before the pointer follows, in pad units;
+; larger is steadier under a pressing finger, 0 turns the filter off
+jitter=1
+; the same, for a finger that landed on the button strip: it is pressing
+; rather than steering there, so it needs a stricter margin
+jitter_buttons=45
+scroll_step=64
+scroll_invert=0
+tap_enable=1
+tap_drag=1
+right_click=1
+button_zone=15
+right_split=50
+
 [mouse]
 speed=4
 acceleration=3

data/en_US/settings/menu.dat

@@ -40,7 +40,7 @@
 06 zSea Image Viewer  * |/kolibrios/media/zsea/zsea
 38 Animage              |media/animage
 38 IconEdit             |media/iconedit
-40 Image Filtering      |media/imgf/imgf
+40 Image Filtering    * |/kolibrios/media/ImgF/ImgF
 52 Palitra              |media/palitra
 39 Pipet                |develop/pipet
 #4 **** MULTIMEDIA

data/es_ES/settings/menu.dat

@@ -41,7 +41,7 @@
 16 Image viewer KIV      |media/kiv
 16 Image viewer zSea  *  |/kolibrios/media/zsea/zsea
 16 Animage               |media/animage
-16 Efectos de imаgenes   |media/imgf/imgf
+16 Efectos de imаgenes * |/kolibrios/media/ImgF/ImgF
 16 Palitra               |media/palitra
 #4   **** MULTIMEDIA
 16 Pixie Audioplayer     |media/pixie

data/et_EE/settings/menu.dat

@@ -41,7 +41,7 @@
 06 Image viewer zSea  *  |/kolibrios/media/zsea/zsea
 38 Animage               |media/animage
 38 IconEdit              |media/iconedit
-16 Image Filtering       |media/imgf/imgf
+16 Image Filtering    *  |/kolibrios/media/ImgF/ImgF
 52 Palitra               |media/palitra
 39 Pipet                 |develop/pipet
 #4   **** MULTIMEDIA ****

data/it_IT/settings/menu.dat

@@ -41,7 +41,7 @@
 06 Image viewer zSea  *  |/kolibrios/media/zsea/zsea
 38 Animage               |media/animage
 38 IconEdit              |media/iconedit
-40 Image Filtering       |media/imgf/imgf
+40 Image Filtering    *  |/kolibrios/media/ImgF/ImgF
 52 Palitra               |media/palitra
 39 Pipet                 |develop/pipet
 #4 **** MULTIMEDIA

data/ru_RU/docs/guide/guide.htm

@@ -20,13 +20,13 @@
 <a href="hardware_support.htm">Поддержка оборудования</a>
 
 <strong>Разработка</strong>
-<a href="/sys/docpack|f">Справочник по системным функциям Колибри</a>
-<a href="/sys/docpack|e">Использование отладчика</a>
-<a href="/sys/docpack|d">Руководство FASM</a>
+<a href="/sys/docpack|d">Справочник по системным функциям Колибри</a>
+<a href="/sys/docpack|c">Использование отладчика</a>
+<a href="/kolibrios/develop/fasm/fasm.txt">Руководство FASM</a>
 
 <strong>Использование библиотек</strong>
-<a href="/sys/docpack|i">Ini</a>
-<a href="/sys/docpack|j">Диалог открытия файла OpenDial</a>
+<a href="/sys/docpack|g">Ini</a>
+<a href="/sys/docpack|h">Диалог открытия файла OpenDial</a>
 
 <strong>Смотрите также</strong>
 Вики проекта <a href="http://wiki.kolibrios.org/">wiki.kolibrios.org</a>

data/ru_RU/settings/menu.dat

@@ -40,7 +40,7 @@
 06 Просмотр фото zSea   |/kolibrios/media/zsea/zsea
 38 Редактор Animage     |media/animage
 38 Редактор иконок      |media/iconedit
-40 Наложение фильтров   |media/imgf/imgf
+40 Наложение фильтров * |/kolibrios/media/ImgF/ImgF
 52 Палитра              |media/palitra
 39 Пипетка              |develop/pipet
 #4 **** Звук и видео

drivers/audio/intel_hda/hdaudio.asm

@@ -273,8 +273,8 @@ RIRB_INT_RESPONSE	 equ  0x01
 RIRB_INT_OVERRUN	 equ  0x04
 RIRB_INT_MASK		 equ  0x05
 
-; STATESTS int mask: SD2,SD1,SD0
-STATESTS_INT_MASK	 equ  0x07
+; STATESTS int mask: SDIWAKE[14:0], a 16-bit register
+STATESTS_INT_MASK	 equ  0x7FFF
 AZX_MAX_CODECS		 equ  4
 
 ; SD_CTL bits
@@ -402,6 +402,7 @@ struc AC_CNTRL		    ;AC controller base class
 	.response_reset     dd 0
 	.polling_mode	    db 0
 	.poll_count	    db 0
+	.no_irq 	    db 0 ; no interrupt line: a thread runs hda_irq
 	.posbuf 	    dd ?
 	.start_wallclk	    dd ? ; start + minimum wallclk
 	.period_wallclk     dd ? ; wallclk for period
@@ -505,8 +506,6 @@ macro spin_unlock
 	pop	ebx eax
 }
 
-data fixups
-end data
 
 include '../../struct.inc'
 include '../../macros.inc'
@@ -607,9 +606,6 @@ if IRQ_REMAP
 	invoke	SysMsgBoardStr
 end if
 
-	mov	ebx, [ctrl.int_line]
-	invoke	AttachIntHandler, ebx, hda_irq, dword 0
-
 ;Asper This part is from "azx_probe" proc. [
 	call	azx_codec_create
 	cmp	eax, 0
@@ -858,13 +854,13 @@ end if
 
 	mov	edx, ICH6_REG_SD_STS + SDO_OFS
 	call	azx_readb
-	mov	bl, al
+	mov	cl, al
 
 	mov	al, SD_INT_MASK
 	mov	edx, ICH6_REG_SD_STS + SDO_OFS
 	call	azx_writeb
 
-	test	bl, SD_INT_COMPLETE
+	test	cl, SD_INT_COMPLETE
 	jz	@f
 
 	mov	eax, [ctrl.civ_val]
@@ -901,13 +897,13 @@ end if
 ;if 0
 	; clear state status int
 	mov	edx, ICH6_REG_STATESTS
-	call	azx_readb
-	test	al, 0x04
+	call	azx_readw
+	and	eax, STATESTS_INT_MASK
 	jz	@f
 
-	mov	al, 0x04
+	; ack all state-change bits, else the level IRQ line stays asserted
 	mov	edx, ICH6_REG_STATESTS
-	call	azx_writeb
+	call	azx_writew
 @@:
 ;end if
 	or	eax, 1
@@ -915,6 +911,24 @@ end if
 	ret
 endp
 
+; Stand-in for the interrupt when no interrupt line is available. A thread,
+; not a TimerHS callback: the kernel calls timer callbacks with the global
+; timer list lock held, and hda_irq may stall and refill the audio buffer.
+; hda_irq runs with interrupts off, as it would in IRQ context: it holds
+; aspinlock, which azx_rirb_get_response waits for with interrupts off, so
+; a holder that could be preempted would hang the machine.
+proc hda_poll_thread
+.loop:
+	pushfd
+	cli
+	call	hda_irq
+	popfd
+	mov	eax, 5		; delay, sleeping instead of spinning
+	mov	ebx, 1		; 10 ms; a 16 KB period is ~85 ms at 48 kHz
+	int	0x40
+	jmp	.loop
+endp
+
 
 align 4
 proc create_primary_buff
@@ -1203,6 +1217,26 @@ end if
 	; allocate CORB/RIRB
 	call	azx_alloc_cmd_io
 
+	; attach the IRQ handler BEFORE the controller may raise interrupts:
+	; azx_init_chip enables GIE/CIE and a codec state-change can fire
+	; right away; an unclaimed level IRQ makes the
+	; kernel poll foreign handlers and relink their IRQH nodes onto our
+	; line (see irq_serv .try_other_irqs), corrupting the dispatch list
+	; a warm reboot can leave GIE and a stale completion bit behind, and
+	; civ_val is never resynced, so silence the controller before attaching
+	call	azx_int_disable
+	mov	ebx, [ctrl.int_line]
+	invoke	AttachIntHandler, ebx, hda_irq, dword 0
+	test	eax, eax
+	jnz	@f
+	; No usable interrupt line (INTLINE=0xFF on boards that route INTx only
+	; through ACPI _PRT or MSI, which the kernel does not parse). The
+	; controller still latches INTSTS/SD_STS, we only lose the wakeup - so
+	; poll the RIRB for codec replies and run the interrupt handler from a
+	; thread. A flag of its own: polling_mode is the timeout recovery state.
+	mov	[ctrl.no_irq], 1
+@@:
+
 	; initialize chip
 	call	azx_init_pci
 
@@ -1210,6 +1244,32 @@ end if
 	call	azx_init_chip
 ;] Asper
 
+	; no IRQ line: let a polling thread stand in for the interrupt.
+	; Started only now - earlier, hda_irq could ack STATESTS in the window
+	; before reset_controller reads it into codec_mask
+	cmp	[ctrl.no_irq], 0
+	je	@f
+	push	esi edi 	; CreateThread clobbers them, START needs esi
+	mov	ebx, 1
+	mov	ecx, hda_poll_thread
+	xor	edx, edx
+	invoke	CreateThread
+	pop	edi esi
+	test	eax, eax
+	jns	.thread_ok
+	; nothing would ever complete a period and playback would stall,
+	; so fail the init rather than register a service that hangs
+	mov	esi, msgNoIrqThread
+	invoke	SysMsgBoardStr
+	xor	eax, eax
+	ret
+.thread_ok:
+if DEBUG
+	mov	esi, msgNoIrqPoll
+	invoke	SysMsgBoardStr
+end if
+@@:
+
 	xor	eax, eax
 	inc	eax
 	ret
@@ -1243,7 +1303,7 @@ endl
 	; clear STATESTS
 	mov	eax, STATESTS_INT_MASK
 	mov	edx, ICH6_REG_STATESTS
-	call	azx_writeb
+	call	azx_writew
 
 	; reset controller
 	mov	edx, ICH6_REG_GCTL
@@ -1454,6 +1514,12 @@ endp
 
 proc azx_alloc_cmd_io
 	push	eax ecx edx
+	; the unsolicited-event ring lives in reserved storage, so its
+	; pointers hold garbage until we clear them here - do it before
+	; reset_controller enables unsolicited responses in GCTL
+	mov	[unsol_events.rp], 0
+	mov	[unsol_events.wp], 0
+
 	; single page (at least 4096 bytes) must suffice for both ringbuffers
 	invoke	KernelAlloc, 4096
 	mov	[ctrl.rb], eax
@@ -1654,14 +1720,24 @@ endl
 	mov	ecx, 1000;+1000
 .next_try:
 	mov	al, [ctrl.polling_mode]
-	test	al, al
+	or	al, [ctrl.no_irq]
 	jnz	.poll
 	mov	ah, [do_poll]
 	test	ah, ah
 	jz	@f
 .poll:
 	spin_lock_irq
+	; take the lock hda_irq holds so two RIRB readers never overlap
+	push	eax ebx
+	mov	eax, aspinlock
+.poll_lock:
+	mov	ebx, SPINLOCK_BUSY
+	lock xchg [eax], ebx
+	cmp	ebx, SPINLOCK_FREE
+	jnz	.poll_lock
+	pop	ebx eax
 	call	azx_update_rirb
+	mov	[aspinlock], SPINLOCK_FREE
 	spin_unlock_irq
 @@:
 	mov	eax, [ctrl.rirb_cmd]
@@ -1942,6 +2018,7 @@ proc azx_int_disable
 	mov	ebx, ICH6_INT_CTRL_EN or ICH6_INT_GLOBAL_EN
 	xor	ebx, -1
 	and	eax, ebx
+	mov	edx, ICH6_REG_INTCTL	; azx_readl turned edx into the absolute address
 	call	azx_writel
 	pop	edx ebx eax
 	ret
@@ -1957,9 +2034,9 @@ proc azx_int_clear
 	call	azx_writeb
 
 	; clear STATESTS
-	mov	al, STATESTS_INT_MASK
+	mov	ax, STATESTS_INT_MASK
 	mov	edx, ICH6_REG_STATESTS
-	call	azx_writeb
+	call	azx_writew
 
 	; clear rirb status
 	mov	al, RIRB_INT_MASK
@@ -2696,6 +2773,7 @@ proc snd_hda_queue_unsol_event stdcall, res:dword, res_ex:dword
     ; only carries EBX across iterations and reloads the rest, and TimerHS
     ; preserves EBX too - so by using scratch regs only we need no saves.
     mov     eax, [unsol_events.wp]
+    and     eax, HDA_UNSOL_QUEUE_SIZE - 1     ; never index the ring wild
     lea     edx, [eax+1]
     and     edx, HDA_UNSOL_QUEUE_SIZE - 1     ; next write pos (size must be 2^n)
     cmp     edx, [unsol_events.rp]
@@ -2998,6 +3076,7 @@ sz_sound_srv	     db 'SOUND',0
 
 msgInit      db 'detect hardware...',13,10,0
 msgFail      db 'device not found',13,10,0
+msgNoIrqThread db 'no IRQ line: cannot create polling thread',13,10,0
 msgAttchIRQ  db 'IRQ line not supported', 13,10,0
 msgInvIRQ    db 'IRQ line not assigned or invalid', 13,10,0
 msgPlay      db 'start play', 13,10,0
@@ -3067,6 +3146,7 @@ if DEBUG
     msgTCSEL		     db 'PCI TCSEL     ',0
     msgTV		     db 'HDA test version ',TEST_VERSION_NUMBER,13,10,0
     msgGCap		     db 'GCAP = ',0
+    msgNoIrqPoll	     db 'no IRQ line: polling thread',13,10,0
 end if
 
 if USE_SINGLE_MODE
@@ -3101,6 +3181,11 @@ end if
 
 ;] Asper
 
+; relocation table lives after all code and initialized data: placed before the
+; code its size feeds back into every relocation site and fasm may never converge
+data fixups
+end data
+
 aspinlock	 dd SPINLOCK_FREE
 
 codec CODEC

drivers/ethernet/i8255x.asm

@@ -130,6 +130,8 @@ RU_STATUS_IDLE          = 0000b shl 2
 RU_STATUS_SUSPENDED     = 0001b shl 2
 RU_STATUS_NO_RESOURCES  = 0010b shl 2
 RU_STATUS_READY         = 0100b shl 2
+SCB_STATUS_CUS          = 11000000b     ; CU Status
+CU_STATUS_ACTIVE        = 10b shl 6
 SCB_STATUS_FCP          = 1 shl 8       ; Flow Control Pause
 SCB_STATUS_SWI          = 1 shl 10      ; Software Interrupt
 SCB_STATUS_MDI          = 1 shl 11      ; MDI read/write complete
@@ -183,6 +185,8 @@ struct  txfd
 
 ends
 
+TXFD_STATUS_C           = 1 shl 15
+
 TXFD_CMD_IA             = 1 shl 0
 TXFD_CMD_CFG            = 1 shl 1
 TXFD_CMD_TX             = 1 shl 2
@@ -628,6 +632,9 @@ reset:
         mov     ax, CU_START or INT_MASK
         out     dx, ax
         call    cmd_wait
+; cmd_wait only means the SCB accepted the command. The Configure below
+; reuses confcmd, so a slow chip would read it instead of our MAC.
+        call    confcmd_wait
 
 ;-------------
 ; Configure CU
@@ -652,6 +659,7 @@ reset:
         mov     ax, CU_START                            ; expect Interrupts from now on
         out     dx, ax
         call    cmd_wait
+        call    confcmd_wait
 
 ; Start media check timer
         mov     [ebx + device.state], ETH_LINK_DOWN
@@ -678,6 +686,11 @@ init_rx_ring:
 ;---------------------
 ; build rxfd structure
 
+        cmp     [ebx + device.rx_desc], 0       ; the RFD of an earlier reset
+        je      @f
+        invoke  NetFree, [ebx + device.rx_desc]
+        mov     [ebx + device.rx_desc], 0
+  @@:
         invoke  NetAlloc, 2000
         test    eax, eax
         jz      .out_of_mem
@@ -689,6 +702,7 @@ init_rx_ring:
         mov     [esi + sizeof.NET_BUFF + rxfd.command], RXFD_CMD_EL or RXFD_CMD_SUSPEND
         mov     [esi + sizeof.NET_BUFF + rxfd.link], eax
         mov     [esi + sizeof.NET_BUFF + rxfd.count], 0
+        mov     [esi + sizeof.NET_BUFF + rxfd.rx_buf_addr], 0xffffffff  ; simplified mode, no RBD
         mov     [esi + sizeof.NET_BUFF + rxfd.size], 1528
 
         ret
@@ -709,6 +723,15 @@ init_tx_ring:
         invoke  GetPhysAddr
         mov     ecx, TX_RING_SIZE
   .next_desc:
+; after a reset with frames still unreclaimed: free them, or transmit sees the
+; descriptor busy (virt_addr) and reports TX overrun forever
+        cmp     [esi + txfd.virt_addr], 0
+        je      @f
+        push    eax ecx
+        invoke  NetFree, [esi + txfd.virt_addr]
+        pop     ecx eax
+        mov     [esi + txfd.virt_addr], 0
+  @@:
         mov     [esi + txfd.status], 0
         mov     [esi + txfd.command], 0
         lea     edx, [eax + txfd.buf_addr]
@@ -763,7 +786,7 @@ proc transmit stdcall bufferptr
         lea     edi, [ebx + device.tx_ring + eax]
 
         ; Check if current descriptor is free or still in use
-        cmp     [edi + txfd.status], 0
+        cmp     [edi + txfd.virt_addr], 0               ; buffer not reclaimed yet
         jne     .overrun
 
         ; Fill in status and command values
@@ -786,6 +809,22 @@ proc transmit stdcall bufferptr
         mov     eax, edi
         invoke  GetPhysAddr
         set_io  [ebx + device.io_addr], 0
+
+        ; CU_START is only valid while the CU is idle or suspended:
+        ; wait for the previous frame to leave (1.2 ms at 10 Mbit)
+        push    eax ecx
+        set_io  [ebx + device.io_addr], REG_SCB_STATUS
+        mov     ecx, 100000
+  .cu_busy:
+        in      al, dx
+        and     al, SCB_STATUS_CUS
+        cmp     al, CU_STATUS_ACTIVE
+        jne     .cu_free
+        dec     ecx
+        jnz     .cu_busy
+        DEBUGF  2, "CU still active, starting anyway\n"
+  .cu_free:
+        pop     ecx eax
         set_io  [ebx + device.io_addr], REG_SCB_PTR
         out     dx, eax
 
@@ -854,7 +893,7 @@ int_handler:
 
         DEBUGF  1,"Status: %x\n", ax
 
-        test    ax, SCB_STATUS_FR               ; did we receive a frame?
+        test    ax, SCB_STATUS_FR or SCB_STATUS_RNR     ; frame received or receiver stopped?
         jz      .no_rx
 
         push    ax
@@ -901,6 +940,7 @@ int_handler:
         mov     [esi + sizeof.NET_BUFF + rxfd.command], RXFD_CMD_EL or RXFD_CMD_SUSPEND
         mov     [esi + sizeof.NET_BUFF + rxfd.link], eax
         mov     [esi + sizeof.NET_BUFF + rxfd.count], 0
+        mov     [esi + sizeof.NET_BUFF + rxfd.rx_buf_addr], 0xffffffff  ; simplified mode, no RBD
         mov     [esi + sizeof.NET_BUFF + rxfd.size], 1528
 
 ; restart RX
@@ -916,11 +956,22 @@ int_handler:
         mov     ax, RX_START
         out     dx, ax
         call    cmd_wait
-  .out_of_mem:
 
 ; Hand the frame over to the kernel
         jmp     [EthInput]
 
+  .out_of_mem:
+; The buffer is still our only RFD: drop the frame and reuse it
+        mov     esi, [esp]                      ; the buffer
+        mov     ecx, [esi + NET_BUFF.length]
+        dec     [ebx + device.packets_rx]       ; counted above, before the allocation
+        sub     dword [ebx + device.bytes_rx], ecx
+        sbb     dword [ebx + device.bytes_rx + 4], 0
+        inc     [ebx + device.packets_rx_drop]
+        add     esp, 12                         ; buffer, .rx_loop, ebx
+        mov     esi, [ebx + device.rx_desc]
+        jmp     .not_ok
+
   .not_ok:
 ; Reset the FD
         mov     [esi + sizeof.NET_BUFF + rxfd.status], 0
@@ -945,6 +996,27 @@ int_handler:
 
   .no_rx_:
         DEBUGF  1, "no more data\n"
+
+; With a single RFD the receiver sits in No Resources/Suspended whenever a
+; frame arrived before we re-armed it. Nothing restarts it but us.
+        set_io  [ebx + device.io_addr], 0
+        set_io  [ebx + device.io_addr], REG_SCB_STATUS
+        in      al, dx
+        and     al, SCB_STATUS_RUS
+        cmp     al, RU_STATUS_READY
+        je      .ru_ready
+        test    byte[esp], 1                    ; bit 0 of the saved status is reserved,
+        jnz     .ru_ready                       ; we mark it: one restart per IRQ
+        or      byte[esp], 1
+        movzx   eax, al
+        DEBUGF  1, "Restarting receiver, RU status %x\n", eax:2
+        mov     esi, [ebx + device.rx_desc]
+        push    ebx
+        test    [esi + sizeof.NET_BUFF + rxfd.status], RXFD_STATUS_C
+        jnz     .rx_loop                        ; a frame completed meanwhile
+        add     esp, 4
+        jmp     .not_ok                         ; re-arm the empty RFD
+  .ru_ready:
         pop     ax
 
   .no_rx:
@@ -955,13 +1027,13 @@ int_handler:
 
         push    eax
   .loop_tx:
-        mov     edi, [ebx + device.last_tx]
-        mov     eax, sizeof.txfd
-        mul     eax
+        mov     eax, [ebx + device.last_tx]
+        mov     edx, sizeof.txfd
+        mul     edx
         lea     edi, [ebx + device.tx_ring + eax]
 
-        cmp     [edi + txfd.status], 0
-        je      .tx_done
+        test    [edi + txfd.status], TXFD_STATUS_C      ; sent by the device?
+        jz      .tx_done
 
         cmp     [edi + txfd.virt_addr], 0
         je      .tx_done
@@ -970,6 +1042,7 @@ int_handler:
 
         push    [edi + txfd.virt_addr]
         mov     [edi + txfd.virt_addr], 0
+        mov     [edi + txfd.status], 0                  ; free for transmit again
         invoke  NetFree
 
         inc     [ebx + device.last_tx]
@@ -980,12 +1053,6 @@ int_handler:
         pop     eax
   .no_tx:
 
-        test    ax, RU_STATUS_NO_RESOURCES
-        jz      .not_out_of_resources
-
-        DEBUGF  2, "Out of resources!\n"
-
-  .not_out_of_resources:
         pop     edi esi ebx
         xor     eax, eax
         inc     eax
@@ -1079,6 +1146,25 @@ proc check_media_mii stdcall dev:dword
 endp
 
 
+; Wait (up to 1 s) until the device has executed the command in confcmd.
+align 4
+confcmd_wait:
+        mov     ecx, 1000                       ; udelay = Sleep(1), 1 ms
+  .loop:
+        test    [ebx + device.confcmd.status], TXFD_STATUS_C
+        jnz     .done
+        call    udelay
+        dec     ecx
+        jnz     .loop
+        movzx   eax, [ebx + device.confcmd.command]
+        DEBUGF  2, "Command 0x%x not completed\n", eax:4
+        ret
+  .done:
+        movzx   eax, [ebx + device.confcmd.command]
+        movzx   ecx, [ebx + device.confcmd.status]
+        DEBUGF  1, "Command 0x%x done, status 0x%x\n", eax:4, ecx:4
+        ret
+
 align 4
 cmd_wait:
 

drivers/i2chid/Tupfile.lua

@@ -0,0 +1,3 @@
+if tup.getconfig("NO_FASM") ~= "" then return end
+ROOT = "../.."
+tup.rule("i2chid.asm", "fasm %f %o " .. tup.getconfig("PESTRIP_CMD") .. tup.getconfig("KPACK_CMD"), "%B.sys")

drivers/i2chid/dwi2c.inc

@@ -0,0 +1,438 @@
+; Synopsys DesignWare I2C master, polled (PIO) mode only.
+; This is the controller inside Intel LPSS blocks (Skylake and newer, plus
+; Atom-derived SoCs); the same IP is used by many other vendors.
+;
+; The register file is always accessed as dwords. The controller must be
+; disabled to change IC_TAR, so every transfer disables it, programs the
+; target address, enables it back and feeds the TX FIFO by hand.
+
+; ------------------------------ Registers ------------------------------------
+DwIcCon             = 0x00
+DwIcTar             = 0x04
+DwIcDataCmd         = 0x10
+DwIcSsSclHcnt       = 0x14
+DwIcSsSclLcnt       = 0x18
+DwIcFsSclHcnt       = 0x1C
+DwIcFsSclLcnt       = 0x20
+DwIcIntrStat        = 0x2C
+DwIcIntrMask        = 0x30
+DwIcRawIntrStat     = 0x34
+DwIcRxTl            = 0x38
+DwIcTxTl            = 0x3C
+DwIcClrIntr         = 0x40
+DwIcClrTxAbrt       = 0x54
+DwIcEnable          = 0x6C
+DwIcStatus          = 0x70
+DwIcTxflr           = 0x74
+DwIcRxflr           = 0x78
+DwIcSdaHold         = 0x7C
+DwIcTxAbrtSource    = 0x80
+DwIcEnableStatus    = 0x9C
+DwIcCompParam1      = 0xF4
+DwIcCompVersion     = 0xF8
+DwIcCompType        = 0xFC
+
+; "DW_apb_i2c" component signature ('DW' + 0x0140).
+DW_COMP_TYPE_VALUE  = 0x44570140
+
+; IC_CON: master, standard speed (100 kHz), restarts allowed, slave disabled.
+DW_CON_MASTER       = 1 shl 0
+DW_CON_SPEED_STD    = 1 shl 1
+DW_CON_SPEED_FAST   = 2 shl 1
+DW_CON_RESTART_EN   = 1 shl 5
+DW_CON_SLAVE_DIS    = 1 shl 6
+
+; IC_DATA_CMD bits above the data byte.
+DW_CMD_READ         = 1 shl 8
+DW_CMD_STOP         = 1 shl 9
+DW_CMD_RESTART      = 1 shl 10
+
+; IC_STATUS bits
+DW_STAT_TFNF        = 1 shl 1   ; TX FIFO not full
+DW_STAT_TFE         = 1 shl 2   ; TX FIFO empty
+DW_STAT_RFNE        = 1 shl 3   ; RX FIFO not empty
+DW_STAT_MST_ACT     = 1 shl 5   ; master FSM busy
+
+; IC_RAW_INTR_STAT bits
+DW_INT_TX_ABRT      = 1 shl 6
+
+; SCL timing for standard mode. The LPSS input clock is 100 MHz on the PCH
+; parts and 133 MHz on the Atom-derived SoCs (Apollo Lake, Gemini Lake), and
+; nothing in PCI config space says which one this is, so the counts are
+; picked to satisfy the fast-mode minimums at the fastest of them: at
+; 133 MHz that is tHIGH = 0.98 us and tLOW = 1.5 us, 400 kHz. On a 100 MHz
+; part the same counts run the bus at 300 kHz, which is just as legal.
+; Fast mode rather than standard: HID over I2C requires every device to
+; support it, that is what Windows drives them at, and a polled host pays
+; for every byte with busy-waiting - at 100 kHz a 64-byte input read costs
+; 6 ms, more than a fast pad takes to produce the next report, and the
+; backlog that builds up plays back as the cursor coasting after the finger
+; has stopped. SDA hold stays well below tLOW.
+DW_FS_HCNT          = 130
+DW_FS_LCNT          = 200
+DW_SDA_HOLD         = 28
+; The AMD FCH feeds its DesignWare masters from a 150 MHz reference, so
+; the same minimums need more counts there: 145/225 give tHIGH = 0.97 us,
+; tLOW = 1.5 us, again 400 kHz; SDA hold 45 = 300 ns.
+DW_AMD_FS_HCNT      = 145
+DW_AMD_FS_LCNT      = 225
+DW_AMD_SDA_HOLD     = 45
+
+; A transfer gives up after [XferTimeout] hundredths of a second. The init
+; value must cover a full report descriptor read: descriptors run up to 2 KB
+; and the bus may be as slow as 77 kHz, so at 9 SCL cycles per byte a single
+; transfer takes ~240 ms of wall-clock time. Polling is generous too now that it runs in its own thread
+; rather than in a timer callback - a wedged bus costs this thread its slice
+; and nothing else.
+DW_TIMEOUT_INIT     = 60
+DW_TIMEOUT_POLL     = 20
+
+; Disables the controller and waits until it reports itself disabled.
+; in: ebx = MMIO base. Clobbers eax, ecx.
+proc dw_ctrl_disable
+        mov     dword [ebx+DwIcEnable], 0
+        mov     ecx, 100000
+@@:
+        test    dword [ebx+DwIcEnableStatus], 1
+        jz      .done
+        pause
+        loop    @b
+.done:
+        ret
+endp
+
+; Aborts a stuck transfer cleanly: the ABORT bit makes the master issue STOP,
+; flush the TX FIFO and raise TX_ABRT, which releases a slave that is holding
+; the bus (a clock-stretching touchpad, for instance). Simply disabling the
+; controller in the middle of a byte would leave SDA held low and every
+; following transfer would time out too. Older cores without the ABORT bit
+; just ignore the write and fall through after the wait.
+; in: ebx = MMIO base. Clobbers eax, ecx.
+proc dw_ctrl_abort
+        mov     dword [ebx+DwIcEnable], 3
+        mov     ecx, 100000
+@@:
+        test    dword [ebx+DwIcRawIntrStat], DW_INT_TX_ABRT
+        jnz     @f
+        pause
+        loop    @b
+@@:
+        mov     eax, [ebx+DwIcClrTxAbrt]
+        call    dw_ctrl_disable
+        ret
+endp
+
+; One idle step of a transfer wait loop.
+; A healthy transfer is over within a few milliseconds, well inside the
+; first two timer ticks, and spins through them: a transfer that yields
+; its slice is at the scheduler's mercy for the rest of it. Past that the
+; bus is stuck or crawling, and the wait gives up the CPU a tick at a time
+; instead of burning a core until XferTimeout.
+; in: eax = ticks elapsed since the transfer began. Preserves all but eax.
+DW_SPIN_TICKS       = 2
+
+proc dw_wait_step
+        cmp     eax, DW_SPIN_TICKS
+        jae     .yield
+        pause
+        ret
+.yield:
+        push    ebx
+        movi    ebx, 1
+        invoke  Delay                   ; preserves every register
+        pop     ebx
+        ret
+endp
+
+; One-time controller setup after mapping.
+; in: ebx = MMIO base.
+proc dw_ctrl_init
+        push    eax ecx
+        call    dw_ctrl_disable
+        mov     dword [ebx+DwIcCon], DW_CON_MASTER + DW_CON_SPEED_FAST + \
+                DW_CON_RESTART_EN + DW_CON_SLAVE_DIS
+; The counts come from variables: they depend on the reference clock, which
+; is decided by whoever found the controller (PCI LPSS or AMD FCH). The
+; standard-mode registers get the same values scaled up so that they are
+; sane should anything ever switch the speed bits back.
+        mov     eax, [DwHcnt]
+        mov     [ebx+DwIcFsSclHcnt], eax
+        lea     eax, [eax*4]
+        mov     [ebx+DwIcSsSclHcnt], eax
+        mov     eax, [DwLcnt]
+        mov     [ebx+DwIcFsSclLcnt], eax
+        lea     eax, [eax*4]
+        mov     [ebx+DwIcSsSclLcnt], eax
+        mov     eax, [DwSdaHold]
+        mov     [ebx+DwIcSdaHold], eax
+        mov     dword [ebx+DwIcIntrMask], 0
+        mov     dword [ebx+DwIcRxTl], 0
+        mov     dword [ebx+DwIcTxTl], 0
+        mov     eax, [ebx+DwIcClrIntr]
+        pop     ecx eax
+        ret
+endp
+
+; One combined I2C transaction: optional write phase, then optional read phase
+; with a repeated START between them. Fully polled.
+; out: eax = 0 on success, IC_TX_ABRT_SOURCE value on NAK/abort, -1 on timeout.
+proc dw_xfer stdcall uses ebx esi edi, addr:dword, wbuf:dword, wlen:dword, \
+        rbuf:dword, rlen:dword
+locals
+        wsent   dd ?
+        rqueued dd ?
+        rgot    dd ?
+        t0      dd ?
+endl
+        mov     ebx, [I2cMmio]
+        call    dw_ctrl_disable
+        mov     eax, [addr]
+        mov     [ebx+DwIcTar], eax
+        mov     eax, [ebx+DwIcClrIntr]
+        mov     dword [ebx+DwIcEnable], 1
+        xor     eax, eax
+        mov     [wsent], eax
+        mov     [rqueued], eax
+        mov     [rgot], eax
+        invoke  GetTimerTicks
+        mov     [t0], eax
+.loop:
+; Abort (NAK and friends) ends the transfer immediately.
+        test    dword [ebx+DwIcRawIntrStat], DW_INT_TX_ABRT
+        jnz     .abort
+; Drain the RX FIFO first so it can never overflow.
+        test    dword [ebx+DwIcStatus], DW_STAT_RFNE
+        jz      .no_rx
+        mov     edx, [rgot]
+        cmp     edx, [rlen]
+        jae     .no_rx
+        mov     eax, [ebx+DwIcDataCmd]
+        mov     edi, [rbuf]
+        mov     [edi+edx], al
+        inc     edx
+        mov     [rgot], edx
+        jmp     .loop
+.no_rx:
+; Feed the write phase.
+        mov     eax, [wsent]
+        cmp     eax, [wlen]
+        jae     .push_reads
+        test    dword [ebx+DwIcStatus], DW_STAT_TFNF
+        jz      .check_done
+        mov     esi, [wbuf]
+        movzx   edx, byte [esi+eax]
+        inc     eax
+        cmp     eax, [wlen]
+        jne     @f
+        cmp     [rlen], 0
+        jne     @f
+        or      edx, DW_CMD_STOP        ; last byte of a write-only transfer
+@@:
+        mov     [ebx+DwIcDataCmd], edx
+        mov     [wsent], eax
+        jmp     .loop
+.push_reads:
+; Feed read commands, at most 8 outstanding so the RX FIFO keeps up.
+        mov     eax, [rqueued]
+        cmp     eax, [rlen]
+        jae     .check_done
+        mov     edx, eax
+        sub     edx, [rgot]
+        cmp     edx, 8
+        jae     .check_done
+        test    dword [ebx+DwIcStatus], DW_STAT_TFNF
+        jz      .check_done
+        mov     edx, DW_CMD_READ
+        test    eax, eax
+        jnz     @f
+        cmp     [wlen], 0
+        je      @f
+        or      edx, DW_CMD_RESTART     ; repeated START after the write phase
+@@:
+        mov     ecx, [rlen]
+        dec     ecx
+        cmp     eax, ecx
+        jne     @f
+        or      edx, DW_CMD_STOP
+@@:
+        mov     [ebx+DwIcDataCmd], edx
+        inc     eax
+        mov     [rqueued], eax
+        jmp     .loop
+.check_done:
+        mov     eax, [rgot]
+        cmp     eax, [rlen]
+        jne     .not_done
+        mov     eax, [wsent]
+        cmp     eax, [wlen]
+        jne     .not_done
+        mov     eax, [ebx+DwIcStatus]
+        test    eax, DW_STAT_TFE
+        jz      .not_done
+        test    eax, DW_STAT_MST_ACT
+        jnz     .not_done
+        call    dw_ctrl_disable
+        xor     eax, eax
+        ret
+.not_done:
+        invoke  GetTimerTicks
+        sub     eax, [t0]
+        cmp     eax, [XferTimeout]
+        jae     .timeout
+        call    dw_wait_step
+        jmp     .loop
+.abort:
+        mov     esi, [ebx+DwIcTxAbrtSource]
+        mov     eax, [ebx+DwIcClrTxAbrt]
+        call    dw_ctrl_disable
+        mov     eax, esi
+        test    eax, eax
+        jnz     @f
+        or      eax, 0x80000000
+@@:
+        ret
+.timeout:
+        call    dw_ctrl_abort
+        or      eax, -1
+        ret
+endp
+
+; Reads one HID input report as a single continuous bus transaction, the way
+; interrupt-driven hosts (Windows, Linux) do it: the optional register-address
+; write and every read command are queued into the TX FIFO before the bytes
+; start flowing, so nothing on the wire depends on how fast the host loop
+; runs. The previous adaptive read (fetch 2 length bytes, decide, fetch the
+; rest, stop short) was cut down on the wire to 3 bytes for an empty poll -
+; and the SIPODEV/Hantick touchpad firmwares are exactly the kind of thing
+; that wedges when an input read is shorter than the frame they prepared.
+; rlen should not exceed the TX FIFO depth (IC_COMP_PARAM_1 bits 23:16, plus
+; one) or the queue loop degrades to feeding the FIFO on the fly, which
+; reintroduces pauses.
+; in: addr = slave, wlen bytes at wbuf written first (0 = plain read),
+;     rlen bytes read into buf.
+; out: eax = 0 ok, IC_TX_ABRT_SOURCE on NAK/abort, -1 on timeout.
+; Timeout check for the dw_read_block wait spots; expands inside the proc so
+; [t0] and [XferTimeout] resolve in the caller's own frame.
+macro BLOCK_TICK fail_label {
+        invoke  GetTimerTicks
+        sub     eax, [t0]
+        cmp     eax, [XferTimeout]
+        jae     fail_label
+        call    dw_wait_step
+}
+
+proc dw_read_block stdcall uses ebx esi edi, addr:dword, wbuf:dword, wlen:dword,         buf:dword, rlen:dword
+locals
+        rgot    dd ?
+        t0      dd ?
+endl
+        mov     ebx, [I2cMmio]
+        call    dw_ctrl_disable
+        mov     eax, [addr]
+        mov     [ebx+DwIcTar], eax
+        mov     eax, [ebx+DwIcClrIntr]
+        mov     dword [ebx+DwIcEnable], 1
+        invoke  GetTimerTicks
+        mov     [t0], eax
+        mov     [rgot], 0
+; Write phase: the register address bytes, no STOP - the read continues the
+; same transaction after a repeated START.
+        xor     ecx, ecx
+.wr:
+        cmp     ecx, [wlen]
+        jae     .queue_reads
+        test    dword [ebx+DwIcRawIntrStat], DW_INT_TX_ABRT
+        jnz     .abort
+        test    dword [ebx+DwIcStatus], DW_STAT_TFNF
+        jz      .wr_wait
+        mov     esi, [wbuf]
+        movzx   edx, byte [esi+ecx]
+        mov     [ebx+DwIcDataCmd], edx
+        inc     ecx
+        jmp     .wr
+.wr_wait:
+        BLOCK_TICK .timeout
+        jmp     .wr
+.queue_reads:
+        xor     ecx, ecx
+.rq:
+        cmp     ecx, [rlen]
+        jae     .drain
+        test    dword [ebx+DwIcRawIntrStat], DW_INT_TX_ABRT
+        jnz     .abort
+        test    dword [ebx+DwIcStatus], DW_STAT_TFNF
+        jz      .rq_wait
+        mov     edx, DW_CMD_READ
+        test    ecx, ecx
+        jnz     .not_first
+        cmp     [wlen], 0
+        je      .not_first
+        or      edx, DW_CMD_RESTART
+.not_first:
+        mov     eax, [rlen]
+        dec     eax
+        cmp     ecx, eax
+        jne     @f
+        or      edx, DW_CMD_STOP
+@@:
+        mov     [ebx+DwIcDataCmd], edx
+        inc     ecx
+; Drain opportunistically so a FIFO shallower than rlen cannot deadlock the
+; queue loop (the RX side fills while the TX side is full).
+        test    dword [ebx+DwIcStatus], DW_STAT_RFNE
+        jz      .rq
+        call    .take_byte
+        jmp     .rq
+.rq_wait:
+        BLOCK_TICK .timeout
+        jmp     .rq
+.drain:
+        test    dword [ebx+DwIcRawIntrStat], DW_INT_TX_ABRT
+        jnz     .abort
+        test    dword [ebx+DwIcStatus], DW_STAT_RFNE
+        jz      .chk
+        call    .take_byte
+        jmp     .drain
+; Moves one byte from the RX FIFO into the caller's buffer. Exactly rlen
+; reads are queued into freshly flushed FIFOs, so a byte beyond rlen should
+; never arrive; if one does anyway it is still taken off the FIFO, so the
+; drain loop cannot stall on it, but it is dropped rather than written
+; past the end of the buffer.
+.take_byte:
+        mov     edx, [rgot]
+        mov     eax, [ebx+DwIcDataCmd]
+        cmp     edx, [rlen]
+        jae     @f
+        mov     edi, [buf]
+        mov     [edi+edx], al
+        inc     edx
+        mov     [rgot], edx
+@@:
+        retn
+.chk:
+        mov     eax, [rgot]
+        cmp     eax, [rlen]
+        jne     .wait
+        test    dword [ebx+DwIcStatus], DW_STAT_MST_ACT
+        jnz     .wait
+        call    dw_ctrl_disable
+        xor     eax, eax
+        ret
+.wait:
+        BLOCK_TICK .timeout
+        jmp     .drain
+.abort:
+        mov     esi, [ebx+DwIcTxAbrtSource]
+        mov     eax, [ebx+DwIcClrTxAbrt]
+        call    dw_ctrl_disable
+        mov     eax, esi
+        test    eax, eax
+        jnz     @f
+        or      eax, 0x80000000
+@@:
+        ret
+.timeout:
+        call    dw_ctrl_abort
+        or      eax, -1
+        ret
+endp

drivers/i2chid/hidmin.inc

@@ -0,0 +1,646 @@
+; Minimal HID report descriptor parser.
+;
+; The goal is not a full HID implementation: the parser walks the descriptor
+; once and remembers the bit layout of the first input report that contains
+; relative X and relative Y (Generic Desktop / X, Y) - i.e. the mouse
+; compatibility collection that every I2C mouse and every Windows Precision
+; Touchpad exposes. Buttons (Usage Page 9) and a relative wheel found in the
+; same report are picked up too. Absolute (digitizer) reports are advanced
+; over but otherwise ignored.
+;
+; Report IDs are honoured: bit positions are counted separately per report ID
+; and the chosen ID is checked in every incoming packet.
+;
+; Alongside the mouse report, the parser also maps the touchpad (Windows
+; Precision Touchpad) report: the report that carries Tip Switch usages.
+; Each Tip Switch opens a new finger block; the absolute X and Y that follow
+; it belong to that finger; Contact Count and the Button-page fields of the
+; same report are recorded too. Once the pad is switched into touchpad mode
+; those offsets are what the frame decoder reads - every vendor lays the
+; frame out differently, and only the descriptor knows how.
+HID_MAX_FINGERS = 5
+
+HID_MAX_USAGES  = 16
+HID_MAX_REPIDS  = 16
+
+; Returns the pointer to the input bit counter of the current report ID,
+; allocating a new slot on first use.
+; out: ebx -> dword bit counter. Clobbers eax, ecx.
+proc hid_get_bits_slot
+        mov     eax, [p_report_id]
+        xor     ecx, ecx
+.scan:
+        cmp     ecx, [p_idcnt]
+        jae     .add
+        cmp     eax, [p_ids+ecx*4]
+        jz      .found
+        inc     ecx
+        jmp     .scan
+.add:
+        cmp     ecx, HID_MAX_REPIDS
+        jb      @f
+        dec     ecx                     ; table full: reuse the last slot
+        jmp     .found
+@@:
+        mov     [p_ids+ecx*4], eax
+        mov     dword [p_bits+ecx*4], 0
+        mov     dword [p_fbits+ecx*4], 0
+        inc     [p_idcnt]
+.found:
+        lea     ebx, [p_bits+ecx*4]
+        ret
+endp
+
+; Handles one Input main item. v_flags = item data (bit 2 = relative).
+proc hid_process_input
+        push    eax ebx ecx edx
+        call    hid_get_bits_slot
+        mov     eax, [ebx]
+        mov     [v_base], eax
+; Advance the counter over this field right away.
+        mov     eax, [p_report_size]
+        imul    eax, [p_report_count]
+        add     [ebx], eax
+; Remember the largest logical maximum ever attached to Generic Desktop X
+; and Y. The mouse collection declares a tiny relative range there while
+; the touchpad collection declares the real surface extents, so the
+; maximum over the whole descriptor is the size of the pad - which is what
+; the button zones are measured against, and it differs per model.
+        cmp     [p_usage_cnt], 0
+        je      .no_extent
+        xor     ecx, ecx
+.ext_loop:
+        cmp     ecx, [p_usage_cnt]
+        jae     .no_extent
+        mov     eax, [p_usages+ecx*4]
+        cmp     eax, 0x00010030
+        jne     .ext_y
+        mov     eax, [p_log_max]
+        cmp     eax, [t_padmaxx]
+        jbe     .ext_next
+        mov     [t_padmaxx], eax
+        jmp     .ext_next
+.ext_y:
+        cmp     eax, 0x00010031
+        jne     .ext_next
+        mov     eax, [p_log_max]
+        cmp     eax, [t_padmaxy]
+        jbe     .ext_next
+        mov     [t_padmaxy], eax
+.ext_next:
+        inc     ecx
+        jmp     .ext_loop
+.no_extent:
+; Button-style field: Usage Minimum/Maximum on the Button page.
+        cmp     [p_usage_min], -1
+        je      .not_buttons
+        cmp     [p_usage_max], -1
+        je      .not_buttons
+        cmp     [p_usage_page], 9
+        jne     .not_buttons
+        cmp     [p_report_size], 1      ; buttons are always 1-bit fields
+        jne     .done
+        mov     eax, [v_base]
+        mov     ecx, [p_report_count]
+        call    hid_note_buttons
+        jmp     .done
+.not_buttons:
+        cmp     [p_usage_cnt], 0
+        je      .done
+        call    hid_note_touchpad
+; Usage-list field: look for GD/X, GD/Y, GD/Wheel among the usages.
+        test    [v_flags], 4            ; only relative fields are interesting
+        jz      .done
+        xor     ecx, ecx
+.usage_loop:
+        cmp     ecx, [p_report_count]
+        jae     .done
+; usage index = min(field index, usage count - 1): a short usage list
+; applies its last usage to the remaining fields.
+        mov     edx, [p_usage_cnt]
+        dec     edx
+        mov     eax, ecx
+        cmp     eax, edx
+        jbe     @f
+        mov     eax, edx
+@@:
+        mov     eax, [p_usages+eax*4]
+        cmp     eax, 0x00010030
+        je      .is_x
+        cmp     eax, 0x00010031
+        je      .is_y
+        cmp     eax, 0x00010038
+        je      .is_wheel
+.next_usage:
+        inc     ecx
+        jmp     .usage_loop
+.is_x:
+        test    [t_have], 1
+        jnz     .next_usage
+        call    hid_claim_report_id
+        jc      .next_usage
+        call    hid_field_offset
+        mov     [t_xoff], eax
+        mov     eax, [p_report_size]
+        mov     [t_xsize], eax
+        or      [t_have], 1
+        jmp     .next_usage
+.is_y:
+        test    [t_have], 2
+        jnz     .next_usage
+        call    hid_claim_report_id
+        jc      .next_usage
+        call    hid_field_offset
+        mov     [t_yoff], eax
+        mov     eax, [p_report_size]
+        mov     [t_ysize], eax
+        or      [t_have], 2
+        jmp     .next_usage
+.is_wheel:
+        test    [t_have], 8
+        jnz     .next_usage
+        mov     eax, [t_id]
+        cmp     eax, -1                 ; wheel never claims a report on its own
+        je      .next_usage
+        cmp     eax, [p_report_id]
+        jne     .next_usage
+        call    hid_field_offset
+        mov     [t_woff], eax
+        mov     eax, [p_report_size]
+        mov     [t_wsize], eax
+        or      [t_have], 8
+        jmp     .next_usage
+.done:
+        pop     edx ecx ebx eax
+        ret
+endp
+
+; Records a run of 1-bit Button-page fields starting at bit eax, ecx of them.
+; The mouse decoder wants the buttons of ITS report: an item from some other
+; report is kept only until a better one turns up, and .eof drops it if it
+; still does not match. The touchpad report keeps its own copy.
+proc hid_note_buttons
+        push    eax ecx edx
+        cmp     ecx, 8
+        jbe     @f
+        mov     ecx, 8
+@@:
+        mov     edx, [p_report_id]
+        cmp     edx, [t_ptp_id]
+        jne     .mouse
+        cmp     [t_pbtn_cnt], 0
+        jne     .mouse
+        mov     [t_pbtn_off], eax
+        mov     [t_pbtn_cnt], ecx
+.mouse:
+        cmp     [t_btncnt], 0
+        je      .take
+        mov     edx, [t_btnid]
+        cmp     edx, [t_id]             ; already the chosen report's: keep
+        je      .done
+        mov     edx, [p_report_id]
+        cmp     edx, [t_btnid]          ; a second run in the same report: keep the first
+        je      .done
+.take:
+        mov     [t_btnoff], eax
+        mov     [t_btncnt], ecx
+        mov     eax, [p_report_id]
+        mov     [t_btnid], eax
+.done:
+        pop     edx ecx eax
+        ret
+endp
+
+; Walks the usages of the current input item looking for the touchpad
+; fields. Field k takes usage min(k, count-1), as in the mouse search.
+proc hid_note_touchpad
+        push    eax ebx ecx edx
+        xor     ecx, ecx
+.field:
+        cmp     ecx, [p_report_count]
+        jae     .done
+        mov     edx, [p_usage_cnt]
+        dec     edx
+        mov     eax, ecx
+        cmp     eax, edx
+        jbe     @f
+        mov     eax, edx
+@@:
+        mov     ebx, [p_usages+eax*4]   ; ebx = usage of field ecx
+        call    hid_field_offset        ; eax = its bit offset
+        cmp     ebx, 0x000D0042         ; Tip Switch: a new finger
+        je      .tip
+        cmp     ebx, 0x000D0054         ; Contact Count
+        je      .count
+        cmp     ebx, 0x000D0051         ; Contact Identifier
+        je      .cid
+        cmp     ebx, 0x00010030
+        je      .x
+        cmp     ebx, 0x00010031
+        je      .y
+        shr     ebx, 16
+        cmp     ebx, 9                  ; buttons written as a usage list
+        jne     .next
+        push    ecx
+        neg     ecx
+        add     ecx, [p_report_count]   ; the fields from here to the end
+        call    hid_note_buttons
+        pop     ecx
+        jmp     .done                   ; one run per item is enough
+.tip:
+        cmp     [t_ptp_id], -1
+        jne     @f
+        mov     edx, [p_report_id]
+        mov     [t_ptp_id], edx         ; the first report with fingers is the one
+@@:
+        call    hid_in_ptp
+        jne     .next
+        mov     edx, [t_fcount]
+        cmp     edx, HID_MAX_FINGERS
+        jae     .next
+        mov     [t_ftip+edx*4], eax
+        mov     dword [t_fxs+edx*4], 0
+        mov     dword [t_fys+edx*4], 0
+        mov     dword [t_fcids+edx*4], 0
+        inc     [t_fcount]
+        jmp     .next
+.cid:
+        call    hid_in_ptp
+        jne     .next
+        mov     edx, [t_fcount]
+        test    edx, edx
+        jz      .next
+        dec     edx
+        cmp     dword [t_fcids+edx*4], 0
+        jne     .next
+        mov     [t_fcid+edx*4], eax
+        mov     eax, [p_report_size]
+        mov     [t_fcids+edx*4], eax
+        jmp     .next
+.count:
+        call    hid_in_ptp
+        jne     .next
+        mov     [t_ccnt_off], eax
+        mov     eax, [p_report_size]
+        mov     [t_ccnt_size], eax
+        jmp     .next
+.x:
+        test    [v_flags], 4            ; absolute only; relative is the mouse
+        jnz     .next
+        call    hid_in_ptp
+        jne     .next
+        mov     edx, [t_fcount]
+        test    edx, edx
+        jz      .next
+        dec     edx
+        cmp     dword [t_fxs+edx*4], 0
+        jne     .next
+        mov     [t_fx+edx*4], eax
+        mov     eax, [p_report_size]
+        mov     [t_fxs+edx*4], eax
+        jmp     .next
+.y:
+        test    [v_flags], 4
+        jnz     .next
+        call    hid_in_ptp
+        jne     .next
+        mov     edx, [t_fcount]
+        test    edx, edx
+        jz      .next
+        dec     edx
+        cmp     dword [t_fys+edx*4], 0
+        jne     .next
+        mov     [t_fy+edx*4], eax
+        mov     eax, [p_report_size]
+        mov     [t_fys+edx*4], eax
+.next:
+        inc     ecx
+        jmp     .field
+.done:
+        pop     edx ecx ebx eax
+        ret
+endp
+
+; ZF set when the current item belongs to the touchpad report.
+proc hid_in_ptp
+        push    eax
+        mov     eax, [p_report_id]
+        cmp     eax, [t_ptp_id]
+        pop     eax
+        ret
+endp
+
+; Size in bytes of the feature report with id eax; 0 when the descriptor
+; never mentioned one.
+proc hid_feature_bytes
+        push    ecx
+        xor     ecx, ecx
+@@:
+        cmp     ecx, [p_idcnt]
+        jae     .none
+        cmp     eax, [p_ids+ecx*4]
+        je      .found
+        inc     ecx
+        jmp     @b
+.found:
+        mov     eax, [p_fbits+ecx*4]
+        add     eax, 7
+        shr     eax, 3
+        pop     ecx
+        ret
+.none:
+        xor     eax, eax
+        pop     ecx
+        ret
+endp
+
+; Locks the candidate report ID to the current one, or fails (CF=1) when a
+; different report already holds X/Y.
+proc hid_claim_report_id
+        mov     eax, [t_id]
+        cmp     eax, -1
+        jne     @f
+        mov     eax, [p_report_id]
+        mov     [t_id], eax
+        clc
+        ret
+@@:
+        cmp     eax, [p_report_id]
+        jne     .taken
+        clc
+        ret
+.taken:
+        stc
+        ret
+endp
+
+; Bit offset of field #ecx of the current item.
+; in: ecx = field index. out: eax = v_base + ecx * report_size.
+proc hid_field_offset
+        mov     eax, [p_report_size]
+        imul    eax, ecx
+        add     eax, [v_base]
+        ret
+endp
+
+; Parses the report descriptor.
+; in: esi -> descriptor, ecx = length.
+; out: eax = 1 and f_* variables filled if a usable report was found.
+proc hid_parse_report_desc uses ebx esi edi
+        lea     edi, [esi+ecx]
+        xor     eax, eax
+        mov     [p_usage_page], eax
+        mov     [p_report_size], eax
+        mov     [p_report_count], eax
+        mov     [p_report_id], eax
+        mov     [p_uses_repid], eax
+        mov     [p_usage_cnt], eax
+        mov     [p_idcnt], eax
+        mov     [p_log_max], eax
+        mov     [t_padmaxx], eax
+        mov     [t_padmaxy], eax
+        mov     [t_have], eax
+        mov     [t_btncnt], eax
+        mov     [t_wsize], eax
+        mov     [p_usage_min], -1
+        mov     [p_usage_max], -1
+        mov     [t_id], -1
+        mov     [t_btnid], -2
+        mov     [t_ptp_id], -1
+        mov     [t_fcount], eax
+        mov     [t_ccnt_size], eax
+        mov     [t_pbtn_cnt], eax
+        mov     [t_ptp_bytes], eax
+.loop:
+        cmp     esi, edi
+        jae     .eof
+        movzx   eax, byte [esi]
+        inc     esi
+        cmp     al, 0xFE
+        je      .long_item
+        mov     ecx, eax
+        and     ecx, 3
+        cmp     ecx, 3
+        jne     @f
+        inc     ecx                     ; size code 3 means 4 bytes
+@@:
+        lea     ebx, [esi+ecx]
+        cmp     ebx, edi
+        ja      .eof
+        xor     edx, edx
+        jecxz   .no_data
+        cmp     ecx, 1
+        jne     @f
+        movzx   edx, byte [esi]
+        jmp     .have_data
+@@:
+        cmp     ecx, 2
+        jne     @f
+        movzx   edx, word [esi]
+        jmp     .have_data
+@@:
+        mov     edx, [esi]
+.have_data:
+        add     esi, ecx
+.no_data:
+        and     eax, 0xFC
+; Global items
+        cmp     al, 0x04
+        je      .g_usage_page
+        cmp     al, 0x84
+        je      .g_report_id
+        cmp     al, 0x74
+        je      .g_report_size
+        cmp     al, 0x94
+        je      .g_report_count
+        cmp     al, 0x24
+        je      .g_log_max
+; Local items
+        cmp     al, 0x08
+        je      .l_usage
+        cmp     al, 0x18
+        je      .l_usage_min
+        cmp     al, 0x28
+        je      .l_usage_max
+; Main items
+        cmp     al, 0x80
+        je      .m_input
+        cmp     al, 0x90
+        je      .m_clear_locals
+        cmp     al, 0xB0
+        je      .m_feature
+        cmp     al, 0xA0
+        je      .m_clear_locals
+        cmp     al, 0xC0
+        je      .m_clear_locals
+        jmp     .loop
+.g_usage_page:
+        and     edx, 0xFFFF
+        mov     [p_usage_page], edx
+        jmp     .loop
+.g_report_id:
+        mov     [p_report_id], edx
+        mov     [p_uses_repid], 1
+        jmp     .loop
+.g_report_size:
+        mov     [p_report_size], edx
+        jmp     .loop
+.g_report_count:
+        mov     [p_report_count], edx
+        jmp     .loop
+.g_log_max:
+        mov     [p_log_max], edx
+        jmp     .loop
+.l_usage:
+        mov     ebx, [p_usage_cnt]
+        cmp     ebx, HID_MAX_USAGES
+        jae     .loop
+        cmp     ecx, 4                  ; 4-byte usage carries its own page
+        je      @f
+        mov     eax, [p_usage_page]
+        shl     eax, 16
+        or      edx, eax
+@@:
+        mov     [p_usages+ebx*4], edx
+        inc     [p_usage_cnt]
+        jmp     .loop
+.l_usage_min:
+        mov     [p_usage_min], edx
+        jmp     .loop
+.l_usage_max:
+        mov     [p_usage_max], edx
+        jmp     .loop
+.m_input:
+        mov     [v_flags], edx
+        call    hid_process_input
+        jmp     .m_clear_locals         ; input bits are not feature bits
+.m_feature:
+; Feature reports are only ever written, never decoded, but writing one
+; needs its exact size: a SET_REPORT shorter than the report is ignored by
+; some firmwares (the Elan Input Mode report has two bytes, mode and the
+; selective-reporting switches - one byte does nothing there).
+        call    hid_get_bits_slot
+        mov     eax, [p_report_size]
+        imul    eax, [p_report_count]
+        add     [p_fbits+ecx*4], eax
+.m_clear_locals:
+        xor     eax, eax
+        mov     [p_usage_cnt], eax
+        mov     [p_usage_min], -1
+        mov     [p_usage_max], -1
+        jmp     .loop
+.long_item:
+        cmp     esi, edi
+        jae     .eof
+        movzx   ecx, byte [esi]
+        lea     esi, [esi+ecx+2]
+        jmp     .loop
+.eof:
+; Size of the touchpad report, when there is one.
+        mov     eax, [t_ptp_id]
+        cmp     eax, -1
+        je      .no_ptp
+        xor     ecx, ecx
+@@:
+        cmp     ecx, [p_idcnt]
+        jae     .no_ptp
+        cmp     eax, [p_ids+ecx*4]
+        je      @f
+        inc     ecx
+        jmp     @b
+@@:
+        mov     eax, [p_bits+ecx*4]
+        add     eax, 7
+        shr     eax, 3
+        mov     [t_ptp_bytes], eax
+.no_ptp:
+; Success needs relative X and Y in the same report - or, failing that, a
+; touchpad report with at least one complete finger, which is then the only
+; thing the device will be driven through.
+        mov     eax, [t_have]
+        and     eax, 3
+        cmp     eax, 3
+        je      .have_mouse
+        cmp     [t_fcount], 0
+        je      .fail
+        cmp     dword [t_fxs], 0
+        je      .fail
+        cmp     dword [t_fys], 0
+        je      .fail
+        mov     [f_repid], -1           ; no mouse report to match
+        mov     [f_bytes], 0
+        mov     [t_btncnt], 0
+        movi    eax, 1
+        ret
+.have_mouse:
+; Keep buttons/wheel only when they live in the chosen report.
+        mov     eax, [t_btnid]
+        cmp     eax, [t_id]
+        je      @f
+        mov     [t_btncnt], 0
+@@:
+; Final layout.
+        mov     eax, [t_id]
+        mov     [f_repid], eax
+        xor     ecx, ecx
+@@:
+        cmp     ecx, [p_idcnt]
+        jae     .no_bits
+        cmp     eax, [p_ids+ecx*4]
+        je      .got_bits
+        inc     ecx
+        jmp     @b
+.got_bits:
+        mov     eax, [p_bits+ecx*4]
+        add     eax, 7
+        shr     eax, 3
+        mov     [f_bytes], eax
+        movi    eax, 1
+        ret
+.no_bits:
+.fail:
+        xor     eax, eax
+        ret
+endp
+
+; Extracts an unsigned field from an input report.
+; in: esi -> report data, eax = bit offset, ecx = bit count (1..24).
+; out: eax = value.
+proc hid_extract_u
+        push    ebx ecx edx
+        mov     edx, eax
+        shr     edx, 3
+        mov     ebx, [esi+edx]          ; buffer has spare bytes at the end
+        and     eax, 7
+        push    ecx
+        mov     ecx, eax
+        shr     ebx, cl
+        pop     ecx
+        neg     ecx
+        add     ecx, 32
+        shl     ebx, cl
+        shr     ebx, cl
+        mov     eax, ebx
+        pop     edx ecx ebx
+        ret
+endp
+
+; Same, sign-extended.
+proc hid_extract_s
+        push    ebx ecx edx
+        mov     edx, eax
+        shr     edx, 3
+        mov     ebx, [esi+edx]
+        and     eax, 7
+        push    ecx
+        mov     ecx, eax
+        shr     ebx, cl
+        pop     ecx
+        neg     ecx
+        add     ecx, 32
+        shl     ebx, cl
+        sar     ebx, cl
+        mov     eax, ebx
+        pop     edx ecx ebx
+        ret
+endp

drivers/i2chid/system.ini

@@ -0,0 +1,26 @@
+[touchpad]
+; pointer speed in percent of the driver's default (on top of the [mouse]
+; speed above; pads differ in resolution)
+speed=100
+; how far a finger must move before the pointer follows, in pad units;
+; larger is steadier under a pressing finger, 0 turns the filter off
+jitter=1
+; the same, for a finger that landed on the button strip: it is pressing
+; rather than steering there, so it needs a stricter margin
+jitter_buttons=45
+; pad units of two-finger travel per wheel step, and the direction
+scroll_step=64
+scroll_invert=0
+; turn a short, still touch into a click: one finger left, two right,
+; three middle
+tap_enable=1
+; tap, then touch again at once and move: the button stays down for that
+; second touch - this is how text is selected and how one draws
+tap_drag=1
+; right click: 0 never, 1 bottom-right corner, 2 by finger count (two
+; fingers right, three middle), 3 either
+right_click=1
+; button strip along the bottom edge, in percent of the surface: its
+; height, and where its left half ends
+button_zone=15
+right_split=50

drivers/i2chid/tpreload.asm

@@ -0,0 +1,72 @@
+; Asks the I2C touchpad driver to read its settings again.
+;
+; The driver takes everything it can be tuned with from the [touchpad] section
+; of /sys/settings/system.ini, and it reads that file once while loading. This
+; tells it to read the file again, so editing a setting takes effect straight
+; away instead of at the next boot.
+;
+; There is no window: the outcome goes to the debug board, the way the driver
+; reports everything else.
+
+format binary as ""
+use32
+        db      'MENUET01'
+        dd      1, start, i_end, mem, mem, 0, 0
+
+include '../../programs/macros.inc'
+
+SRV_RELOAD_CONF = 1
+
+start:
+; 1. Find the driver's service by name.
+        mcall   68, 16, srv_name
+        test    eax, eax
+        jz      .no_driver
+        mov     [ioctl.handle], eax
+; 2. Ask it to reload. No data travels either way: the driver reads the file
+;    itself, so the settings file stays the only place they are written down.
+        mcall   68, 17, ioctl
+        test    eax, eax
+        jnz     .refused
+        mov     esi, msg_ok
+        jmp     .say
+.no_driver:
+        mov     esi, msg_no_driver
+        jmp     .say
+.refused:
+        mov     esi, msg_refused
+.say:
+        call    board_str
+        or      eax, -1
+        mcall
+
+; Writes a zero-terminated string to the debug board.
+; in: esi -> string
+board_str:
+        lodsb
+        test    al, al
+        jz      .done
+        mov     cl, al
+        mcall   63, 1
+        jmp     board_str
+.done:
+        ret
+
+srv_name        db 'I2CHID', 0
+
+align 4
+ioctl:
+.handle         dd 0
+.io_code        dd SRV_RELOAD_CONF
+.input          dd 0
+.inp_size       dd 0
+.output         dd 0
+.out_size       dd 0
+
+msg_ok          db 'tpreload: touchpad settings reloaded', 13, 10, 0
+msg_no_driver   db 'tpreload: the I2C touchpad driver is not loaded', 13, 10, 0
+msg_refused     db 'tpreload: the driver refused the request', 13, 10, 0
+
+i_end:
+        rb      1024
+mem:

drivers/peimport.inc

@@ -139,6 +139,7 @@ import core,\
               RegKeyboard,\
               DelKeyboard,\
               Sleep,\
+              Delay,\
               GetTimerTicks,\
 \
               strncat,\

drivers/usb/usbstor.asm

@@ -1754,8 +1754,14 @@ proc disk_read_write stdcall uses ebx esi edi, \
         mov     eax, [numsectors]
         mov     eax, [eax]
 ; 2. The transfer length for SCSI_{READ,WRITE}10 commands can not be greater
-; than 0xFFFF, so split the request to slices with <= 0xFFFF sectors.
-max_sectors_at_time = 0xFFFF
+; than 0xFFFF sectors, but that is not the practical limit: the field size is
+; the only thing the standard bounds, while real sticks hang their firmware on
+; multi-megabyte single commands. The kernel disk cache asks for as much as
+; CACHE_MAX_ALLOC_SIZE (4 MB) in one call, i.e. 8192 sectors, and that is
+; exactly where such sticks stop answering and re-enumerate. Linux caps
+; usb-storage at 240 sectors (120 KB) for the same reason; do the same and
+; split larger requests into slices.
+max_sectors_at_time = 240
 .split:
         push    eax     ; .length_rest
         cmp     eax, max_sectors_at_time

drivers/usb/xhci/Tupfile.lua

@@ -0,0 +1,3 @@
+if tup.getconfig("NO_FASM") ~= "" then return end
+ROOT = "../../.."
+tup.rule("xhci.asm", "fasm %f %o " .. tup.getconfig("PESTRIP_CMD") .. tup.getconfig("KPACK_CMD"), "%B.sys")

drivers/usb/xhci/xhci_context.inc

@@ -0,0 +1,993 @@
+; Device slots, contexts and pipes for the xHCI driver.
+; This file is included from xhci.asm.
+
+; =============================================================================
+; ============================ Context helpers ================================
+; =============================================================================
+
+; Returns the address of one context inside a context structure.
+; in: eax = base of the structure, ecx = index of the context,
+;     esi -> usb_controller.
+; out: eax = address of the context.
+proc xhci_context
+        push    ecx edx
+        mov     edx, [esi+xhci_controller.ContextSize-XCD]
+        imul    edx, ecx
+        add     eax, edx
+        pop     edx ecx
+        ret
+endp
+
+; Zeroes the Input Context. The caller must hold CmdLock.
+; in: esi -> usb_controller.
+proc xhci_clear_input_ctx
+        push    eax ecx edi
+        mov     edi, [esi+xhci_controller.InputCtx-XCD]
+        xor     eax, eax
+        mov     ecx, 33*64/4
+        rep stosd
+        pop     edi ecx eax
+        ret
+endp
+
+; Returns the pipe of the default control endpoint of the same device, which
+; represents the device as a whole.
+; in: ebx -> xhci_pipe. out: eax -> xhci_pipe or 0.
+proc xhci_ctrl_pipe
+        mov     eax, [ebx+xhci_pipe.DevPage]
+        test    eax, eax
+        jz      .nothing
+        mov     eax, [eax+XHCI_PIPES_OFS+1*4]
+        test    eax, eax
+        jz      .nothing
+        sub     eax, sizeof.xhci_pipe
+.nothing:
+        ret
+endp
+
+; Fills the Slot Context of the Input Context from the device description kept
+; in a pipe. The caller must hold CmdLock.
+; in: esi -> usb_controller, ebx -> xhci_pipe describing the device,
+;     ecx = value for the Context Entries field.
+proc xhci_fill_slot_ctx
+        push    eax ecx edx edi
+        mov     eax, [esi+xhci_controller.InputCtx-XCD]
+        push    ecx
+        movi    ecx, 1
+        call    xhci_context
+        pop     ecx
+        mov     edi, eax
+; dword 0: Route String, Speed and Context Entries.
+        mov     eax, [ebx+xhci_pipe.Route]
+        and     eax, 0FFFFFh
+if XHCI_SUPERSPEED
+; A SuperSpeed root-port device carries the raw PORTSC speed identifier;
+; its kernel-visible Speed pretends high-speed and must not get here.
+        mov     edx, [ebx+xhci_pipe.PSIV]
+        test    edx, edx
+        jnz     .havespeed
+end if
+        mov     edx, [ebx+xhci_pipe.Speed]
+        cmp     edx, USB_SPEED_FS
+        jz      .fs
+        cmp     edx, USB_SPEED_LS
+        jz      .ls
+        movi    edx, XHCI_PSI_HS
+        jmp     .havespeed
+.fs:
+        movi    edx, XHCI_PSI_FS
+        jmp     .havespeed
+.ls:
+        movi    edx, XHCI_PSI_LS
+.havespeed:
+        shl     edx, 20
+        or      eax, edx
+        shl     ecx, 27
+        or      eax, ecx
+        mov     [edi], eax
+; dword 1: Root Hub Port Number.
+        mov     eax, [ebx+xhci_pipe.RootPort]
+        shl     eax, 16
+        mov     [edi+4], eax
+; dword 2: the Transaction Translator, if the device is served by one.
+        mov     eax, [ebx+xhci_pipe.TTSlot]
+        test    eax, eax
+        jz      @f
+        mov     edx, [ebx+xhci_pipe.TTPort]
+        shl     edx, 8
+        or      eax, edx
+@@:
+        mov     [edi+8], eax
+        and     dword [edi+12], 0
+        pop     edi edx ecx eax
+        ret
+endp
+
+; Fills the Endpoint Context of the Input Context for the given pipe.
+; The caller must hold CmdLock.
+; in: esi -> usb_controller, ebx -> xhci_pipe.
+proc xhci_fill_ep_ctx
+        push    eax ecx edx edi
+        mov     eax, [esi+xhci_controller.InputCtx-XCD]
+        mov     ecx, [ebx+xhci_pipe.DCI]
+        inc     ecx
+        call    xhci_context
+        mov     edi, eax
+; dword 0: Interval.
+        mov     eax, [ebx+xhci_pipe.Interval]
+        shl     eax, 16
+        mov     [edi], eax
+; dword 1: Error Count, Endpoint Type, Max Burst Size and Max Packet Size.
+        mov     eax, [ebx+xhci_pipe.EpType]
+        shl     eax, 3
+        or      eax, 3 shl 1            ; three retries before giving up
+        mov     edx, [ebx+xhci_pipe.MaxPacket]
+        mov     ecx, edx
+        shr     ecx, 11
+        and     ecx, 3                  ; extra transactions per microframe
+        shl     ecx, 8
+        or      eax, ecx
+if XHCI_SUPERSPEED
+; A SuperSpeed bulk endpoint may allow bursts of several packets; the field
+; shares its bits with the extra-transactions ones, which are zero then.
+        mov     ecx, [ebx+xhci_pipe.MaxBurst]
+        shl     ecx, 8
+        or      eax, ecx
+end if
+        and     edx, 7FFh
+        shl     edx, 16
+        or      eax, edx
+        mov     [edi+4], eax
+; dwords 2 and 3: the TR Dequeue Pointer with the Dequeue Cycle State.
+        mov     eax, [ebx+xhci_pipe.Enqueue]
+        shl     eax, 4
+        add     eax, [ebx+xhci_pipe.RingPhys]
+        or      eax, [ebx+xhci_pipe.Cycle]
+        mov     [edi+8], eax
+        and     dword [edi+12], 0
+; dword 4: Average TRB Length, only a hint for the bandwidth scheduler.
+        mov     eax, [ebx+xhci_pipe.MaxPacket]
+        and     eax, 7FFh
+        cmp     [ebx+xhci_pipe.EpType], XHCI_EP_CONTROL
+        jnz     @f
+        movi    eax, 8
+@@:
+        mov     [edi+16], eax
+        and     dword [edi+20], 0
+        pop     edi edx ecx eax
+        ret
+endp
+
+; =============================================================================
+; ============================== Slot commands ================================
+; =============================================================================
+
+; Issues the Address Device command for a newly created default control pipe.
+; in: esi -> usb_controller, ebx -> xhci_pipe. out: eax = 0 on failure.
+proc xhci_address_device
+        push    ebx ecx edx edi
+        lea     ecx, [esi+xhci_controller.CmdLock-XCD]
+        invoke  MutexLock
+        call    xhci_clear_input_ctx
+        mov     edx, [esi+xhci_controller.InputCtx-XCD]
+        mov     dword [edx+4], 3        ; add the Slot Context and endpoint zero
+        movi    ecx, 1
+        call    xhci_fill_slot_ctx
+        call    xhci_fill_ep_ctx
+        mov     [ebx+xhci_pipe.MaxDCI], 1
+; Point the DCBAA entry of the slot at the Device Context.
+        mov     eax, [esi+xhci_controller.CommonPage-XCD]
+        mov     ecx, [ebx+xhci_pipe.SlotId]
+        mov     edx, [ebx+xhci_pipe.DevPagePhys]
+        mov     [eax+ecx*8], edx
+        and     dword [eax+ecx*8+4], 0
+        mov     edx, [ebx+xhci_pipe.SlotId]
+        shl     edx, 24
+        or      edx, XHCI_TRB_ADDRESS_DEV shl 10
+        mov     eax, [esi+xhci_controller.InputCtxPhys-XCD]
+        push    ebx
+        xor     ebx, ebx
+        xor     ecx, ecx
+        call    xhci_cmd_submit
+        pop     ebx
+        push    eax
+        lea     ecx, [esi+xhci_controller.CmdLock-XCD]
+        invoke  MutexUnlock
+        pop     eax
+        cmp     eax, XHCI_CC_SUCCESS
+        jz      .ok
+        DEBUGF 1,'K : XHCI: Address Device failed, code %d\n',eax
+        xor     eax, eax
+        jmp     .nothing
+.ok:
+        movi    eax, 1
+.nothing:
+        pop     edi edx ecx ebx
+        ret
+endp
+
+; Issues a Configure Endpoint command that adds the given endpoint.
+; in: esi -> usb_controller, ebx -> xhci_pipe of the endpoint,
+;     edi -> xhci_pipe of the default control endpoint of the same device.
+; out: eax = completion code.
+proc xhci_configure_endpoint
+        push    ebx ecx edx esi edi
+        lea     ecx, [esi+xhci_controller.CmdLock-XCD]
+        invoke  MutexLock
+        call    xhci_clear_input_ctx
+; Drop the endpoint and add it again, so that reopening an endpoint that has
+; been closed before also works.
+        mov     ecx, [ebx+xhci_pipe.DCI]
+        movi    eax, 1
+        shl     eax, cl
+        mov     edx, [esi+xhci_controller.InputCtx-XCD]
+        mov     [edx], eax
+        or      eax, 1                  ; the Slot Context is always added
+        mov     [edx+4], eax
+        call    xhci_fill_ep_ctx
+; The Context Entries field must cover every endpoint in use by the device.
+        mov     ecx, [ebx+xhci_pipe.DCI]
+        cmp     ecx, [edi+xhci_pipe.MaxDCI]
+        jae     @f
+        mov     ecx, [edi+xhci_pipe.MaxDCI]
+@@:
+        mov     [edi+xhci_pipe.MaxDCI], ecx
+        push    ebx
+        mov     ebx, edi
+        call    xhci_fill_slot_ctx
+        pop     ebx
+        mov     edx, [ebx+xhci_pipe.SlotId]
+        shl     edx, 24
+        or      edx, XHCI_TRB_CONFIG_EP shl 10
+        mov     eax, [esi+xhci_controller.InputCtxPhys-XCD]
+        push    ebx
+        xor     ebx, ebx
+        xor     ecx, ecx
+        call    xhci_cmd_submit
+        pop     ebx
+        push    eax
+        lea     ecx, [esi+xhci_controller.CmdLock-XCD]
+        invoke  MutexUnlock
+        pop     eax
+        pop     edi esi edx ecx ebx
+        ret
+endp
+
+; Tells the controller where the endpoint has to continue from.
+; in: esi -> usb_controller, ebx -> xhci_pipe.
+proc xhci_set_dequeue
+        pushad
+        cmp     [ebx+xhci_pipe.SlotId], 0
+        jz      .nothing
+        mov     eax, [ebx+xhci_pipe.Dequeue]
+        shl     eax, 4
+        add     eax, [ebx+xhci_pipe.RingPhys]
+        or      eax, [ebx+xhci_pipe.Cycle]
+        mov     ecx, [ebx+xhci_pipe.DCI]
+        shl     ecx, 16
+        mov     edx, [ebx+xhci_pipe.SlotId]
+        shl     edx, 24
+        or      edx, ecx
+        or      edx, XHCI_TRB_SET_DEQ shl 10
+        xor     ebx, ebx
+        xor     ecx, ecx
+        call    xhci_cmd_sync
+        cmp     eax, XHCI_CC_SUCCESS
+        jz      .nothing
+        DEBUGF 1,'K : XHCI: Set TR Dequeue Pointer failed, code %d\n',eax
+.nothing:
+        popad
+        ret
+endp
+
+; =============================================================================
+; =============================== New device ==================================
+; =============================================================================
+
+; Called from xhci_port_init and from the hub support code when a new device
+; has been connected and reset. Enables a device slot for it and hands it over
+; to the protocol layer.
+; in: esi -> usb_controller, eax = speed. out: eax = 0 on failure.
+proc xhci_new_device
+        push    ebx ecx edx edi
+        mov     [esi+usb_controller.ResettingSpeed], al
+; 1. Reserve room for the pseudo-pipe that describes the new device. Only the
+; fields read by xhci_init_pipe are filled in; the layout matches a real pipe,
+; so the result can be passed to usb_new_device as a usb_pipe pointer.
+        sub     esp, sizeof.xhci_pipe + 4
+        mov     ebx, esp
+        push    ebx
+        mov     edi, ebx
+        xor     eax, eax
+        mov     ecx, (sizeof.xhci_pipe + 4)/4
+        rep stosd
+        pop     ebx
+        mov     [ebx+sizeof.xhci_pipe], esi     ; usb_pipe.Controller
+        movzx   eax, [esi+usb_controller.ResettingSpeed]
+        mov     [ebx+xhci_pipe.Speed], eax
+; 2. Compute the root hub port, the route string and the tier of the device.
+        movzx   edx, [esi+usb_controller.ResettingPort]
+        mov     edi, [esi+usb_controller.ResettingHub]
+        test    edi, edi
+        jnz     .behind_hub
+        mov     eax, [esi+xhci_controller.PortMap-XCD+edx*4]
+        mov     [ebx+xhci_pipe.RootPort], eax
+if XHCI_SUPERSPEED
+        mov     eax, [esi+xhci_controller.ResettingPSIV-XCD]
+        mov     [ebx+xhci_pipe.PSIV], eax
+end if
+        jmp     .have_topology
+.behind_hub:
+; The parent hub is a device too, and its own pipe carries the topology data.
+        mov     eax, [edi+USB_HUB_CONFIGPIPE]
+        sub     eax, sizeof.xhci_pipe
+        mov     ecx, [eax+xhci_pipe.RootPort]
+        mov     [ebx+xhci_pipe.RootPort], ecx
+        mov     ecx, [eax+xhci_pipe.Depth]
+        mov     eax, [eax+xhci_pipe.Route]
+; The route string holds one nibble per tier, at most five of them.
+        cmp     ecx, 5
+        jae     .have_route
+        push    ecx edx
+        shl     ecx, 2
+        inc     edx                     ; the route uses 1-based port numbers
+        cmp     edx, 15
+        jbe     @f
+        movi    edx, 15
+@@:
+        shl     edx, cl
+        or      eax, edx
+        pop     edx ecx
+.have_route:
+        inc     ecx
+        mov     [ebx+xhci_pipe.Depth], ecx
+        mov     [ebx+xhci_pipe.Route], eax
+; 3. Low- and full-speed devices behind a high-speed hub need the address of
+; the Transaction Translator that serves them.
+        cmp     [ebx+xhci_pipe.Speed], USB_SPEED_HS
+        jz      .have_topology
+        push    ebx
+        mov     ecx, edx
+        mov     edx, edi
+        invoke  usbhc_api.usb_get_tt
+        pop     ebx
+        test    edx, edx
+        jz      .have_topology
+        sub     edx, sizeof.xhci_pipe
+        mov     eax, [edx+xhci_pipe.SlotId]
+        mov     [ebx+xhci_pipe.TTSlot], eax
+        inc     ecx
+        mov     [ebx+xhci_pipe.TTPort], ecx
+.have_topology:
+; 4. Ask the controller for a device slot.
+        xor     eax, eax
+        push    ebx
+        xor     ebx, ebx
+        xor     ecx, ecx
+        mov     edx, XHCI_TRB_ENABLE_SLOT shl 10
+        call    xhci_cmd_sync
+        pop     ebx
+        cmp     eax, XHCI_CC_SUCCESS
+        jnz     .fail
+        test    edx, edx
+        jz      .fail
+        mov     [ebx+xhci_pipe.SlotId], edx
+        DEBUGF 1,'K : XHCI new device: slot %d, root port %d, route %x, speed %d\n',edx,[ebx+xhci_pipe.RootPort],[ebx+xhci_pipe.Route],[ebx+xhci_pipe.Speed]
+; 5. Hand the device over to the protocol layer. It opens the pipe of the
+; default control endpoint, and that is where the slot is really addressed.
+        lea     ecx, [ebx+sizeof.xhci_pipe]
+        invoke  usbhc_api.usb_new_device
+        test    eax, eax
+        jnz     .done
+; 6. The protocol layer has failed; release the slot again.
+        mov     edx, [ebx+xhci_pipe.SlotId]
+        shl     edx, 24
+        or      edx, XHCI_TRB_DISABLE_SLOT shl 10
+        xor     eax, eax
+        xor     ebx, ebx
+        xor     ecx, ecx
+        call    xhci_cmd_sync
+        xor     eax, eax
+        jmp     .done
+.fail:
+        dbgstr 'XHCI: Enable Slot failed'
+        xor     eax, eax
+.done:
+        add     esp, sizeof.xhci_pipe + 4
+        pop     edi edx ecx ebx
+        ret
+endp
+
+; =============================================================================
+; ================================= Pipes =====================================
+; =============================================================================
+
+proc xhci_alloc_pipe
+        push    ebx ecx edi
+        mov     ebx, xhci_ep_mutex
+        invoke  usbhc_api.usb_allocate_common, sizeof.xhci_pipe + sizeof.usb_pipe
+        test    eax, eax
+        jz      @f
+; Zero the whole block: the kernel may call FreePipe for a pipe whose InitPipe
+; has never run, and that path must not act on garbage.
+        mov     edi, eax
+        push    eax
+        xor     eax, eax
+        movi    ecx, (sizeof.xhci_pipe + sizeof.usb_pipe + 3)/4
+        rep stosd
+        pop     eax
+        add     eax, sizeof.xhci_pipe
+@@:
+        pop     edi ecx ebx
+        ret
+endp
+
+proc xhci_free_pipe stdcall, ptr:dword
+        push    ebx esi edi
+        mov     ebx, [ptr]
+        sub     ebx, sizeof.xhci_pipe
+        mov     esi, [ebx+sizeof.xhci_pipe+usb_pipe.Controller]
+        test    esi, esi
+        jz      .free
+        cmp     [ebx+xhci_pipe.SlotId], 0
+        jz      .drop_pipe
+        cmp     [ebx+xhci_pipe.DCI], 1
+        jnz     .drop_pipe
+; This is the default control endpoint, that is, the handle of the whole
+; device: release the slot together with the per-device page.
+        mov     ecx, [ebx+xhci_pipe.SlotId]
+        mov     edx, ecx
+        shl     edx, 24
+        or      edx, XHCI_TRB_DISABLE_SLOT shl 10
+        xor     eax, eax
+        push    ebx
+        xor     ebx, ebx
+        xor     ecx, ecx
+        call    xhci_cmd_sync
+        pop     ebx
+; The controller may access the Device Context until the command above
+; completes, so only now may the DCBAA entry and the slot table forget it.
+        mov     ecx, [ebx+xhci_pipe.SlotId]
+        mov     edi, [esi+xhci_controller.CommonPage-XCD]
+        and     dword [edi+ecx*8], 0
+        and     dword [edi+ecx*8+4], 0
+        and     [esi+xhci_controller.SlotPages-XCD+ecx*4], 0
+        mov     eax, [ebx+xhci_pipe.DevPage]
+        test    eax, eax
+        jz      .drop_ring
+        invoke  KernelFree, eax
+        and     [ebx+xhci_pipe.DevPage], 0
+        jmp     .drop_ring
+.drop_pipe:
+; Remove the pipe from the per-device table, if it is still listed there.
+; On disconnect the kernel releases the pipes of a device in arbitrary order,
+; and the pipe of the default control endpoint - the owner of the per-device
+; page - may well be freed before this one, taking the page with it. The
+; authoritative pointer is the one in SlotPages: it is cleared when the slot
+; is released, so it can never point to freed memory.
+        mov     ecx, [ebx+xhci_pipe.SlotId]
+        test    ecx, ecx
+        jz      .drop_ring
+        mov     eax, [esi+xhci_controller.SlotPages-XCD+ecx*4]
+        test    eax, eax
+        jz      .drop_ring
+        mov     ecx, [ebx+xhci_pipe.DCI]
+        cmp     ecx, 32
+        jae     .drop_ring
+        and     dword [eax+XHCI_PIPES_OFS+ecx*4], 0
+.drop_ring:
+        mov     eax, [ebx+xhci_pipe.Ring]
+        test    eax, eax
+        jz      @f
+        invoke  KernelFree, eax
+        and     [ebx+xhci_pipe.Ring], 0
+@@:
+        mov     eax, [ebx+xhci_pipe.RingMap]
+        test    eax, eax
+        jz      .free
+        invoke  KernelFree, eax
+        and     [ebx+xhci_pipe.RingMap], 0
+.free:
+        pop     edi esi ebx
+        mov     eax, [ptr]
+        sub     eax, sizeof.xhci_pipe
+        invoke  usbhc_api.usb_free_common, eax
+        ret
+endp
+
+; Allocates the transfer ring of a pipe.
+; in: ebx -> xhci_pipe. out: eax = 0 on failure.
+proc xhci_alloc_ring
+        push    ecx edx
+        call    xhci_alloc_page
+        test    eax, eax
+        jz      .nothing
+        mov     [ebx+xhci_pipe.RingMap], eax
+        call    xhci_alloc_page
+        test    eax, eax
+        jnz     @f
+        invoke  KernelFree, [ebx+xhci_pipe.RingMap]
+        and     [ebx+xhci_pipe.RingMap], 0
+        xor     eax, eax
+        jmp     .nothing
+@@:
+        mov     [ebx+xhci_pipe.Ring], eax
+        mov     [ebx+xhci_pipe.RingPhys], edx
+; The last entry returns to the beginning of the ring and toggles the cycle.
+        lea     ecx, [eax+(XHCI_RING_TRBS-1)*16]
+        mov     [ecx], edx
+        mov     dword [ecx+12], (XHCI_TRB_LINK shl 10) + XHCI_TRB_TC
+        and     [ebx+xhci_pipe.Enqueue], 0
+        and     [ebx+xhci_pipe.Dequeue], 0
+        and     [ebx+xhci_pipe.Reserved], 0
+        and     [ebx+xhci_pipe.FirstFix], 0
+        mov     [ebx+xhci_pipe.Cycle], XHCI_TRB_C
+        movi    eax, 1
+.nothing:
+        pop     edx ecx
+        ret
+endp
+
+; Stores the pipe in the per-device table used by the event ring handler.
+; in: ebx -> xhci_pipe.
+proc xhci_register_pipe
+        push    eax ecx edx
+        mov     eax, [ebx+xhci_pipe.DevPage]
+        test    eax, eax
+        jz      @f
+        mov     ecx, [ebx+xhci_pipe.DCI]
+        cmp     ecx, 32
+        jae     @f
+        lea     edx, [ebx+sizeof.xhci_pipe]
+        mov     [eax+XHCI_PIPES_OFS+ecx*4], edx
+@@:
+        pop     edx ecx eax
+        ret
+endp
+
+; Computes the Interval field of the Endpoint Context of an interrupt endpoint.
+; The kernel passes bInterval from the endpoint descriptor: for high-speed
+; devices it is an exponent of 125 us periods, for low- and full-speed devices
+; it is a number of frames.
+; in: ebx -> xhci_pipe, eax = bInterval.
+proc xhci_calc_interval
+        push    eax ecx
+        cmp     [ebx+xhci_pipe.Speed], USB_SPEED_HS
+        jnz     .low_speed
+        test    eax, eax
+        jz      @f
+        dec     eax
+@@:
+        jmp     .clamp
+.low_speed:
+; One frame equals eight periods of 125 us, hence the offset of three.
+        test    eax, eax
+        jnz     @f
+        movi    eax, 1
+@@:
+        bsr     ecx, eax
+        lea     eax, [ecx+3]
+.clamp:
+        cmp     eax, 15
+        jbe     @f
+        movi    eax, 15
+@@:
+        mov     [ebx+xhci_pipe.Interval], eax
+        pop     ecx eax
+        ret
+endp
+
+; Called from usb_open_pipe.
+; in: edi -> usb_pipe for the target, ecx -> usb_pipe of the config pipe,
+;     esi -> usb_controller, eax -> usb_gtd of the first descriptor,
+;     [ebp+12] = endpoint, [ebp+16] = maxpacket, [ebp+20] = type,
+;     [ebp+24] = interval.
+; out: eax = 0 on failure.
+proc xhci_init_pipe
+virtual at ebp+8
+.config_pipe    dd      ?
+.endpoint       dd      ?
+.maxpacket      dd      ?
+.type           dd      ?
+.interval       dd      ?
+end virtual
+        push    ebx
+        sub     edi, sizeof.xhci_pipe
+        sub     ecx, sizeof.xhci_pipe
+        mov     ebx, edi                ; ebx -> xhci_pipe being initialized
+; 1. Copy the description of the device from the config pipe.
+        mov     eax, [ecx+xhci_pipe.SlotId]
+        mov     [ebx+xhci_pipe.SlotId], eax
+        mov     eax, [ecx+xhci_pipe.DevPage]
+        mov     [ebx+xhci_pipe.DevPage], eax
+        mov     eax, [ecx+xhci_pipe.DevPagePhys]
+        mov     [ebx+xhci_pipe.DevPagePhys], eax
+        mov     eax, [ecx+xhci_pipe.Speed]
+        mov     [ebx+xhci_pipe.Speed], eax
+        mov     eax, [ecx+xhci_pipe.RootPort]
+        mov     [ebx+xhci_pipe.RootPort], eax
+        mov     eax, [ecx+xhci_pipe.Route]
+        mov     [ebx+xhci_pipe.Route], eax
+        mov     eax, [ecx+xhci_pipe.Depth]
+        mov     [ebx+xhci_pipe.Depth], eax
+        mov     eax, [ecx+xhci_pipe.TTSlot]
+        mov     [ebx+xhci_pipe.TTSlot], eax
+        mov     eax, [ecx+xhci_pipe.TTPort]
+        mov     [ebx+xhci_pipe.TTPort], eax
+if XHCI_SUPERSPEED
+        mov     eax, [ecx+xhci_pipe.PSIV]
+        mov     [ebx+xhci_pipe.PSIV], eax
+end if
+; 2. Work out the Device Context Index and the endpoint type.
+        mov     eax, [.endpoint]
+        test    eax, eax
+        jz      .control_ep
+        mov     ecx, eax
+        and     ecx, 15
+        add     ecx, ecx
+        test    al, 80h
+        jz      @f
+        inc     ecx
+        or      [ebx+xhci_pipe.Flags], XHCI_PIPE_IN
+@@:
+        mov     [ebx+xhci_pipe.DCI], ecx
+        mov     eax, [.type]
+        cmp     eax, BULK_PIPE
+        jz      .bulk
+        cmp     eax, INTERRUPT_PIPE
+        jz      .interrupt
+; A control endpoint other than endpoint zero is unusual but legal.
+        mov     [ebx+xhci_pipe.EpType], XHCI_EP_CONTROL
+        jmp     .have_type
+.bulk:
+        movi    eax, XHCI_EP_BULK_OUT
+        test    [ebx+xhci_pipe.Flags], XHCI_PIPE_IN
+        jz      @f
+        movi    eax, XHCI_EP_BULK_IN
+@@:
+        mov     [ebx+xhci_pipe.EpType], eax
+        jmp     .have_type
+.interrupt:
+        movi    eax, XHCI_EP_INT_OUT
+        test    [ebx+xhci_pipe.Flags], XHCI_PIPE_IN
+        jz      @f
+        movi    eax, XHCI_EP_INT_IN
+@@:
+        mov     [ebx+xhci_pipe.EpType], eax
+        mov     eax, [.interval]
+        call    xhci_calc_interval
+.have_type:
+        mov     eax, [.maxpacket]
+        mov     [ebx+xhci_pipe.MaxPacket], eax
+if XHCI_SUPERSPEED
+        call    xhci_ss_get_burst
+end if
+        jmp     .alloc_ring
+.control_ep:
+; The default control endpoint always has index one. Until the real packet size
+; is known from the device descriptor, use the value the specification requires
+; for the speed of the device: the kernel asks for 64 regardless of the speed,
+; which is wrong for low- and full-speed devices.
+        mov     [ebx+xhci_pipe.DCI], 1
+        mov     [ebx+xhci_pipe.EpType], XHCI_EP_CONTROL
+        movi    eax, 64
+        cmp     [ebx+xhci_pipe.Speed], USB_SPEED_HS
+        jz      @f
+        movi    eax, 8
+@@:
+if XHCI_SUPERSPEED
+; A SuperSpeed device always uses 512 bytes on the default control endpoint.
+        cmp     [ebx+xhci_pipe.PSIV], XHCI_PSI_SS
+        jb      @f
+        mov     eax, 512
+@@:
+end if
+        mov     [ebx+xhci_pipe.MaxPacket], eax
+.alloc_ring:
+; 3. Allocate the transfer ring.
+        call    xhci_alloc_ring
+        test    eax, eax
+        jz      .fail
+; 4. Choose the software list the pipe belongs to and insert it there. The
+; hardware knows nothing about these lists; they exist only so that the kernel
+; can walk all pipes and reinsert an aborted one.
+        lea     edx, [esi+xhci_controller.ControlED-XCD]
+        cmp     [.type], BULK_PIPE
+        jb      @f
+        lea     edx, [esi+xhci_controller.BulkED-XCD]
+        jz      @f
+        lea     edx, [esi+xhci_controller.IntED-XCD]
+@@:
+        lea     edi, [ebx+sizeof.xhci_pipe]
+        mov     [edi+usb_pipe.BaseList], edx
+        mov     ecx, [edx+usb_pipe.NextVirt]
+        mov     [edi+usb_pipe.NextVirt], ecx
+        mov     [edi+usb_pipe.PrevVirt], edx
+        mov     [ecx+usb_pipe.PrevVirt], edi
+        mov     [edx+usb_pipe.NextVirt], edi
+; 5. Now do the hardware part.
+        cmp     [.endpoint], 0
+        jz      .address_device
+        call    xhci_register_pipe
+        mov     edi, [.config_pipe]
+        sub     edi, sizeof.xhci_pipe
+        call    xhci_configure_endpoint
+        xhci_trace 'K : XHCI open pipe: slot %d dci %d\n', [ebx+xhci_pipe.SlotId], [ebx+xhci_pipe.DCI]
+        cmp     eax, XHCI_CC_SUCCESS
+        jz      .success
+        DEBUGF 1,'K : XHCI: Configure Endpoint failed, code %d\n',eax
+        jmp     .unlink_fail
+.address_device:
+; 6. This is a new device: allocate its Device Context, publish the slot and
+; let the controller give the device a bus address.
+        call    xhci_alloc_page
+        test    eax, eax
+        jz      .unlink_fail
+        mov     [ebx+xhci_pipe.DevPage], eax
+        mov     [ebx+xhci_pipe.DevPagePhys], edx
+        mov     ecx, [ebx+xhci_pipe.SlotId]
+        mov     [esi+xhci_controller.SlotPages-XCD+ecx*4], eax
+        call    xhci_register_pipe
+        call    xhci_address_device
+        test    eax, eax
+        jz      .free_devpage
+.success:
+; usb_open_pipe keeps the pointer to the new pipe in edi across this call, so
+; it has to be handed back untouched; the branch above has overwritten it with
+; the pipe of the default control endpoint.
+        lea     edi, [ebx+sizeof.xhci_pipe]
+        pop     ebx
+        movi    eax, 1
+        ret
+.free_devpage:
+        mov     ecx, [ebx+xhci_pipe.SlotId]
+        and     [esi+xhci_controller.SlotPages-XCD+ecx*4], 0
+        mov     eax, [ebx+xhci_pipe.DevPage]
+        invoke  KernelFree, eax
+        and     [ebx+xhci_pipe.DevPage], 0
+.unlink_fail:
+; Undo the insertion into the software list before failing.
+        lea     edi, [ebx+sizeof.xhci_pipe]
+        mov     eax, [edi+usb_pipe.NextVirt]
+        mov     edx, [edi+usb_pipe.PrevVirt]
+        mov     [eax+usb_pipe.PrevVirt], edx
+        mov     [edx+usb_pipe.NextVirt], eax
+.fail:
+        lea     edi, [ebx+sizeof.xhci_pipe]
+        pop     ebx
+        xor     eax, eax
+        ret
+endp
+
+; Removes the pipe from the hardware configuration of the device.
+; in: esi -> usb_controller, ebx -> usb_pipe.
+proc xhci_unlink_pipe
+        pushad
+        sub     ebx, sizeof.xhci_pipe
+        cmp     [ebx+xhci_pipe.SlotId], 0
+        jz      .nothing
+        cmp     [ebx+xhci_pipe.DCI], 1
+        jz      .nothing        ; the slot itself is released by FreePipe
+        call    xhci_ctrl_pipe
+        test    eax, eax
+        jz      .nothing
+        mov     edi, eax                ; edi -> pipe of endpoint zero
+        lea     ecx, [esi+xhci_controller.CmdLock-XCD]
+        invoke  MutexLock
+        call    xhci_clear_input_ctx
+        mov     ecx, [ebx+xhci_pipe.DCI]
+        movi    eax, 1
+        shl     eax, cl
+        mov     edx, [esi+xhci_controller.InputCtx-XCD]
+        mov     [edx], eax              ; drop this endpoint
+        mov     dword [edx+4], 1        ; keep the Slot Context
+        mov     ecx, [edi+xhci_pipe.MaxDCI]
+        test    ecx, ecx
+        jnz     @f
+        movi    ecx, 1
+@@:
+        push    ebx
+        mov     ebx, edi
+        call    xhci_fill_slot_ctx
+        pop     ebx
+        mov     edx, [ebx+xhci_pipe.SlotId]
+        shl     edx, 24
+        or      edx, XHCI_TRB_CONFIG_EP shl 10
+        mov     eax, [esi+xhci_controller.InputCtxPhys-XCD]
+        xor     ebx, ebx
+        xor     ecx, ecx
+        call    xhci_cmd_submit
+        lea     ecx, [esi+xhci_controller.CmdLock-XCD]
+        invoke  MutexUnlock
+.nothing:
+        popad
+        ret
+endp
+
+; Temporarily removes the pipe from the hardware schedule.
+; in: esi -> usb_controller, ebx -> usb_pipe.
+proc xhci_disable_pipe
+        pushad
+        sub     ebx, sizeof.xhci_pipe
+        cmp     [ebx+xhci_pipe.SlotId], 0
+        jz      .nothing
+        mov     ecx, [ebx+xhci_pipe.DCI]
+        shl     ecx, 16
+        mov     edx, [ebx+xhci_pipe.SlotId]
+        shl     edx, 24
+        or      edx, ecx
+        or      edx, XHCI_TRB_STOP_EP shl 10
+        xor     eax, eax
+        xor     ebx, ebx
+        xor     ecx, ecx
+        call    xhci_cmd_sync
+.nothing:
+        popad
+        ret
+endp
+
+; Reinserts the pipe into the hardware schedule after xhci_disable_pipe, with
+; an empty transfer queue.
+; in: esi -> usb_controller, ebx -> usb_pipe,
+;     edx -> current descriptor, eax -> new last descriptor.
+proc xhci_enable_pipe
+        pushad
+        sub     ebx, sizeof.xhci_pipe
+        mov     eax, [ebx+xhci_pipe.Ring]
+        test    eax, eax
+        jz      .nothing
+; The queue is empty now, so the ring can be restarted from its beginning.
+        mov     edi, eax
+        xor     eax, eax
+        mov     ecx, 0x1000/4
+        rep stosd
+        mov     edi, [ebx+xhci_pipe.RingMap]
+        mov     ecx, 0x1000/4
+        rep stosd
+        mov     edi, [ebx+xhci_pipe.Ring]
+        add     edi, (XHCI_RING_TRBS-1)*16
+        mov     eax, [ebx+xhci_pipe.RingPhys]
+        mov     [edi], eax
+        mov     dword [edi+12], (XHCI_TRB_LINK shl 10) + XHCI_TRB_TC
+        and     [ebx+xhci_pipe.Enqueue], 0
+        and     [ebx+xhci_pipe.Dequeue], 0
+        and     [ebx+xhci_pipe.Reserved], 0
+        and     [ebx+xhci_pipe.FirstFix], 0
+        mov     [ebx+xhci_pipe.Cycle], XHCI_TRB_C
+        call    xhci_set_dequeue
+.nothing:
+        popad
+        ret
+endp
+
+; =============================================================================
+; ========================== Address and packet size ==========================
+; =============================================================================
+
+; Called from usb_set_address_callback. The bus address is chosen by the
+; controller itself while executing the Address Device command, so the only
+; thing left to do is to remember the number the kernel has reserved: it asks
+; for it back on disconnect in order to mark it free again.
+; in: esi -> usb_controller, ebx -> usb_pipe, cl = address.
+proc xhci_set_device_address
+        movzx   eax, cl
+        mov     [ebx+xhci_pipe.Address-sizeof.xhci_pipe], eax
+        jmp     [usbhc_api.usb_subscribe_control]
+endp
+
+; in: esi -> usb_controller, ebx -> usb_pipe. out: eax = address.
+proc xhci_get_device_address
+        mov     eax, [ebx+xhci_pipe.Address-sizeof.xhci_pipe]
+        ret
+endp
+
+; Called from usb_get_descr8_callback once the real packet size of the default
+; control endpoint is known. On xHCI that value lives in the Endpoint Context,
+; so an Evaluate Context command is needed to update it.
+; in: esi -> usb_controller, ebx -> usb_pipe, ecx = packet size.
+proc xhci_set_endpoint_packet_size
+        push    ebx ecx edx edi
+        sub     ebx, sizeof.xhci_pipe
+if XHCI_SUPERSPEED
+; For a SuperSpeed device bMaxPacketSize0 is an exponent, and the only value
+; the specification allows is 9: use 512 directly instead of the raw byte.
+        cmp     [ebx+xhci_pipe.PSIV], XHCI_PSI_SS
+        jb      @f
+        mov     ecx, 512
+@@:
+end if
+        mov     [ebx+xhci_pipe.MaxPacket], ecx
+        lea     ecx, [esi+xhci_controller.CmdLock-XCD]
+        invoke  MutexLock
+        call    xhci_clear_input_ctx
+        mov     edx, [esi+xhci_controller.InputCtx-XCD]
+        mov     dword [edx+4], 2        ; evaluate the default control endpoint
+        call    xhci_fill_ep_ctx
+        mov     edx, [ebx+xhci_pipe.SlotId]
+        shl     edx, 24
+        or      edx, XHCI_TRB_EVAL_CTX shl 10
+        mov     eax, [esi+xhci_controller.InputCtxPhys-XCD]
+        push    ebx
+        xor     ebx, ebx
+        xor     ecx, ecx
+        call    xhci_cmd_submit
+        pop     ebx
+        cmp     eax, XHCI_CC_SUCCESS
+        jz      @f
+        DEBUGF 1,'K : XHCI: Evaluate Context failed, code %d\n',eax
+@@:
+        lea     ecx, [esi+xhci_controller.CmdLock-XCD]
+        invoke  MutexUnlock
+        pop     edi edx ecx ebx
+        jmp     [usbhc_api.usb_subscribe_control]
+endp
+
+; If the descriptor that has just completed is the answer to a request for the
+; hub descriptor, tell the controller that the device is a hub: it has to know
+; that in order to serve the devices behind it.
+; in: ebx -> usb_gtd.
+proc xhci_check_hub_descriptor
+        pushad
+        cmp     [ebx+usb_gtd.Callback], 0
+        jz      .nothing
+        mov     eax, [ebx+usb_gtd.Pipe]
+        mov     esi, [eax+usb_pipe.Controller]
+        sub     eax, sizeof.xhci_pipe
+        test    [eax+xhci_pipe.Flags], XHCI_PIPE_HUBPEND
+        jz      .nothing
+        and     [eax+xhci_pipe.Flags], not XHCI_PIPE_HUBPEND
+        cmp     [ebx+xhci_gtd.Status-sizeof.xhci_gtd], XHCI_CC_SUCCESS
+        jnz     .nothing
+        mov     edx, [eax+xhci_pipe.HubBuf]
+        test    edx, edx
+        jz      .nothing
+        cmp     byte [edx+1], 29h       ; bDescriptorType
+        jnz     .nothing
+        or      [eax+xhci_pipe.Flags], XHCI_PIPE_ISHUB
+        mov     ebx, eax                ; ebx -> xhci_pipe of the hub
+        movzx   edi, byte [edx+2]       ; bNbrPorts
+        movzx   ebp, byte [edx+3]       ; low byte of wHubCharacteristics
+        shr     ebp, 5
+        and     ebp, 3                  ; Transaction Translator Think Time
+        DEBUGF 1,'K : XHCI: device on slot %d is a hub with %d ports\n',[ebx+xhci_pipe.SlotId],edi
+        lea     ecx, [esi+xhci_controller.CmdLock-XCD]
+        invoke  MutexLock
+        call    xhci_clear_input_ctx
+        mov     eax, [esi+xhci_controller.InputCtx-XCD]
+        mov     dword [eax+4], 1        ; only the Slot Context is evaluated
+        mov     ecx, [ebx+xhci_pipe.MaxDCI]
+        test    ecx, ecx
+        jnz     @f
+        movi    ecx, 1
+@@:
+        call    xhci_fill_slot_ctx
+; Patch the Slot Context with the fields that describe a hub. Only Configure
+; Endpoint takes them into account, Evaluate Context would ignore them.
+        mov     eax, [esi+xhci_controller.InputCtx-XCD]
+        movi    ecx, 1
+        call    xhci_context
+        or      dword [eax], 1 shl 26   ; this device is a hub
+        cmp     [ebx+xhci_pipe.Speed], USB_SPEED_HS
+        jnz     @f
+        shl     ebp, 16
+        or      [eax+8], ebp
+@@:
+        shl     edi, 24
+        or      [eax+4], edi            ; Number of Ports
+        mov     edx, [ebx+xhci_pipe.SlotId]
+        shl     edx, 24
+        or      edx, XHCI_TRB_CONFIG_EP shl 10
+        mov     eax, [esi+xhci_controller.InputCtxPhys-XCD]
+        xor     ebx, ebx
+        xor     ecx, ecx
+        call    xhci_cmd_submit
+        cmp     eax, XHCI_CC_SUCCESS
+        jz      @f
+        DEBUGF 1,'K : XHCI: cannot mark the device as a hub, code %d\n',eax
+@@:
+        lea     ecx, [esi+xhci_controller.CmdLock-XCD]
+        invoke  MutexUnlock
+.nothing:
+        popad
+        ret
+endp

drivers/usb/xhci/xhci_ss.inc

@@ -0,0 +1,135 @@
+; SuperSpeed support for the xHCI driver, included from xhci.asm when
+; XHCI_SUPERSPEED is enabled. The pieces too small to live here - the port
+; map, the Slot Context speed, the 512-byte default control endpoint, the
+; Max Burst field of the Endpoint Context - sit in the other files under
+; "if XHCI_SUPERSPEED" guards; this file keeps the self-contained procedures.
+;
+; The kernel USB stack knows nothing about SuperSpeed, so it never learns
+; about it: a SuperSpeed device is announced as a high-speed one, which also
+; keeps the Transaction Translator logic away from it. The raw PORTSC speed
+; identifier travels in xhci_pipe.PSIV instead and reaches everything that
+; really needs it. SuperSpeed hubs are not supported: the kernel only speaks
+; the USB2 hub protocol, so a device behind a USB3 hub is served by the USB2
+; half of that hub at high speed.
+
+; =============================================================================
+; ============================ SuperSpeed ports ===============================
+; =============================================================================
+
+; Called from xhci_new_port for a SuperSpeed root port instead of the USB2
+; reset entry point: a SuperSpeed link trains itself and the port enables
+; itself on connect, reporting the speed with no reset at all. When the link
+; is already up, skip straight to the reset recovery stage, after which
+; xhci_port_init picks the device up; otherwise try a warm reset.
+; in: esi -> usb_controller, ecx = port index.
+proc xhci_ss_start_port
+        push    eax edx
+        and     [esi+usb_controller.ResettingHub], 0
+        mov     [esi+usb_controller.ResettingPort], cl
+        push    ecx
+        invoke  GetTimerTicks
+        pop     ecx
+        mov     [esi+usb_controller.ResetTime], eax
+        mov     [esi+xhci_controller.ResetStart-XCD], eax
+        call    xhci_port_reg
+        mov     eax, [edx]
+        test    al, XHCI_PORT_PED
+        jz      .warm_reset
+        mov     [esi+usb_controller.ResettingStatus], 2
+        DEBUGF 1,'K : XHCI %x: SS port %d link is up, PORTSC=%x\n',esi,ecx,eax
+        jmp     .nothing
+.warm_reset:
+; Connected, but the link has not come up. The Port Reset bit is set while a
+; warm reset is running, so the ordinary reset-wait machinery in
+; xhci_port_reset_done serves this case unchanged; the extra WRC change bit is
+; acknowledged there together with the usual ones.
+        mov     dword [edx], XHCI_PORT_PP + XHCI_PORT_WPR
+        mov     [esi+usb_controller.ResettingStatus], 1
+        DEBUGF 1,'K : XHCI %x: warm-resetting port %d, PORTSC=%x\n',esi,ecx,eax
+.nothing:
+        pop     edx eax
+        ret
+endp
+
+; The kernel restarts pending root ports through the NewPortReset entry of
+; usb_hardware_func directly, bypassing xhci_new_port; SuperSpeed ports must
+; not get the USB2 reset on that path either, so the entry points here.
+; in: esi -> usb_controller, ecx = port index.
+proc xhci_ss_port_reset
+        bt      [esi+xhci_controller.SSPortMask-XCD], ecx
+        jnc     xhci_new_port.reset
+        call    xhci_ss_start_port
+        ret
+endp
+
+; =============================================================================
+; ===================== Endpoint Companion descriptors ========================
+; =============================================================================
+
+; Reads the SuperSpeed Endpoint Companion descriptor of a bulk endpoint and
+; keeps its bMaxBurst for the Endpoint Context, allowing bursts of up to
+; bMaxBurst+1 packets per transaction. Zero (no bursts) is always legal, so
+; any failure just leaves the field alone. Only bulk endpoints are considered:
+; a nonzero burst on a periodic endpoint would also require the Max ESIT
+; Payload fields, and real interrupt endpoints do not burst anyway.
+; Relies on the stack frame of usb_open_pipe exactly like xhci_init_pipe does:
+; [ebp+8] is the config pipe of the device.
+; in: ebx -> xhci_pipe of a non-control endpoint being opened.
+proc xhci_ss_get_burst
+        push    eax ecx edx edi
+; 1. Only SuperSpeed bulk endpoints are of interest.
+        cmp     [ebx+xhci_pipe.PSIV], XHCI_PSI_SS
+        jb      .nothing
+        mov     eax, [ebx+xhci_pipe.EpType]
+        and     al, 3
+        cmp     al, 2                   ; bulk endpoint types are 2 and 6
+        jnz     .nothing
+; 2. Find the configuration data kept by the kernel behind the device
+; descriptor; it holds the endpoint descriptors with their companions.
+        mov     eax, [ebp+8]
+        mov     eax, [eax+usb_pipe.DeviceData]
+        test    eax, eax
+        jz      .nothing
+        movzx   edx, [eax+usb_device_data.DeviceDescrSize]
+        mov     edi, [eax+usb_device_data.ConfigDataSize]
+        lea     ecx, [eax+edx+usb_device_data.DeviceDescriptor]
+        add     edi, ecx                ; edi -> end of the configuration data
+; 3. Reconstruct bEndpointAddress from the Device Context Index.
+        mov     edx, [ebx+xhci_pipe.DCI]
+        shr     edx, 1
+        test    [ebx+xhci_pipe.Flags], XHCI_PIPE_IN
+        jz      @f
+        or      dl, 80h
+@@:
+; 4. Walk the descriptors to the one of this endpoint.
+.walk:
+        lea     eax, [ecx+2]
+        cmp     eax, edi
+        ja      .nothing
+        movzx   eax, byte [ecx]         ; bLength
+        test    eax, eax
+        jz      .nothing                ; malformed data, stop
+        cmp     byte [ecx+1], 5         ; an endpoint descriptor?
+        jnz     .advance
+        cmp     [ecx+2], dl             ; bEndpointAddress
+        jnz     .advance
+; 5. The companion descriptor (type 30h) immediately follows its endpoint.
+        add     ecx, eax
+        lea     eax, [ecx+3]
+        cmp     eax, edi
+        ja      .nothing
+        cmp     byte [ecx+1], 30h       ; SuperSpeed Endpoint Companion
+        jnz     .nothing
+        movzx   eax, byte [ecx+2]       ; bMaxBurst
+        cmp     eax, 15
+        ja      .nothing
+        mov     [ebx+xhci_pipe.MaxBurst], eax
+        DEBUGF 1,'K : XHCI: slot %d dci %d bursts %d+1 packets\n',[ebx+xhci_pipe.SlotId],[ebx+xhci_pipe.DCI],eax
+        jmp     .nothing
+.advance:
+        add     ecx, eax
+        jmp     .walk
+.nothing:
+        pop     edi edx ecx eax
+        ret
+endp

kernel/trunk/blkdev/disk.inc

@@ -842,8 +842,7 @@ disk_scan_partitions:
         cmp     ebx, mbr_buffer
         jz      .release_partition_buffer
 ; 15b. If we have allocated it, free it.
-        xchg    eax, ebx
-        call    free
+        stdcall kernel_free, ebx
         jmp     .nothing
 ; 15c. Otherwise, release reference.
 .release_partition_buffer:

kernel/trunk/const.inc

@@ -183,6 +183,8 @@ EFLAGS_ID      =    0x200000  ; id flag
 IRQ_PIC        =    0
 IRQ_APIC       =    1
 
+IO_MAP_SIZE    =    65536/8 ; 8192 or 2 PAGE_SIZE
+
 struct  TSS
         _back       rw 2
         _esp0       rd 1
@@ -290,6 +292,9 @@ sys_proc            = OS_BASE + 0x008E000
 
 SLOT_BASE           = OS_BASE + 0x0090000
 
+APPDATA_IDLE_THREAD = SLOT_BASE + sizeof.APPDATA*1 ; poiner to IDLE thread
+APPDATA_OS_THREAD   = SLOT_BASE + sizeof.APPDATA*2 ; poiner to OS thread
+
 VGABasePtr          = OS_BASE + 0x00A0000
 
 virtual at            OS_BASE + 0x05FFF80
@@ -570,7 +575,7 @@ struct  APPDATA
         ipc_start       dd ?            ;+160
         ipc_size        dd ?            ;+164
         occurred_events dd ?            ;+168  ; mask which accumulates occurred events
-        debugger_slot   dd ?            ;+172  ; index in SLOT_BASE array
+        debugger_thread dd ?            ;+172  ; pointer to APPDATA or 0
         terminate_protection dd ?       ;+176
         keyboard_mode   db ?            ;+180
                         rb 3

kernel/trunk/core/debug.inc

@@ -8,8 +8,8 @@
 
 ; diamond, 2006
 sys_debug_services:
-        cmp     ebx, 9
-        ja      @f
+        cmp     ebx, sys_debug_services_table.count
+        jae     @f
         jmp     dword [sys_debug_services_table + ebx*4]
 @@:
         ret
@@ -26,7 +26,9 @@ sys_debug_services_table:
         dd      debug_write_process_memory
         dd      debug_terminate
         dd      debug_set_drx
+.count  = ($ - sys_debug_services_table)/4
 endg
+
 debug_set_event_data:
 ; in: ecx = pointer
 ; destroys eax
@@ -34,22 +36,20 @@ debug_set_event_data:
         mov     [eax + APPDATA.dbg_event_mem], ecx
         ret
 
-get_debuggee_slot:
+get_debuggee_thread:
 ; in: ecx=PID
 ; out: CF=1 if error
-;      CF=0 and eax=slot*0x20 if ok
+;      CF=0 and eax = pointer to APPDATA if ok
 ; out: interrupts disabled
         cli
         mov     eax, ecx
-        call    pid_to_slot
-        ;call     pid_to_appdata
+        call    pid_to_thread
         test    eax, eax
         jz      .ret_bad
-        shl     eax, BSF sizeof.APPDATA
+
         push    ebx
-        mov     ebx, [current_slot_idx]
-        cmp     [SLOT_BASE + eax + APPDATA.debugger_slot], ebx
-        ;cmp     [eax + APPDATA.debugger_slot], ebx
+        mov     ebx, [current_slot]
+        cmp     [eax + APPDATA.debugger_thread], ebx
         pop     ebx
         jnz     .ret_bad
 ;       clc     ; automatically
@@ -61,10 +61,9 @@ get_debuggee_slot:
 debug_detach:
 ; in: ecx=pid
 ; destroys eax,ebx
-        call    get_debuggee_slot
+        call    get_debuggee_thread
         jc      .ret
-        and     dword [eax + SLOT_BASE + APPDATA.debugger_slot], 0
-        ;and     dword [eax + APPDATA.debugger_slot], 0
+        and     dword [eax + APPDATA.debugger_thread], 0
         call    do_resume
 .ret:
         sti
@@ -72,40 +71,27 @@ debug_detach:
 
 debug_terminate:
 ; in: ecx=pid
-        call    get_debuggee_slot
+        call    get_debuggee_thread
         jc      debug_detach.ret
         mov     ecx, eax
 
-        shr     ecx, BSF sizeof.APPDATA
-        ;movzx   ecx, ch ; del when sysfn_term... will using APPDATA
-
-;        push    2
-;        pop     ebx
-        mov     edx, esi ; what?
+        movzx   ecx, ch ; del when sysfn_term... will using APPDATA
         jmp     sysfn_terminate
 
 debug_suspend:
 ; in: ecx=pid
 ; destroys eax,ecx
-; { Patch by Coldy (rev. 7125), reason: http://board.kolibrios.org/viewtopic.php?f=1&t=1712&p=75957#p75957
-;        cli
-;        mov     eax, ecx
-;        call    pid_to_slot
-;        shl     eax, 5
-;        jz      .ret
-        call    get_debuggee_slot
+        call    get_debuggee_thread
         jc      .ret
-; } End patch
-        mov     cl, [SLOT_BASE + eax + APPDATA.state] ; process state
-        ;mov     cl, [eax + APPDATA.state] ; process state
+
+        mov     cl, [eax + APPDATA.state] ; process state
         test    cl, cl
         jz      .1
         cmp     cl, TSTATE_WAITING
         jnz     .ret
         mov     cl, TSTATE_WAIT_SUSPENDED
 .2:
-        mov     [SLOT_BASE + eax + APPDATA.state], cl
-        ;mov     [eax + APPDATA.state], cl
+        mov     [eax + APPDATA.state], cl
 .ret:
         sti
         ret
@@ -114,16 +100,14 @@ debug_suspend:
         jmp     .2
 
 do_resume:
-        mov     cl, [SLOT_BASE + eax + APPDATA.state]
-        ;mov     cl, [eax + APPDATA.state]
+        mov     cl, [eax + APPDATA.state]
         cmp     cl, TSTATE_RUN_SUSPENDED
         jz      .1
         cmp     cl, TSTATE_WAIT_SUSPENDED
         jnz     .ret
         mov     cl, TSTATE_WAITING
 .2:
-        mov     [SLOT_BASE + eax + APPDATA.state], cl
-        ;mov     [eax + APPDATA.state], cl
+        mov     [eax + APPDATA.state], cl
 .ret:
         ret
 .1:
@@ -133,11 +117,9 @@ do_resume:
 debug_resume:
 ; in: ecx=pid
 ; destroys eax,ebx
-        cli
-        mov     eax, ecx
-        call    pid_to_slot
-        shl     eax, BSF sizeof.APPDATA
-        jz      .ret
+        call    get_debuggee_thread
+        jc      .ret
+
         call    do_resume
 .ret:
         sti
@@ -160,21 +142,15 @@ debug_getcontext:
         inc     ebx                 ; 1 - get sse context
                                   ; TODO legacy 32-bit FPU/MMX context
 .std_ctx:
-        call    get_debuggee_slot
+        call    get_debuggee_thread
         jc      .ret
 
-        shr     eax, BSF sizeof.APPDATA
-        ;movzx   ebp, ah
-
         cmp     eax, [fpu_owner]
-        ;cmp     ebp, [fpu_owner]
         jne     @f
         inc     bh                ; set swap context flag
 @@:
-        shl     eax, BSF sizeof.APPDATA
         mov     edi, esi
-        mov     eax, [SLOT_BASE + eax + APPDATA.pl0_stack]
-        ;mov     eax, [eax + APPDATA.pl0_stack]
+        mov     eax, [eax + APPDATA.pl0_stack]
         lea     esi, [eax + RING0_STACK_SIZE]
 
 .ring0:
@@ -240,11 +216,10 @@ debug_setcontext:
         cmp     edx, 28h
         jnz     .ret
 
-        call    get_debuggee_slot
+        call    get_debuggee_thread
         jc      .stiret
-;        mov     esi, edx
-        mov     eax, [eax + SLOT_BASE+APPDATA.pl0_stack]
-        ;mov     eax, [eax + APPDATA.pl0_stack]
+
+        mov     eax, [eax + APPDATA.pl0_stack]
         lea     edi, [eax + RING0_STACK_SIZE]
 
 .ring0:
@@ -277,15 +252,16 @@ debug_setcontext:
         ret
 
 debug_set_drx:
-        call    get_debuggee_slot
+        call    get_debuggee_thread
         jc      .errret
+
         mov     ebp, eax
-        lea     eax, [eax + SLOT_BASE + APPDATA.dbg_regs]
-        ;lea     eax, [eax + APPDATA.dbg_regs]
+        lea     eax, [eax + APPDATA.dbg_regs]
 ; [eax]=dr0, [eax+4]=dr1, [eax+8]=dr2, [eax+C]=dr3
 ; [eax+10]=dr7
         cmp     esi, OS_BASE
         jae     .errret
+
         cmp     dl, 3
         ja      .errret
         mov     ecx, dr7
@@ -306,10 +282,7 @@ debug_set_drx:
         btr     dword [eax + 10h], edx    ; clear L<i> bit
         test    byte [eax + 10h], 55h
         jnz     .okret
-;        imul    eax, ebp, tss_step/32
-;        and     byte [eax + tss_data + TSS._trap], not 1
-        and     [SLOT_BASE + ebp + APPDATA.dbg_state], not 1
-        ;and     [ebp + APPDATA.dbg_state], not 1
+        and     [ebp + APPDATA.dbg_state], not 1
 .okret:
         and     dword [esp + SYSCALL_STACK.eax], 0
         sti
@@ -354,10 +327,7 @@ debug_set_drx:
         not     ebx
         and     [eax + 10h+2], bx
         or      [eax + 10h+2], dx         ; set R/W and LEN fields
-;        imul    eax, ebp, tss_step/32
-;        or      byte [eax + tss_data + TSS._trap], 1
-        or      [SLOT_BASE + ebp + APPDATA.dbg_state], 1
-        ;or      [ebp + APPDATA.dbg_state], 1
+        or      [ebp + APPDATA.dbg_state], 1
         jmp     .okret
 
 debug_read_process_memory:
@@ -368,10 +338,9 @@ debug_read_process_memory:
 ; esi=address in debuggee
 ; out: [esp+36]=sizeof(read)
 ; destroys all
-        call    get_debuggee_slot
+        call    get_debuggee_thread
         jc      .err
-        shr     eax, BSF sizeof.APPDATA
-        ;movzx   eax,ah
+
         mov     ecx, edi
         call    read_process_memory
         sti
@@ -389,10 +358,9 @@ debug_write_process_memory:
 ; esi=address in debuggee
 ; out: [esp+36]=sizeof(write)
 ; destroys all
-        call    get_debuggee_slot
+        call    get_debuggee_thread
         jc      debug_read_process_memory.err
-        shr     eax, BSF sizeof.APPDATA
-        ;movzx   eax,ah
+
         mov     ecx, edi
         call    write_process_memory
         sti
@@ -400,7 +368,7 @@ debug_write_process_memory:
         ret
 
 debugger_notify:
-; in: eax=debugger slot
+; in: eax=debugger thread
 ;     ecx=size of debug message
 ;     [esp+4]..[esp+4+ecx]=message
 ; interrupts must be disabled!
@@ -410,9 +378,7 @@ debugger_notify:
         mov     edi, [timer_ticks]
         add     edi, 500        ; 5 sec timeout
 .1:
-        mov     eax, ebp
-        shl     eax, BSF sizeof.APPDATA
-        mov     esi, [SLOT_BASE + eax + APPDATA.dbg_event_mem]
+        mov     esi, [ebp + APPDATA.dbg_event_mem]
         test    esi, esi
         jz      .ret
 ; read buffer header
@@ -466,8 +432,6 @@ debugger_notify:
         pop     edx
         call    write_process_memory
 ; new debug event
-        mov     eax, ebp
-        shl     eax, BSF sizeof.APPDATA
-        or      [SLOT_BASE + eax + APPDATA.occurred_events], EVENT_DEBUG
+        or      [ebp + APPDATA.occurred_events], EVENT_DEBUG
 .ret:
         ret

kernel/trunk/core/fpu.inc

@@ -150,10 +150,10 @@ init_avx512:
 
         ret
 
-; param
-;  eax= 512 bytes memory area aligned on a 16-byte boundary
-
 align 4
+; @brief Export function - Save FPU context in buffer
+; @param eax - 512 bytes memory area aligned on a 16-byte boundary
+; @return  not return
 fpu_save:
         push    ecx
         push    esi
@@ -166,7 +166,7 @@ fpu_save:
         mov     edi, eax
 
         mov     ecx, [fpu_owner]
-        mov     esi, [current_slot_idx]
+        mov     esi, [current_slot]
         cmp     ecx, esi
         jne     .save
 
@@ -175,14 +175,11 @@ fpu_save:
 .save:
         mov     [fpu_owner], esi
 
-        shl     ecx, BSF sizeof.APPDATA
-        mov     eax, [SLOT_BASE + ecx + APPDATA.fpu_state]
-
+        mov     eax, [ecx + APPDATA.fpu_state]
         call    save_context
 
 ; first 512 bytes of XSAVE area have the same format as FXSAVE
-        shl     esi, BSF sizeof.APPDATA
-        mov     esi, [SLOT_BASE + esi + APPDATA.fpu_state]
+        mov     esi, [esi + APPDATA.fpu_state]
         mov     ecx, 512/4
         cld
         rep movsd
@@ -194,14 +191,16 @@ fpu_save:
         pop     ecx
         ret
 
+; @brief Export function - Get size from AVX context
+; @return  size of buffer
 avx_save_size:
         mov     eax, [xsave_area_size]
         ret
 
-; param
-;  eax= avx_save_size() bytes memory area aligned on a 64-byte boundary
-
 align 4
+; @brief Export function - Save AVX context in buffer
+; @param eax - avx_save_size() bytes memory area aligned on a 64-byte boundary
+; @return  not return
 avx_save:
         push    ecx
         push    esi
@@ -214,7 +213,7 @@ avx_save:
         mov     edi, eax
 
         mov     ecx, [fpu_owner]
-        mov     esi, [current_slot_idx]
+        mov     esi, [current_slot]
         cmp     ecx, esi
         jne     .save
 
@@ -223,13 +222,10 @@ avx_save:
 .save:
         mov     [fpu_owner], esi
 
-        shl     ecx, BSF sizeof.APPDATA
-        mov     eax, [SLOT_BASE + ecx + APPDATA.fpu_state]
-
+        mov     eax, [ecx + APPDATA.fpu_state]
         call    save_context
 
-        shl     esi, BSF sizeof.APPDATA
-        mov     esi, [SLOT_BASE + esi + APPDATA.fpu_state]
+        mov     esi, [esi + APPDATA.fpu_state]
         mov     ecx, [xsave_area_size]
         add     ecx, 3
         shr     ecx, 2
@@ -264,9 +260,13 @@ save_fpu_context:
 
 
 align 4
+; @brief Export function - Restore FPU context
+; @param eax - 512 bytes memory area aligned on a 16-byte boundary
+; @return  not return
 fpu_restore:
         push    ecx
         push    esi
+        push    edi
 
         mov     esi, eax
 
@@ -274,7 +274,7 @@ fpu_restore:
         cli
 
         mov     ecx, [fpu_owner]
-        mov     eax, [current_slot_idx]
+        mov     eax, [current_slot]
         cmp     ecx, eax
         jne     .copy
 
@@ -284,6 +284,8 @@ fpu_restore:
 
         fxrstor [esi]
         popfd
+
+        pop     edi
         pop     esi
         pop     ecx
         ret
@@ -291,24 +293,31 @@ fpu_restore:
         fnclex                  ;fix possible problems
         frstor  [esi]
         popfd
+
+        pop     edi
         pop     esi
         pop     ecx
         ret
 .copy:
-        shl     eax, BSF sizeof.APPDATA
-        mov     edi, [SLOT_BASE + eax + APPDATA.fpu_state]
+        mov     edi, [eax + APPDATA.fpu_state]
         mov     ecx, 512/4
         cld
         rep movsd
         popfd
+
+        pop     edi
         pop     esi
         pop     ecx
         ret
 
 align 4
+; @brief Export function - Restore AVX context
+; @param eax - avx_save_size() bytes memory area aligned on a 64-byte boundary
+; @return  not return
 avx_restore:
         push    ecx
         push    esi
+        push    edi
 
         mov     esi, eax
 
@@ -316,7 +325,7 @@ avx_restore:
         cli
 
         mov     ecx, [fpu_owner]
-        mov     eax, [current_slot_idx]
+        mov     eax, [current_slot]
         cmp     ecx, eax
         jne     .copy
 
@@ -329,6 +338,8 @@ avx_restore:
         xrstor  [esi]
         pop     edx
         popfd
+
+        pop     edi
         pop     esi
         pop     ecx
         ret
@@ -338,6 +349,8 @@ avx_restore:
 
         fxrstor [esi]
         popfd
+
+        pop     edi
         pop     esi
         pop     ecx
         ret
@@ -345,18 +358,21 @@ avx_restore:
         fnclex                  ;fix possible problems
         frstor  [esi]
         popfd
+
+        pop     edi
         pop     esi
         pop     ecx
         ret
 .copy:
-        shl     eax, BSF sizeof.APPDATA
-        mov     edi, [SLOT_BASE + eax + APPDATA.fpu_state]
+        mov     edi, [eax + APPDATA.fpu_state]
         mov     ecx, [xsave_area_size]
         add     ecx, 3
         shr     ecx, 2
         cld
         rep movsd
         popfd
+
+        pop     edi
         pop     esi
         pop     ecx
         ret
@@ -370,21 +386,19 @@ except_7:                  ;#NM exception handler
         mov     es, ax
 
         mov     ebx, [fpu_owner]
-        cmp     ebx, [current_slot_idx]
+        cmp     ebx, [current_slot]
         je      .exit
 
-        shl     ebx, BSF sizeof.APPDATA
-        mov     eax, [SLOT_BASE + ebx + APPDATA.fpu_state]
+        mov     eax, [ebx + APPDATA.fpu_state]
         bt      [cpu_caps + (CAPS_OSXSAVE/32)*4], CAPS_OSXSAVE mod 32
         jnc     .no_xsave
         mov     ecx, eax
         mov     eax, [xsave_eax]
         mov     edx, [xsave_edx]
         xsave   [ecx]
-        mov     ebx, [current_slot_idx]
+        mov     ebx, [current_slot]
         mov     [fpu_owner], ebx
-        shl     ebx, BSF sizeof.APPDATA
-        mov     ecx, [SLOT_BASE + ebx + APPDATA.fpu_state]
+        mov     ecx, [ebx + APPDATA.fpu_state]
         xrstor  [ecx]
 .exit:
         restore_ring3_context
@@ -394,24 +408,22 @@ except_7:                  ;#NM exception handler
         jnc     .no_SSE
 
         fxsave  [eax]
-        mov     ebx, [current_slot_idx]
+        mov     ebx, [current_slot]
         mov     [fpu_owner], ebx
-        shl     ebx, BSF sizeof.APPDATA
-        mov     eax, [SLOT_BASE + ebx + APPDATA.fpu_state]
+        mov     eax, [ebx + APPDATA.fpu_state]
         fxrstor [eax]
         restore_ring3_context
         iret
 
 .no_SSE:
         fnsave  [eax]
-        mov     ebx, [current_slot_idx]
+        mov     ebx, [current_slot]
         mov     [fpu_owner], ebx
-        shl     ebx, BSF sizeof.APPDATA
-        mov     eax, [SLOT_BASE + ebx + APPDATA.fpu_state]
+        mov     eax, [ebx + APPDATA.fpu_state]
         frstor  [eax]
         restore_ring3_context
         iret
 
 iglobal
-  fpu_owner dd 2
+  fpu_owner dd APPDATA_OS_THREAD
 endg

kernel/trunk/core/memory.inc

@@ -605,7 +605,7 @@ proc page_fault_handler
 
         mov     edi, [.err_addr]
         and     edi, -PAGE_SIZE
-        mov     ecx, 1024
+        mov     ecx, PAGE_SIZE/4
         xor     eax, eax
        ;cld     ;caller is duty for this
         rep stosd
@@ -664,7 +664,7 @@ proc page_fault_handler
         cmp     ebx, tss._io_map_0
         jb      .fail
 
-        cmp     ebx, tss._io_map_0+8192
+        cmp     ebx, tss._io_map_0 + IO_MAP_SIZE
         jae     .fail
 
 ; io permission map
@@ -695,7 +695,7 @@ proc page_fault_handler
 endp
 
 ; returns number of mapped bytes
-proc map_mem_ipc stdcall, lin_addr:dword,slot:dword,\
+proc map_mem_ipc stdcall, lin_addr:dword,thread:dword,\
                       ofs:dword,buf_size:dword,req_access:dword
         locals
              count   dd ?
@@ -706,9 +706,8 @@ proc map_mem_ipc stdcall, lin_addr:dword,slot:dword,\
         cmp     [buf_size], 0
         jz      .exit
 
-        mov     eax, [slot]
-        shl     eax, BSF sizeof.APPDATA
-        mov     eax, [SLOT_BASE + eax + APPDATA.process]
+        mov     eax, [thread]
+        mov     eax, [eax + APPDATA.process]
         test    eax, eax
         jz      .exit
 
@@ -732,7 +731,7 @@ proc map_mem_ipc stdcall, lin_addr:dword,slot:dword,\
         shr     edx, 12
         and     edx, 0x3FF
 .map:
-        stdcall safe_map_page, [slot], [req_access], [ofs]
+        stdcall safe_map_page, [process], [req_access], [ofs]
         jnc     .exit
         add     [count], PAGE_SIZE
         add     [ofs], PAGE_SIZE
@@ -758,7 +757,7 @@ proc map_mem_ipc stdcall, lin_addr:dword,slot:dword,\
         ret
 endp
 
-proc map_memEx stdcall, lin_addr:dword,slot:dword,\
+proc map_memEx stdcall, lin_addr:dword,thread:dword,\
                         ofs:dword,buf_size:dword,req_access:dword
         locals
              count   dd ?
@@ -769,9 +768,8 @@ proc map_memEx stdcall, lin_addr:dword,slot:dword,\
         cmp     [buf_size], 0
         jz      .exit
 
-        mov     eax, [slot]
-        shl     eax, BSF sizeof.APPDATA
-        mov     eax, [SLOT_BASE + eax + APPDATA.process]
+        mov     eax, [thread]
+        mov     eax, [eax + APPDATA.process]
         test    eax, eax
         jz      .exit
 
@@ -795,7 +793,7 @@ proc map_memEx stdcall, lin_addr:dword,slot:dword,\
         shr     edx, 12
         and     edx, 0x3FF
 .map:
-        stdcall safe_map_page, [slot], [req_access], [ofs]
+        stdcall safe_map_page, [process], [req_access], [ofs]
         jnc     .exit
         add     [count], PAGE_SIZE
         add     [ofs], PAGE_SIZE
@@ -826,7 +824,7 @@ endp
 ; in: edi = linear address to map
 ; out: CF cleared <=> failed
 ; destroys: only eax
-proc safe_map_page stdcall, slot:dword, req_access:dword, ofs:dword
+proc safe_map_page stdcall, process:dword, req_access:dword, ofs:dword
         mov     eax, [esi + edx*4]
         test    al, PG_READ
         jz      .not_present
@@ -861,9 +859,7 @@ proc safe_map_page stdcall, slot:dword, req_access:dword, ofs:dword
         cli
         cld
         push    ebx ecx
-        mov     eax, [slot]
-        shl     eax, BSF sizeof.APPDATA
-        mov     eax, [SLOT_BASE + eax + APPDATA.process]
+        mov     eax, [process]
         mov     eax, [eax + PROC.dlls_list_ptr]
         test    eax, eax
         jz      .no_hdll
@@ -953,7 +949,7 @@ sys_IPC:
 
 proc sys_ipc_send stdcall, PID:dword, msg_addr:dword, msg_size:dword
            locals
-             dst_slot   dd ?
+             dst_thread dd ?
              dst_offset dd ?
              buf_size   dd ?
              used_buf   dd ?
@@ -963,13 +959,12 @@ proc sys_ipc_send stdcall, PID:dword, msg_addr:dword, msg_size:dword
         cli
 
         mov     eax, [PID]
-        call    pid_to_slot
+        call    pid_to_thread
         test    eax, eax
         jz      .no_pid
 
-        mov     [dst_slot], eax
-        shl     eax, BSF sizeof.APPDATA
-        mov     edi, [SLOT_BASE + eax + APPDATA.ipc_start] ;is ipc area defined?
+        mov     [dst_thread], eax
+        mov     edi, [eax + APPDATA.ipc_start] ;is ipc area defined?
         test    edi, edi
         jz      .no_ipc_area
 
@@ -977,7 +972,7 @@ proc sys_ipc_send stdcall, PID:dword, msg_addr:dword, msg_size:dword
         and     ebx, 0xFFF
         mov     [dst_offset], ebx
 
-        mov     esi, [SLOT_BASE + eax + APPDATA.ipc_size]
+        mov     esi, [eax + APPDATA.ipc_size]
         mov     [buf_size], esi
 
         mov     ecx, [ipc_tmp]
@@ -990,7 +985,7 @@ proc sys_ipc_send stdcall, PID:dword, msg_addr:dword, msg_size:dword
         pop     edi esi
 @@:
         mov     [used_buf], ecx
-        stdcall map_mem_ipc, ecx, [dst_slot], \
+        stdcall map_mem_ipc, ecx, [dst_thread], \
                 edi, esi, PG_SWR
 
         mov     edi, [dst_offset]
@@ -1039,9 +1034,8 @@ proc sys_ipc_send stdcall, PID:dword, msg_addr:dword, msg_size:dword
         mov     [page_tabs + ebx*4], eax
         invlpg  [edx]
 
-        mov     eax, [dst_slot]
-        shl     eax, BSF sizeof.APPDATA
-        or      [SLOT_BASE + eax + APPDATA.occurred_events], EVENT_IPC
+        mov     eax, [dst_thread]
+        or      [eax + APPDATA.occurred_events], EVENT_IPC
         push    0
         jmp     .ret
 .no_pid:

kernel/trunk/core/sched.inc

@@ -139,14 +139,13 @@ do_change_task:
         mov     eax, [xsave_eax]
         mov     edx, [xsave_edx]
         xsave   [ecx]
-        mov     ecx, [current_slot_idx]
-        mov     [fpu_owner], ecx
         mov     ecx, [current_slot]
+        mov     [fpu_owner], ecx
         mov     ecx, [ecx + APPDATA.fpu_state]
         xrstor  [ecx]
   .no_xsave:
       ; set CR0.TS
-        cmp     bh, byte[fpu_owner] ;bh == incoming task (new)
+        cmp     ebx, [fpu_owner]    ;ebx == incoming task (new)
         clts                        ;clear a task switch flag
         je      @f
         mov     eax, cr0            ;and set it again if the owner

kernel/trunk/core/sys32.inc

@@ -165,7 +165,7 @@ exc_c:                            ; exceptions (all but 7th - #NM)
         jnz     IRetToUserHook
   @@:
         cli
-        mov     eax, [esi + APPDATA.debugger_slot]
+        mov     eax, [esi + APPDATA.debugger_thread]
         test    eax, eax
         jnz     .debug
 ; not debuggee => say error and terminate
@@ -514,28 +514,30 @@ destroy_thread:
 @@:
         mov     eax, [esi + APPOBJ.fd]
         test    eax, eax
-        jz      @F
+        jz      @f
 
         cmp     eax, esi
-        je      @F
+        je      @f
 
         push    esi
         call    [eax + APPOBJ.destroy]
            DEBUGF 1,"%s",msg_obj_destroy
         pop     esi
-        jmp     @B
+        jmp     @b
 @@:
         mov     esi, [.slot]
+        shl     esi, BSF sizeof.APPDATA
+        add     esi, SLOT_BASE
         cmp     [fpu_owner], esi ; if user fpu last -> fpu user = 2
-        jne     @F
+        jne     @f
 
-        mov     [fpu_owner], 2
-        mov     eax, [SLOT_BASE + sizeof.APPDATA*2 + APPDATA.fpu_state]
+        mov     [fpu_owner], APPDATA_OS_THREAD
+        mov     eax, [APPDATA_OS_THREAD + APPDATA.fpu_state]
         clts
         bt      [cpu_caps], CAPS_SSE
         jnc     .no_SSE
         fxrstor [eax]
-        jmp     @F
+        jmp     @f
 .no_SSE:
         fnclex
         frstor  [eax]
@@ -544,8 +546,8 @@ destroy_thread:
         mov     [KEY_COUNT], byte 0    ; empty keyboard buffer
         mov     [BTN_COUNT], byte 0    ; empty button buffer
 
-
 ; remove defined hotkeys
+        mov     esi, [.slot]
         mov     eax, hotkey_list
 .loop:
         cmp     [eax+8], esi
@@ -638,7 +640,7 @@ destroy_thread:
         pushad
         mov     edi, esi
         shl     edi, BSF sizeof.APPDATA
-        mov     eax, [SLOT_BASE + edi + APPDATA.debugger_slot]
+        mov     eax, [SLOT_BASE + edi + APPDATA.debugger_thread]
         test    eax, eax
         jz      .nodebug
         movi    ecx, 8
@@ -666,12 +668,12 @@ destroy_thread:
         add     edi, SLOT_BASE
 
         mov     eax, [edi + APPDATA.io_map]
-        cmp     eax, [SLOT_BASE + sizeof.APPDATA+APPDATA.io_map]
+        cmp     eax, [APPDATA_IDLE_THREAD + APPDATA.io_map]
         je      @F
         call    free_page
 @@:
         mov     eax, [edi+APPDATA.io_map+4]
-        cmp     eax, [SLOT_BASE+sizeof.APPDATA+APPDATA.io_map+4]
+        cmp     eax, [APPDATA_IDLE_THREAD + APPDATA.io_map+4]
         je      @F
         call    free_page
 @@:
@@ -712,12 +714,12 @@ destroy_thread:
         call    free_cd_channel
         and     [cd_status], 0
 @@:
-        pop     esi
         cmp     [bgrlockpid], esi
         jnz     @f
         and     [bgrlockpid], 0
         and     [bgrlock], 0
 @@:
+        pop     esi
 
         pusha                 ; remove all port reservations
         mov     edx, esi
@@ -778,14 +780,15 @@ destroy_thread:
 
         mov     edi, esi ; do not run this process slot
         shl     edi, BSF sizeof.APPDATA
-        mov     [SLOT_BASE + edi + APPDATA.state], TSTATE_FREE
+        add     edi, SLOT_BASE
+        mov     [edi + APPDATA.state], TSTATE_FREE
 ; debugger test - terminate all debuggees
         mov     eax, 2
-        mov     ecx, SLOT_BASE + 2*sizeof.APPDATA + APPDATA.debugger_slot
+        mov     ecx, APPDATA_OS_THREAD + APPDATA.debugger_thread
 .xd0:
         cmp     eax, [thread_count]
         ja      .xd1
-        cmp     dword [ecx], esi
+        cmp     dword [ecx], edi
         jnz     @f
         and     dword [ecx], 0
         pushad

kernel/trunk/core/taskman.inc

@@ -537,21 +537,21 @@ pid_to_slot:
 align 4
 proc read_process_memory
 ;Input:
-;  eax - process slot
+;  eax - process thread
 ;  ecx - buffer address
 ;  edx - buffer size
 ;  esi - start address in other process
 ;Output:
 ;  eax - number of bytes read.
        locals
-         slot   dd ?
+         thread dd ?
          buff   dd ?
          r_count    dd ?
          offset dd ?
          tmp_r_cnt  dd ?
        endl
 
-        mov     [slot], eax
+        mov     [thread], eax
         mov     [buff], ecx
         and     [r_count], 0
         mov     [tmp_r_cnt], edx
@@ -577,7 +577,7 @@ proc read_process_memory
 
         push    ecx
         stdcall map_memEx, [proc_mem_map], \
-                [slot], ebx, ecx, PG_READ
+                [thread], ebx, ecx, PG_READ
         pop     ecx
 
         mov     esi, [offset]
@@ -607,7 +607,7 @@ endp
 align 4
 proc write_process_memory
 ;Input:
-;  eax - process slot
+;  eax - process thread
 ;  ecx - buffer address
 ;  edx - buffer size
 ;  esi - start address in other process
@@ -615,14 +615,14 @@ proc write_process_memory
 ;  eax - number of bytes written
 
        locals
-         slot   dd ?
+         thread dd ?
          buff   dd ?
          w_count    dd ?
          offset dd ?
          tmp_w_cnt  dd ?
        endl
 
-        mov     [slot], eax
+        mov     [thread], eax
         mov     [buff], ecx
         and     [w_count], 0
         mov     [tmp_w_cnt], edx
@@ -647,7 +647,7 @@ proc write_process_memory
         mov     ebx, [offset]
         push    ecx
         stdcall map_memEx, [proc_mem_map], \
-                [slot], ebx, ecx, PG_SWR
+                [thread], ebx, ecx, PG_SWR
         pop     ecx
 
         mov     edi, [offset]
@@ -906,7 +906,7 @@ common_app_entry:
         stdcall free_kernel_space, [ebp+APP_HDR.img_base]
         stdcall kernel_free, ebp
         mov     ebx, [current_slot]
-        cmp     [ebx + APPDATA.debugger_slot], 0
+        cmp     [ebx + APPDATA.debugger_thread], 0
         je      .exit
         mov     [ebx + APPDATA.state], TSTATE_RUN_SUSPENDED
         mov     [ebx + APPDATA.state], TSTATE_RUN_SUSPENDED
@@ -1054,8 +1054,8 @@ proc set_app_params stdcall,slot:dword, params:dword, flags:dword
 ; set if debuggee
         test    byte [flags], 1
         jz      .no_debug
-        mov     eax, [current_slot_idx]
-        mov     [SLOT_BASE + ebx + APPDATA.debugger_slot], eax
+        mov     eax, [current_slot]
+        mov     [SLOT_BASE + ebx + APPDATA.debugger_thread], eax
 .no_debug:
         mov     [SLOT_BASE + ebx + APPDATA.state], dl
         lea     edx, [SLOT_BASE + ebx]
@@ -1074,27 +1074,24 @@ get_curr_slot:
         mov     eax, [current_slot]
         ret
 
-pid_to_appdata:
-;Input:
-;  eax - pid of process
-;Output:
-;  eax - 0 - not found or pointer on APPDATA
+; @brief  Get pointer to APPDATA by PID
+; @param eax - TID of thread
+; @return  eax - 0 - not found or pointer on APPDATA
+pid_to_thread:
         push    ebx
         push    ecx
         mov     ebx, [thread_count]
-        shl     ebx, BSF sizeof.APPDATA ; multiply by size
         ; skip first process in the task table
-        mov     ecx, SLOT_BASE
-        add     ebx, ecx
+        mov     ecx, APPDATA_IDLE_THREAD
 .loop:
-        add     ecx, sizeof.APPDATA
         cmp     [ecx + APPDATA.state], TSTATE_FREE
         jz      @f  ;skip empty slots
         cmp     [ecx + APPDATA.tid], eax
         jz      .pid_found
 @@:
-        cmp     ecx, ebx
-        jb      .loop
+        add     ecx, sizeof.APPDATA
+        dec     ebx
+        jnz     .loop
 
         pop     ecx
         pop     ebx

kernel/trunk/core/timers.inc

@@ -48,7 +48,7 @@ endg
 
 ; This internal function acquires the lock for the global list.
 lock_timer_list:
-        mov     edx, [current_slot_idx]
+        mov     edx, [current_slot]
 @@:
         xor     eax, eax
         lock cmpxchg [timer_list_owner], edx

kernel/trunk/docs/sysfuncs.txt

@@ -4148,6 +4148,7 @@ Structure of block of data for folder entry (BDFE):
       and after archiving this bit is cleared - it can be useful
       for automatically creating of backup-archives as at writing
       this bit is usually set
+    * bit 6 (mask 0x40): this is a symbolic link (currently only set by the ext driver).
   * +4: dword: encoding, equals to field +8 in the information structure
   * +8: 4*byte: time of file creation
   * +12 = +0xC: 4*byte: date of file creation

kernel/trunk/fs/fat.inc

@@ -2405,11 +2405,12 @@ fat_Write:
 ; hd_extend_file can return three error codes: FAT table error, device error or disk full.
 ; First two cases are fatal errors, in third case we may write some data
         cmp     al, ERROR_DISK_FULL
-        jnz     @f
+        jnz     .noWrite
 ; correct number of bytes to write
         mov     ecx, [edi+28]
         cmp     ecx, ebx
         ja      .length_ok
+.noWrite:
         push    0
 .ret:
         pop     eax eax eax ecx ecx

kernel/trunk/gui/background.inc

@@ -193,7 +193,8 @@ align 4
 ;--------------------------------------
 align 4
 @@:
-        mov     eax, [current_slot_idx]
+        mov     eax, [current_slot]
+        mov     eax, [eax + APPDATA.tid]
         mov     [bgrlockpid], eax
         cmp     [img_background], static_background_data
         jz      .nomem
@@ -241,7 +242,8 @@ nosb6:
         jnz     nosb7
         cmp     [bgrlock], 0
         jz      .err
-        mov     eax, [current_slot_idx]
+        mov     eax, [current_slot]
+        mov     eax, [eax + APPDATA.tid]
         cmp     [bgrlockpid], eax
         jnz     .err
         mov     eax, ecx

kernel/trunk/gui/font.inc

@@ -48,6 +48,10 @@ dtext:
         add     esp, 28
         ret
 
+.nomem:
+        add     esp, 40
+        ret
+
 .redirect:
         mov     ebp, [edi]
         add     edi, 8
@@ -81,6 +85,8 @@ dtext:
         mov     [esp+32], eax
         imul    ebp, eax
         stdcall kernel_alloc, ebp
+        test    eax, eax
+        jz      .nomem
         mov     ecx, ebp
         shr     ecx, 2
         mov     [esp+36], eax

kernel/trunk/gui/window.inc

@@ -1997,6 +1997,8 @@ align 4
         ; esi - slot
         push    eax ecx edx esi
         xchg    eax, esi
+        shl     eax, BSF sizeof.APPDATA
+        add     eax, SLOT_BASE
         lea     ecx, [esp + 12]
         mov     edx, 1
         call    read_process_memory

kernel/trunk/kernel.asm

@@ -55,7 +55,6 @@ launcher_start db 1
 endg
 
 max_processes  =  255
-tss_step       =  128 + 8192 ; tss & i/o - 65535 ports, * 256=557056*4
 
 os_stack       =  os_data_l - gdts    ; GDTs
 os_code        =  os_code_l - gdts
@@ -389,7 +388,7 @@ high_code:
         mov     edi, tss._io_map_0
         xor     eax, eax
         not     eax
-        mov     ecx, 8192/4
+        mov     ecx, IO_MAP_SIZE/4
         rep stosd                    ; access to 4096*8=65536 ports
 
         mov     ax, tss0
@@ -410,7 +409,7 @@ high_code:
         stdcall alloc_kernel_space, 0x50000         ; FIXME check size
         mov     [default_io_map], eax
 
-        add     eax, 0x2000
+        add     eax, IO_MAP_SIZE
         mov     [ipc_tmp], eax
         mov     ebx, 0x1000
 
@@ -535,7 +534,7 @@ high_code:
         dec     ecx
         jns     @b
 
-        mov     edx, SLOT_BASE + sizeof.APPDATA*1
+        mov     edx, APPDATA_IDLE_THREAD
         mov     ebx, [os_stack_seg]
         add     ebx, RING0_STACK_SIZE
         add     ebx, [xsave_area_size]
@@ -547,7 +546,7 @@ high_code:
         mov     ecx, IDLE_PRIORITY
         call    scheduler_add_thread
 
-        mov     edx, SLOT_BASE + sizeof.APPDATA*2
+        mov     edx, APPDATA_OS_THREAD
         mov     ebx, [os_stack_seg]
         call    setup_os_slot
         mov     dword [edx], 'OS'
@@ -556,7 +555,7 @@ high_code:
 
         mov     [current_slot_idx], 2
         mov     [thread_count], 2
-        mov     [current_slot], SLOT_BASE + sizeof.APPDATA*2
+        mov     [current_slot], APPDATA_OS_THREAD
 
 ; Move other CPUs to deep sleep, if it is useful
 uglobal
@@ -848,15 +847,15 @@ include "detect/vortex86.inc"                     ; Vortex86 SoC detection code
         call    reserve_irqs_ports
 
         mov     [SLOT_BASE + APPDATA.window], window_data
-        mov     [SLOT_BASE + sizeof.APPDATA + APPDATA.window], window_data + sizeof.WDATA
-        mov     [SLOT_BASE + sizeof.APPDATA*2 + APPDATA.window], window_data + sizeof.WDATA*2
+        mov     [APPDATA_IDLE_THREAD + APPDATA.window], background_window
+        mov     [APPDATA_OS_THREAD + APPDATA.window], window_data + sizeof.WDATA*2
         mov     [window_data + WDATA.thread], SLOT_BASE
-        mov     [window_data + sizeof.WDATA + WDATA.thread], SLOT_BASE + sizeof.APPDATA
-        mov     [window_data + sizeof.WDATA*2 + WDATA.thread], SLOT_BASE + sizeof.APPDATA*2
+        mov     [background_window + WDATA.thread], APPDATA_IDLE_THREAD
+        mov     [window_data + sizeof.WDATA*2 + WDATA.thread], APPDATA_OS_THREAD
 
         call    init_display
         mov     eax, [def_cursor]
-        mov     [window_data + sizeof.WDATA + WDATA.cursor], eax
+        mov     [background_window + WDATA.cursor], eax
         mov     [window_data + sizeof.WDATA*2 + WDATA.cursor], eax
 
 ; PRINT CPU FREQUENCY
@@ -953,14 +952,14 @@ include "detect/vortex86.inc"                     ; Vortex86 SoC detection code
 ; Protect I/O permission map
 
         mov     esi, [default_io_map]
-        stdcall map_page, esi, [SLOT_BASE + sizeof.APPDATA + APPDATA.io_map], PG_READ
+        stdcall map_page, esi, [APPDATA_IDLE_THREAD + APPDATA.io_map], PG_READ
         add     esi, 0x1000
-        stdcall map_page, esi, [SLOT_BASE + sizeof.APPDATA + APPDATA.io_map + 4], PG_READ
+        stdcall map_page, esi, [APPDATA_IDLE_THREAD + APPDATA.io_map + 4], PG_READ
 
         stdcall map_page, tss._io_map_0, \
-                [SLOT_BASE + sizeof.APPDATA + APPDATA.io_map], PG_READ
+                [APPDATA_IDLE_THREAD + APPDATA.io_map], PG_READ
         stdcall map_page, tss._io_map_1, \
-                [SLOT_BASE + sizeof.APPDATA + APPDATA.io_map + 4], PG_READ
+                [APPDATA_IDLE_THREAD + APPDATA.io_map + 4], PG_READ
 
 ; SET KEYBOARD PARAMETERS
         mov     al, 0xf6       ; reset keyboard, scan enabled
@@ -1865,7 +1864,7 @@ sysfn_shutdown:          ; 18.9 = system shutdown
 ; out: Z/z -- is/not kernel thread
 is_kernel_thread:
         mov     eax, [eax + APPDATA.process]
-        cmp     eax, [SLOT_BASE + 2*sizeof.APPDATA + APPDATA.process]       ; OS
+        cmp     eax, [APPDATA_OS_THREAD + APPDATA.process]       ; OS
         ret
 ;------------------------------------------------------------------------------
 sysfn_terminate:        ; 18.2 = TERMINATE
@@ -2065,7 +2064,7 @@ sysfn_zmodif:
 
 ;------------------------------------------------------------------------------
 sysfn_getidletime:              ; 18.4 = GET IDLETIME
-        mov     eax, [SLOT_BASE + sizeof.APPDATA + APPDATA.cpu_usage]
+        mov     eax, [APPDATA_IDLE_THREAD + APPDATA.cpu_usage]
         mov     [esp + SYSCALL_STACK.eax], eax
         ret
 ;------------------------------------------------------------------------------
@@ -2825,7 +2824,7 @@ no_mark_system_shutdown:
 align 4
 noshutdown:
         mov     eax, [thread_count]           ; termination
-        mov     ebx, SLOT_BASE + sizeof.APPDATA + APPDATA.state
+        mov     ebx, APPDATA_IDLE_THREAD + APPDATA.state
         mov     esi, 1
 ;--------------------------------------
 align 4

kernel/trunk/network/ARP.inc

@@ -83,6 +83,31 @@ macro arp_init {
 
 }
 
+;-----------------------------------------------------------------;
+;                                                                 ;
+; arp_clear_device: Empty the ARP table and counters of a removed ;
+;                   device.                                       ;
+;                                                                 ;
+;  IN:  edi = device number * 4                                   ;
+;                                                                 ;
+;-----------------------------------------------------------------;
+macro arp_clear_device {
+
+        xor     eax, eax
+        mov     [ARP_entries + edi], eax
+        mov     [ARP_packets_tx + edi], eax
+        mov     [ARP_packets_rx + edi], eax
+        mov     [ARP_conflicts + edi], eax
+
+        push    edi
+        imul    edi, (ARP_TABLE_SIZE * sizeof.ARP_entry)/4
+        add     edi, ARP_table
+        mov     ecx, (ARP_TABLE_SIZE * sizeof.ARP_entry)/2
+        rep stosw
+        pop     edi
+
+}
+
 ;-----------------------------------------------------------------;
 ;                                                                 ;
 ; arp_decrease_entry_ttls                                         ;
@@ -313,6 +338,9 @@ arp_output_request:
         DEBUGF  DEBUG_NETWORK_VERBOSE, "ARP_output_request: ip=%u.%u.%u.%u device=0x%x\n",\
         [esp]:1, [esp + 1]:1, [esp + 2]:1, [esp + 3]:1, ebx
 
+        test    ebx, ebx                        ; device is gone
+        jz      .exit
+
         mov     ax, ETHER_PROTO_ARP
         mov     ecx, sizeof.ARP_header
         mov     edx, ETH_BROADCAST              ; broadcast mac

kernel/trunk/network/IPv4.inc

@@ -105,6 +105,29 @@ macro   ipv4_init {
 }
 
 
+;-----------------------------------------------------------------;
+;                                                                 ;
+; ipv4_clear_device: Reset the IPv4 settings of a removed device, ;
+;                    so no route points to it anymore.            ;
+;                                                                 ;
+;  IN:  edi = device number * 4                                   ;
+;                                                                 ;
+;-----------------------------------------------------------------;
+macro   ipv4_clear_device {
+
+        xor     eax, eax
+        mov     [IPv4_address + edi], eax
+        mov     [IPv4_subnet + edi], eax
+        mov     [IPv4_nameserver + edi], eax
+        mov     [IPv4_gateway + edi], eax
+        mov     [IPv4_broadcast + edi], eax
+        mov     [IPv4_packets_tx + edi], eax
+        mov     [IPv4_packets_rx + edi], eax
+        mov     [IPv4_packets_dumped + edi], eax
+
+}
+
+
 ;-----------------------------------------------------------------;
 ;                                                                 ;
 ; Decrease TimeToLive of all fragment slots                       ;
@@ -757,6 +780,8 @@ ipv4_output_raw:
         push    esi eax
 
         call    ipv4_route
+        test    eax, eax
+        jz      .arp_error
         call    arp_ip_to_mac
 
         test    eax, 0xffff0000         ; error bits
@@ -967,6 +992,8 @@ ipv4_route:
 ; Check for on-link
         xor     edi, edi
   .loop:
+        cmp     [net_device_list + edi], 0      ; skip slots without a device
+        je      .next
         mov     ebx, [IPv4_address + edi]
         and     ebx, [IPv4_subnet + edi]
         jz      .next
@@ -983,7 +1010,10 @@ ipv4_route:
         mov     edi, 4                  ; skip loopback device
   .loop_gw:
         cmp     [IPv4_gateway + edi], 0
+        je      .next_gw
+        cmp     [net_device_list + edi], 0
         jne     .found_gw
+  .next_gw:
         add     edi, 4
         cmp     edi, 4*NET_DEVICES_MAX
         jb      .loop_gw
@@ -1049,14 +1079,23 @@ ipv4_route:
         cmp     eax, 0xffffffff
         je      @f
 
+; An interface without an address (DHCP not done yet, or the lease
+; dropped when the link went down) can send nothing but broadcasts: with
+; address and mask both zero every destination compared as on-link, and
+; the stack went asking ARP for internet addresses on behalf of 0.0.0.0.
+        test    edx, edx
+        jz      .fail
+
 ; Check if we should route to gateway or not
-        mov     ebx, [IPv4_address + edi]
+        mov     ebx, edx
         and     ebx, [IPv4_subnet + edi]
         mov     ecx, eax
         and     ecx, [IPv4_subnet + edi]
         cmp     ecx, ebx
         je      @f
         mov     eax, [IPv4_gateway + edi]
+        test    eax, eax
+        jz      .fail                                ; off-link and no gateway
   @@:
         DEBUGF  DEBUG_NETWORK_VERBOSE, "IPv4_route: %u\n", edi
         ret

kernel/trunk/network/ethernet.inc

@@ -61,6 +61,41 @@ macro   eth_init {
 
 }
 
+;-----------------------------------------------------------------;
+;                                                                 ;
+; eth_clear_device: Drop the frames of a removed device that are  ;
+;                   still waiting in the input queue, their       ;
+;                   NET_BUFF.device will not be valid anymore.    ;
+;                                                                 ;
+;  IN:  ebx = device ptr                                          ;
+;                                                                 ;
+;-----------------------------------------------------------------;
+macro   eth_clear_device {
+
+local   .loop, .done
+
+        spin_lock_irqsave
+        mov     esi, [ETH_frame_head]
+  .loop:
+        cmp     esi, ETH_frame_head
+        je      .done
+        mov     eax, esi
+        mov     esi, [esi + NET_BUFF.NextPtr]
+        cmp     [eax + NET_BUFF.device], ebx
+        jne     .loop
+
+; unlink it, ETH_frame_head/tail double as the NextPtr/PrevPtr of the list head
+        mov     ecx, [eax + NET_BUFF.PrevPtr]
+        mov     [ecx + NET_BUFF.NextPtr], esi
+        mov     [esi + NET_BUFF.PrevPtr], ecx
+        dec     [ETH_frame_queued]
+        stdcall net_buff_free, eax
+        jmp     .loop
+  .done:
+        spin_unlock_irqrestore
+
+}
+
 align 4
 ; This function is called by ethernet drivers.
 ; Push the received ethernet packet onto the ethernet input queue.
@@ -245,6 +280,9 @@ eth_output:
 
         DEBUGF  DEBUG_NETWORK_VERBOSE, "ETH_output: size=%u device=%x\n", ecx, ebx
 
+        test    ebx, ebx
+        jz      .no_device
+
         cmp     ecx, [ebx + ETH_DEVICE.mtu]
         ja      .too_large
 
@@ -295,11 +333,16 @@ eth_output:
         ret
 
   .too_large:
-        inc     [eax + NET_DEVICE.packets_tx_err]
+        inc     [ebx + NET_DEVICE.packets_tx_err]
         DEBUGF  DEBUG_NETWORK_VERBOSE, "ETH_output: Packet too large!\n"
         xor     eax, eax
         ret
 
+  .no_device:
+        DEBUGF  DEBUG_NETWORK_ERROR, "ETH_output: no device!\n"
+        xor     eax, eax
+        ret
+
 
 
 ;-----------------------------------------------------------------;

kernel/trunk/network/socket.inc

@@ -2124,11 +2124,11 @@ socket_free:
         mov     ebx, eax
         cmp     [eax + STREAM_SOCKET.rcv.start_ptr], 0
         je      @f
-        stdcall free_kernel_space, [eax + STREAM_SOCKET.rcv.start_ptr]
+        stdcall kernel_free, [eax + STREAM_SOCKET.rcv.start_ptr]
   @@:
         cmp     [ebx + STREAM_SOCKET.snd.start_ptr], 0
         je      @f
-        stdcall free_kernel_space, [ebx + STREAM_SOCKET.snd.start_ptr]
+        stdcall kernel_free, [ebx + STREAM_SOCKET.snd.start_ptr]
   @@:
         mov     eax, ebx
   .no_stream:

kernel/trunk/network/stack.inc

@@ -118,7 +118,7 @@ SS_MORETOCOME           = 0x4000
 SS_BLOCKED              = 0x8000
 
 
-SOCKET_BUFFER_SIZE      = 4096*8        ; must be 4096*(power of 2) where 'power of 2' is at least 8
+SOCKET_BUFFER_SIZE      = 4096*32       ; must be 4096*(power of 2) where 'power of 2' is at least 8
 MAX_backlog             = 20            ; maximum backlog for stream sockets
 
 ; Error Codes
@@ -597,6 +597,16 @@ net_remove_device:
         mov     dword [edi-4], eax
         dec     [net_device_count]
 
+;-----------------------------------------------------------------
+; Forget everything the protocols still know about this interface.
+; Otherwise its IP address keeps matching in ipv4_route, and output
+; ends up calling through the now empty net_device_list slot.
+
+        sub     edi, net_device_list + 4        ; device number * 4
+        ipv4_clear_device
+        arp_clear_device
+        eth_clear_device
+
         call    net_send_event
 
         xor     eax, eax

kernel/trunk/network/tcp_input.inc

@@ -1422,10 +1422,14 @@ endl
         test    [edx + TCP_header.Flags], TH_ACK
         jz      .drop
 
-        mov     eax, ebx
+        push    ebx
+        lea     ecx, [ebx + SOCKET.mutex]
+        call    mutex_unlock
+        pop     eax
+
         mov     ebx, ECONNREFUSED
         call    tcp_drop
-        jmp     .drop
+        jmp     .drop_no_socket
        @@:
 
 ;-----------------------------------------------------------------------------------

kernel/trunk/network/tcp_output.inc

@@ -234,23 +234,25 @@ endl
         mov     ebx, TCP_max_win
         shl     ebx, cl
         pop     ecx
-        sub     ebx, [eax + TCP_SOCKET.RCV_ADV]
-        add     ebx, [eax + TCP_SOCKET.RCV_NXT]
+
+; Clamp to available window first, then subtract what the peer already knows (as in BSD)
 
         cmp     ebx, ecx
-        jl      @f
+        jbe     @f
         mov     ebx, ecx
        @@:
+        sub     ebx, [eax + TCP_SOCKET.RCV_ADV]
+        add     ebx, [eax + TCP_SOCKET.RCV_NXT]
 
         DEBUGF  DEBUG_NETWORK_VERBOSE, "TCP_output: we can increase window by %d bytes\n", ebx
 
         mov     edi, [eax + TCP_SOCKET.t_maxseg]
         shl     edi, 1
-        cmp     ebx, edi
-        jae     .send
+        cmp     ebx, edi                                ; signed: may be negative if window shrank
+        jge     .send
 
         cmp     ebx, SOCKET_BUFFER_SIZE/2
-        jae     .send
+        jge     .send
 
   .no_window:
 

kernel/trunk/network/tcp_usreq.inc

@@ -180,6 +180,8 @@ tcp_connect:
 
   .nomem:
         pop     edx eax
+        lea     ecx, [eax + SOCKET.mutex]
+        call    mutex_unlock
         xor     eax, eax
         dec     eax
         mov     ebx, ENOMEM
@@ -194,6 +196,8 @@ tcp_connect:
   .enoroute:
         pop     eax
         popa
+        lea     ecx, [eax + SOCKET.mutex]
+        call    mutex_unlock
         xor     eax, eax
         dec     eax
         mov     ebx, EADDRNOTAVAIL

kernel/trunk/sound/playnote.inc

@@ -148,9 +148,10 @@ ReadNoteByte:
         push    esi
 
         mov     eax, [pidProcessNote]
-        call    pid_to_slot
+        call    pid_to_thread
         test    eax, eax
         jz      .failed
+
         lea     ecx, [esp+12]
         mov     edx, 1
         mov     esi, [memAdrNote]

programs/demos/free3d04/build.bat

@@ -0,0 +1,2 @@
+@fasm free3d04.asm free3d04
+if not exist free3d04 ( @pause )

programs/demos/free3d04/build_en.bat

@@ -1,5 +0,0 @@
-@erase lang.inc
-@echo lang fix en_US >lang.inc
-@fasm free3d04.asm free3d04
-@erase lang.inc
-@pause

programs/demos/free3d04/build_ru.bat

@@ -1,5 +0,0 @@
-@erase lang.inc
-@echo lang fix ru_RU >lang.inc
-@fasm free3d04.asm free3d04
-@erase lang.inc
-@pause

programs/demos/free3d04/free3d04.asm

@@ -9,9 +9,10 @@
 ;
 ;   Compile with FASM for Menuet (requires .INC files - see DATA Section)
 ;
-;   Willow - greatly srinked code size by using GIF texture and FPU to calculate sine table
+;   Willow - greatly srinked code size by using packed texture and FPU to calculate sine table
 ;
-;   !!!! Don't use GIF_LITE.INC in your apps - it's modified for FREE3D !!!!
+;   Textures are stored as a single 64x512 PNG embedded into the binary and
+;   decoded at startup with libimg.
 ;
 ;   Heavyiron - new 0-function of drawing window from kolibri (do not work correctly with menuet)
 
@@ -38,15 +39,17 @@ use32
                dd     APP_MEM;0x100000        ; memory for app
                dd     APP_MEM;0x100000        ; esp
                dd     0x0 , 0x0               ; I_Param , I_Icon
-include 'lang.inc'
 include '..\..\macros.inc'
-COLOR_ORDER equ OTHER
-include 'gif_lite.inc'
+include '..\..\proc32.inc'
+include '..\..\dll.inc'
+include '..\..\develop\libraries\libs-dev\libimg\libimg.inc'
 
 START:                          ; start of execution
-		mov  esi,textures
-		mov  edi,ceil-8
-		call ReadGIF
+		mcall 68,11                 ; init heap, libimg needs it
+		stdcall dll.Load,@IMPORT
+		test eax,eax
+		jnz  finish
+		call load_textures
 		mov  esi,sinus
 		mov  ecx,360*10
 		fninit
@@ -291,6 +294,34 @@ m_right:                                  ; turn right
     mcall
 
 
+;   *********************************************
+;   *******        LOAD TEXTURES         ********
+;   *********************************************
+; Decodes the embedded 64x512 PNG (8 stacked 64x64 tiles) into the raw
+; 0x00RRGGBB texture bank at [ceil].
+
+load_textures:
+	invoke img.decode,textures,textures.size,0
+	test eax,eax
+	jz   finish
+	mov  ebx,eax
+	invoke img.convert,ebx,0,Image.bpp32,0,0
+	push eax
+	invoke img.destroy,ebx
+	pop  ebx
+	test ebx,ebx
+	jz   finish
+	mov  esi,[ebx+Image.Data]
+	mov  edi,ceil
+	mov  ecx,TEX_SIZE*8/4
+    .copy:
+	lodsd
+	and  eax,0x00FFFFFF     ; drop alpha, the renderer stores 32 bit dwords
+	stosd                   ; into a 24 bit image buffer
+	loop .copy
+	invoke img.destroy,ebx
+	ret
+
 ;   *********************************************
 ;   *******  WINDOW DEFINITIONS AND DRAW ********
 ;   *********************************************
@@ -980,13 +1011,21 @@ dd 0x0001FFFF ; initial player position * 0xFFFF
  vpy:
 dd 0x0001FFFF
 
-title    db   'FISHEYE RAYCASTING ENGINE ETC. FREE3D',0
+title    db   'Fisheye Raycasting Engine Etc. FREE3D',0
 
 sindegree dd 0.0
 sininc    dd 0.0017453292519943295769236907684886
 sindiv    dd 6553.5
 textures:
-	file 'texture.gif'
+	file 'texture.png'
+  .size = $ - textures
+
+align 16
+@IMPORT:
+
+library                       	libimg , 'libimg.obj'
+
+import	libimg                      , 	libimg.init  , 'lib_init'   , 	img.decode   , 'img_decode' , 	img.convert  , 'img_convert', 	img.destroy  , 'img_destroy'
 
 align 4
 

programs/demos/free3d04/gif_lite.inc

@@ -1,487 +0,0 @@
-; GIF LITE v3.0 by Willow
-; Written in pure assembler by Ivushkin Andrey aka Willow
-; Modified by Diamond
-;
-; This include file will contain functions to handle GIF image format
-;
-; Created: August 15, 2004
-; Last changed: June 24, 2007
-
-; Requires kglobals.inc (iglobal/uglobal macro)
-; (program must 'include "kglobals.inc"' and say 'IncludeUGlobal'
-; somewhere in uninitialized data area).
-
-; Configuration: [changed from program which includes this file]
-; 1. The constant COLOR_ORDER: must be one of
-;    PALETTE - for 8-bit image with palette (sysfunction 65)
-;    MENUETOS - for MenuetOS and KolibriOS color order (sysfunction 7)
-;    OTHER - for standard color order
-; 2. Define constant GIF_SUPPORT_INTERLACED if you want to support interlaced
-;    GIFs.
-; 3. Single image mode vs multiple image mode:
-;    if the program defines the variable 'gif_img_count' of type dword
-;    somewhere, ReadGIF will enter multiple image mode: gif_img_count
-;    will be initialized with image count, output format is GIF_list,
-;    the function GetGIFinfo retrieves Nth image info. Otherwise, ReadGIF
-;    uses single image mode: exit after end of first image, output is
-;    <dd width,height, times width*height[*3] db image>
-
-if ~ (COLOR_ORDER in <PALETTE,MENUETOS,OTHER>)
-; This message may not appear under MenuetOS, so watch...
-  display 'Please define COLOR_ORDER: PALETTE, MENUETOS or OTHER',13,10
-end if
-
-if defined gif_img_count
-; virtual structure, used internally
-
-struct GIF_list
-    NextImg rd 1
-    Left   rw 1
-    Top    rw 1
-    Width  rw 1
-    Height rw 1
-    Delay  rd 1
-    Displacement rd 1 ; 0 = not specified
-                       ; 1 = do not dispose
-                       ; 2 = restore to background color
-                       ; 3 = restore to previous
-if COLOR_ORDER eq PALETTE
-    Image rd 1
-end if
-ends
-
-struct GIF_info
-    Left   rw 1
-    Top    rw 1
-    Width  rw 1
-    Height rw 1
-    Delay  rd 1
-    Displacement rd 1
-if COLOR_ORDER eq PALETTE
-    Palette rd 1
-end if
-ends
-
-; ****************************************
-;   FUNCTION GetGIFinfo - retrieve Nth image info
-; ****************************************
-; in:
-;   esi - pointer to image list header
-;   ecx - image_index (0...img_count-1)
-;   edi - pointer to GIF_info structure to be filled
-
-; out:
-;   eax - pointer to RAW data, or 0, if error
-
-GetGIFinfo:
-    push  esi ecx edi
-    xor   eax,eax
-    jecxz .eloop
-  .lp:
-    mov   esi,[esi]
-    test  esi,esi
-    jz    .error
-    loop  .lp
-  .eloop:
-    lodsd
-    movsd
-    movsd
-    movsd
-    movsd
-if COLOR_ORDER eq PALETTE
-    lodsd
-    mov   [edi],esi
-else
-    mov   eax,esi
-end if
-  .error:
-    pop   edi ecx esi
-    ret
-
-end if
-
-_null fix 0x1000
-
-; ****************************************
-;   FUNCTION ReadGIF - unpacks GIF image
-; ****************************************
-; in:
-;   esi - pointer to GIF file in memory
-;   edi - pointer to output image list
-
-; out:
-;   eax - 0, all OK;
-;   eax - 1, invalid signature;
-;   eax >=8, unsupported image attributes
-;
-
-ReadGIF:
-    push esi edi
-    mov  [.cur_info],edi
-    xor  eax,eax
-    mov  [.globalColor],eax
-if defined gif_img_count
-    mov  [gif_img_count],eax
-    mov  [.anim_delay],eax
-    mov  [.anim_disp],eax
-end if
-    inc  eax
-    cmp  dword[esi],'GIF8'
-    jne  .ex            ; signature
-    mov  ecx,[esi+0xa]
-    add  esi,0xd
-    mov  edi,esi
-    test cl,cl
-    jns  .nextblock
-    mov  [.globalColor],esi
-    call .Gif_skipmap
-  .nextblock:
-    cmp  byte[edi],0x21
-    jne  .noextblock
-    inc  edi
-if defined gif_img_count
-    cmp  byte[edi],0xf9 ; Graphic Control Ext
-    jne  .no_gc
-    movzx eax,word [edi+3]
-    mov  [.anim_delay],eax
-    mov  al,[edi+2]
-    shr  al,2
-    and  eax,7
-    mov  [.anim_disp],eax
-    add  edi,7
-    jmp  .nextblock
-  .no_gc:
-end if
-    inc  edi
-  .block_skip:
-    movzx eax,byte[edi]
-    lea  edi,[edi+eax+1]
-    test eax,eax
-    jnz  .block_skip
-    jmp  .nextblock
-  .noextblock:
-    mov  al,8
-    cmp  byte[edi],0x2c    ; image beginning
-    jne  .ex
-if defined gif_img_count
-    inc  [gif_img_count]
-end if
-    inc  edi
-    mov  esi,[.cur_info]
-if defined gif_img_count
-    add  esi,4
-end if
-    xchg esi,edi
-if defined GIF_SUPPORT_INTERLACED
-    movzx ecx,word[esi+4]
-    mov  [.width],ecx
-    movzx eax,word[esi+6]
-    imul eax,ecx
-if ~(COLOR_ORDER eq PALETTE)
-    lea  eax,[eax*3]
-end if
-    mov  [.img_end],eax
-    inc  eax
-    mov  [.row_end],eax
-    and  [.pass],0
-    test byte[esi+8],40h
-    jz   @f
-if ~(COLOR_ORDER eq PALETTE)
-    lea  ecx,[ecx*3]
-end if
-    mov  [.row_end],ecx
-@@:
-end if
-if defined gif_img_count
-    movsd
-    movsd
-    mov  eax,[.anim_delay]
-    stosd
-    mov  eax,[.anim_disp]
-    stosd
-else
-    movzx eax,word[esi+4]
-    stosd
-    movzx eax,word[esi+6]
-    stosd
-    add   esi,8
-end if
-    push edi
-    mov  ecx,[esi]
-    inc  esi
-    test cl,cl
-    js   .uselocal
-    push [.globalColor]
-    mov  edi,esi
-    jmp  .setPal
-  .uselocal:
-    call .Gif_skipmap
-    push esi
-  .setPal:
-    movzx ecx,byte[edi]
-    inc  ecx
-    mov  [.codesize],ecx
-    dec  ecx
-if ~(COLOR_ORDER eq PALETTE)
-    pop  [.Palette]
-end if
-    lea  esi,[edi+1]
-    mov  edi,.gif_workarea
-    xor  eax,eax
-    lodsb               ; eax - block_count
-    add  eax,esi
-    mov  [.block_ofs],eax
-    mov  [.bit_count],8
-    mov  eax,1
-    shl  eax,cl
-    mov  [.CC],eax
-    mov  ecx,eax
-    inc  eax
-    mov  [.EOI],eax
-    mov  eax, _null shl 16
-  .filltable:
-    stosd
-    inc  eax
-    loop .filltable
-if COLOR_ORDER eq PALETTE
-    pop  eax
-    pop  edi
-    push edi
-    scasd
-    push esi
-    mov  esi,eax
-    mov  ecx,[.CC]
-@@:
-    lodsd
-    dec  esi
-    bswap eax
-    shr  eax,8
-    stosd
-    loop @b
-    pop  esi
-    pop  eax
-    mov  [eax],edi
-else
-    pop  edi
-end if
-if defined GIF_SUPPORT_INTERLACED
-    mov  [.img_start],edi
-    add  [.img_end],edi
-    add  [.row_end],edi
-end if
-  .reinit:
-    mov  edx,[.EOI]
-    inc  edx
-    push [.codesize]
-    pop  [.compsize]
-    call .Gif_get_sym
-    cmp  eax,[.CC]
-    je   .reinit
-    call .Gif_output
-  .cycle:
-    movzx ebx,ax
-    call .Gif_get_sym
-    cmp  eax,edx
-    jae  .notintable
-    cmp  eax,[.CC]
-    je   .reinit
-    cmp  eax,[.EOI]
-    je   .end
-    call .Gif_output
-  .add:
-    mov  dword [.gif_workarea+edx*4],ebx
-    cmp  edx,0xFFF
-    jae  .cycle
-    inc  edx
-    bsr  ebx,edx
-    cmp  ebx,[.compsize]
-    jne  .noinc
-    inc  [.compsize]
-  .noinc:
-    jmp  .cycle
-  .notintable:
-    push eax
-    mov  eax,ebx
-    call .Gif_output
-    push ebx
-    movzx eax,bx
-    call .Gif_output
-    pop  ebx eax
-    jmp  .add
-  .end:
-if defined GIF_SUPPORT_INTERLACED
-    mov  edi,[.img_end]
-end if
-if defined gif_img_count
-    mov  eax,[.cur_info]
-    mov  [eax],edi
-    mov  [.cur_info],edi
-    add  esi,2
-    xchg esi,edi
-  .nxt:
-    cmp  byte[edi],0
-    jnz  .continue
-    inc  edi
-    jmp  .nxt
-  .continue:
-    cmp  byte[edi],0x3b
-    jne  .nextblock
-    xchg esi,edi
-    and  dword [eax],0
-end if
-    xor  eax,eax
-  .ex:
-    pop  edi esi
-    ret
-
-.Gif_skipmap:
-; in: ecx - image descriptor, esi - pointer to colormap
-; out: edi - pointer to area after colormap
-
-    and  ecx,111b
-    inc  ecx            ; color map size
-    mov  ebx,1
-    shl  ebx,cl
-    lea  ebx,[ebx*2+ebx]
-    lea  edi,[esi+ebx]
-    ret
-
-.Gif_get_sym:
-    mov  ecx,[.compsize]
-    push ecx
-    xor  eax,eax
-  .shift:
-    ror  byte[esi],1
-    rcr  eax,1
-    dec  [.bit_count]
-    jnz  .loop1
-    inc  esi
-    cmp  esi,[.block_ofs]
-    jb   .noblock
-    push eax
-    xor  eax,eax
-    lodsb
-    test eax,eax
-    jnz  .nextbl
-    mov  eax,[.EOI]
-    sub  esi,2
-    add  esp,8
-    jmp  .exx
-  .nextbl:
-    add  eax,esi
-    mov  [.block_ofs],eax
-    pop  eax
-  .noblock:
-    mov  [.bit_count],8
-  .loop1:
-    loop .shift
-    pop  ecx
-    rol  eax,cl
-  .exx:
-    xor  ecx,ecx
-    ret
-
-.Gif_output:
-    push esi eax edx
-    mov  edx,.gif_workarea
-  .next:
-    push word[edx+eax*4]
-    mov  ax,word[edx+eax*4+2]
-    inc  ecx
-    cmp  ax,_null
-    jnz  .next
-    shl  ebx,16
-    mov  bx,[esp]
-  .loop2:
-    pop  ax
-
-    if COLOR_ORDER eq PALETTE
-        stosb
-    else
-        lea  esi,[eax+eax*2]
-        add  esi,[.Palette]
-
-    if COLOR_ORDER eq MENUETOS
-        mov  esi,[esi]
-        bswap esi
-        shr  esi,8
-        mov  [edi],esi
-        add  edi,3
-    else
-        movsb
-        movsb
-        movsb
-	mov  byte [edi],0
-	inc  edi
-    end if
-    end if
-
-if defined GIF_SUPPORT_INTERLACED
-    cmp  edi,[.row_end]
-    jb   .norowend
-    mov  eax,[.width]
-if ~(COLOR_ORDER eq PALETTE)
-    lea  eax,[eax*3]
-end if
-    push eax
-    sub  edi,eax
-    add  eax,eax
-    cmp  [.pass],3
-    jz   @f
-    add  eax,eax
-    cmp  [.pass],2
-    jz   @f
-    add  eax,eax
-@@:
-    add  edi,eax
-    pop  eax
-    cmp  edi,[.img_end]
-    jb   .nextrow
-    mov  edi,[.img_start]
-    inc  [.pass]
-    add  edi,eax
-    cmp  [.pass],3
-    jz   @f
-    add  edi,eax
-    cmp  [.pass],2
-    jz   @f
-    add  edi,eax
-    add  edi,eax
-@@:
-.nextrow:
-    add  eax,edi
-    mov  [.row_end],eax
-    xor  eax,eax
-.norowend:
-end if
-
-    loop .loop2
-    pop  edx eax esi
-    ret
-
-uglobal
-align 4
-    ReadGIF.globalColor rd 1
-    ReadGIF.cur_info rd 1        ; image table pointer
-    ReadGIF.codesize rd 1
-    ReadGIF.compsize rd 1
-    ReadGIF.bit_count rd 1
-    ReadGIF.CC rd 1
-    ReadGIF.EOI rd 1
-if ~(COLOR_ORDER eq PALETTE)
-    ReadGIF.Palette rd 1
-end if
-    ReadGIF.block_ofs rd 1
-if defined GIF_SUPPORT_INTERLACED
-    ReadGIF.row_end rd 1
-    ReadGIF.img_end rd 1
-    ReadGIF.img_start rd 1
-    ReadGIF.pass rd 1
-    ReadGIF.width rd 1
-end if
-if defined gif_img_count
-    ReadGIF.anim_delay rd 1
-    ReadGIF.anim_disp rd 1
-end if
-    ReadGIF.gif_workarea rb 16*1024
-endg

programs/demos/free3d04/readme.txt

@@ -10,11 +10,11 @@ By Dieter Marfurt
 
 --------------------------------------------
 
-Format of texture include files:
+Format of the texture file:
 
-dd 0x00RRGGBB,0x00RRGGBB....
-
-for 64*64 pixels. 
+texture.png - a 64*512 image holding the 8 textures of 64*64 pixels
+stacked vertically. It is embedded into the binary and decoded at
+startup with libimg.
 
 Have fun!
 

programs/develop/cedit/SRC/Menu.ob07

@@ -238,7 +238,6 @@ BEGIN
     m.active := FALSE;
    	resetTimer;
     m.tid := 0;
-    G.destroy(m.canvas);
     K.Exit
 END exit;
 
@@ -250,7 +249,6 @@ BEGIN
     	resetTimer
     END;
     m.tid := 0;
-    G.destroy(m.canvas);
     K.Exit
 END escape;
 
@@ -356,7 +354,6 @@ BEGIN
     	END;
     	temp := m.tid;
     	m.tid := 0;
-    	G.destroy(m.canvas);
         K.ExitID(temp);
         m.active := FALSE
     END
@@ -512,7 +509,6 @@ BEGIN
 	            	IF m.parent # NIL THEN
 	            		temp := m.parent.tid;
 	            		m.parent.tid := 0;
-	            		G.destroy(m.parent.canvas);
 	            		K.ExitID(temp)
 	            	END;
 	                exit(m)

programs/develop/ktcc/bin/lib/libc.def

@@ -181,6 +181,7 @@ con_cls
 con_get_cursor_pos
 con_set_cursor_pos
 ;____TIME____
+clock
 mktime
 time
 localtime

programs/develop/ktcc/lib/libshell/shell_cls.c

@@ -5,7 +5,7 @@ void shell_cls()
     __shell_init();
     if (__shell_is_init == __SHELL_INIT_OK)
     {
-        *__shell_shm = SHELL_CLS;
+        __shell_shm->cmd = SHELL_CLS;
         __SHELL_WAIT();
     }
 }

programs/develop/ktcc/lib/libshell/shell_exit.c

@@ -5,7 +5,7 @@ void shell_exit()
 {
     if (__shell_is_init == __SHELL_INIT_OK)
     {
-        *__shell_shm = SHELL_EXIT;
+        __shell_shm->cmd = SHELL_EXIT;
         __SHELL_WAIT();
         _ksys_shm_close(__shell_shm_name);
     }

programs/develop/ktcc/lib/libshell/shell_get_pid.c

@@ -8,9 +8,9 @@ unsigned shell_get_pid()
     __shell_init();
     if (__shell_is_init == __SHELL_INIT_OK)
     {
-        *__shell_shm = SHELL_PID;
+        __shell_shm->cmd = SHELL_PID;
         __SHELL_WAIT();
-        memcpy(&pid, __shell_shm + 1, sizeof(unsigned));
+        memcpy(&pid, __shell_shm->data, sizeof(unsigned));
     }
 
     return pid;

programs/develop/ktcc/lib/libshell/shell_getc.c

@@ -1,15 +1,15 @@
 #include <shell_api.h>
 
-char shell_getc()
+int shell_getc()
 {
     __shell_init();
 
     if (__shell_is_init == __SHELL_INIT_OK)
     {
-        *__shell_shm = SHELL_GETC;
+        __shell_shm->cmd = SHELL_GETC;
         __SHELL_WAIT();
 
-        return *(__shell_shm + 1);
+        return *((int*)__shell_shm->data);
     }
     else
     {

programs/develop/ktcc/lib/libshell/shell_gets.c

@@ -1,15 +1,28 @@
 #include <shell_api.h>
+#include <stdio.h>
 #include <string.h>
 
-void shell_gets(char *str, int n)
+char *shell_gets(char *str, int n)
 {
     __shell_init();
 
+    char *ptr = NULL;
     if (__shell_is_init == __SHELL_INIT_OK)
     {
-        *__shell_shm = SHELL_GETS;
+        __shell_shm->cmd = SHELL_GETS;
         __SHELL_WAIT();
 
-        strncpy(str, __shell_shm + 1, n);
+        strncpy(str, __shell_shm->data, n);
+
+        if (__shell_shm->return_value.str == EOF || !__shell_shm->return_value.str)
+        {
+            ptr = __shell_shm->return_value.str;
+        }
+        else
+        {
+            ptr = str;
+        }
     }
+
+    return ptr;
 }

programs/develop/ktcc/lib/libshell/shell_init.c

@@ -6,7 +6,7 @@
 
 char app_name[13];
 char __shell_shm_name[32];
-char *__shell_shm = NULL;
+struct shell_shm_buffer *__shell_shm = NULL;
 enum __SHELL_INIT_STATE __shell_is_init = __SHELL_NOT_LOADED;
 
 int __shell_shm_init()
@@ -17,12 +17,12 @@ int __shell_shm_init()
 
     _ksys_thread_info(&proc_info, -1);
     PID = proc_info.pid;
-    strncpy(app_name, (&proc_info)->name, 12);
+    strncpy(app_name, proc_info.name, sizeof(app_name) - 1);
 
     itoa(PID, __shell_shm_name);
     strcat(__shell_shm_name, "-SHELL");
 
-    return _ksys_shm_open(__shell_shm_name, KSYS_SHM_OPEN_ALWAYS | KSYS_SHM_WRITE, SHELL_SHM_MAX, &__shell_shm);
+    return _ksys_shm_open(__shell_shm_name, KSYS_SHM_OPEN_ALWAYS | KSYS_SHM_WRITE, sizeof(*__shell_shm), (char **)&__shell_shm);
 }
 
 void __shell_init()

programs/develop/ktcc/lib/libshell/shell_ping.c

@@ -7,12 +7,12 @@
 
 int shell_ping()
 {
-    *__shell_shm = SHELL_PING;
+    __shell_shm->cmd = SHELL_PING;
 
     _ksys_thread_yield(); // hope shell is fast enough
 
     size_t i = 0;
-    while (*__shell_shm != SHELL_OK)
+    while (__shell_shm->cmd != SHELL_OK)
     {
         if (i > (SHELL_PING_TIMEOUT / SHELL_PING_MIN_DELAY))
         {

programs/develop/ktcc/lib/libshell/shell_printf.c

@@ -1,12 +1,13 @@
 #include <shell_api.h>
 #include <stdio.h>
 
-void shell_printf(const char *format, ...)
+int shell_printf(const char *format, ...)
 {
     va_list ap;
     va_start(ap, format);
-    vsnprintf(__shell_shm + 1, SHELL_SHM_MAX, format, ap);
-    *__shell_shm = SHELL_PUTS;
+    int val = vsnprintf(__shell_shm->data, sizeof(__shell_shm->data), format, ap);
+    __shell_shm->cmd = SHELL_PUTS;
     va_end(ap);
     __SHELL_WAIT();
+    return val;
 }

programs/develop/ktcc/lib/libshell/shell_putc.c

@@ -6,8 +6,8 @@ void shell_putc(char c)
 
     if (__shell_is_init == __SHELL_INIT_OK)
     {
-        *(__shell_shm + 1) = c;
-        *__shell_shm = SHELL_PUTC;
+        __shell_shm->data[0] = c;
+        __shell_shm->cmd = SHELL_PUTC;
         __SHELL_WAIT();
     }
 }

programs/develop/ktcc/lib/libshell/shell_puts.c

@@ -1,7 +1,7 @@
 #include <shell_api.h>
 #include <string.h>
 
-void shell_puts(const char *str)
+size_t shell_puts(const char *str)
 {
-    shell_write_string(str, strlen(str));
+    return shell_write_string(str, strlen(str));
 }

programs/develop/ktcc/lib/libshell/shell_write_string.c

@@ -1,22 +1,27 @@
 #include <shell_api.h>
 #include <string.h>
 
-void shell_write_string(const char *s, size_t len)
+size_t shell_write_string(const char *s, size_t len)
 {
     __shell_init();
     if (__shell_is_init == __SHELL_INIT_OK)
     {
-        if (len > SHELL_SHM_MAX - 1)
+        size_t written_bytes = 0;
+        if (len > sizeof(__shell_shm->data) - 1)
         {
-            shell_write_string(s, SHELL_SHM_MAX - 1);                               // Outputs as much as it can.
-            shell_write_string(s + (SHELL_SHM_MAX - 1), len - (SHELL_SHM_MAX - 1)); // Outputs the rest.
+            written_bytes += shell_write_string(s, sizeof(__shell_shm->data) - 1);                                           // Outputs as much as it can.
+            written_bytes += shell_write_string(s + (sizeof(__shell_shm->data) - 1), len - (sizeof(__shell_shm->data) - 1)); // Outputs the rest.
         }
         else
         {
-            memset(__shell_shm, 0, SHELL_SHM_MAX);  // without int shell show \t, \n, lose chars and other trash
-            memcpy(__shell_shm + 1, s, len);
-            *__shell_shm = SHELL_PUTS;
+            memset(__shell_shm->data, 0, sizeof(__shell_shm->data)); // without int shell show \t, \n, lose chars and other trash
+            memcpy(__shell_shm->data, s, len);
+            __shell_shm->cmd = SHELL_PUTS;
             __SHELL_WAIT();
+            written_bytes += __shell_shm->return_value.written_bytes;
         }
+
+        return written_bytes;
     }
+    return 0;
 }

programs/develop/ktcc/libc.obj/include/shell_api.h

@@ -9,15 +9,17 @@
 #define __EXTERN extern
 #endif
 
-#define SHELL_OK   0
-#define SHELL_EXIT 1
-#define SHELL_PUTC 2
-#define SHELL_PUTS 3
-#define SHELL_GETC 4
-#define SHELL_GETS 5
-#define SHELL_CLS  6
-#define SHELL_PID  7
-#define SHELL_PING 8
+enum SHELL_API_COMMANDS {
+    SHELL_OK = 0,
+    SHELL_EXIT = 1,
+    SHELL_PUTC = 2,
+    SHELL_PUTS = 3,
+    SHELL_GETC = 4,
+    SHELL_GETS = 5,
+    SHELL_CLS = 6,
+    SHELL_PID = 7,
+    SHELL_PING = 8
+};
 
 #define SHELL_SHM_MAX 1024 * 16
 
@@ -28,28 +30,42 @@ enum __SHELL_INIT_STATE {
     __SHELL_INIT_FAILED = 3 // fail init shell
 };
 
+#pragma pack(push, 1)
+struct return_value {
+    size_t written_bytes;
+    char* str;
+};
+
+struct shell_shm_buffer {
+    uint8_t cmd;
+    char data[SHELL_SHM_MAX - sizeof(((struct shell_shm_buffer*)NULL)->cmd) - sizeof(struct return_value)];
+    struct return_value return_value;
+};
+#pragma pack(pop)
+
 __EXTERN char __shell_shm_name[32];
-__EXTERN char* __shell_shm;
+__EXTERN struct shell_shm_buffer* __shell_shm;
 __EXTERN enum __SHELL_INIT_STATE __shell_is_init;
 __EXTERN void __shell_init();
 
 #define __SHELL_WAIT()        \
     do {                      \
         _ksys_thread_yield(); \
-    } while (*__shell_shm);
+    } while (__shell_shm->cmd);
 
 __EXTERN int shell_ping();
 __EXTERN unsigned shell_get_pid();
 __EXTERN void shell_exit();
 
-__EXTERN char shell_getc();
-__EXTERN void shell_gets(char* str, int n);
+__EXTERN int shell_getc();
+__EXTERN char* shell_gets(char* str, int n);
 
 __EXTERN void shell_putc(char c);
-__EXTERN void shell_puts(const char* str);
-__EXTERN void shell_printf(const char* format, ...);
+__EXTERN size_t shell_puts(const char* str);
+__EXTERN int shell_printf(const char* format, ...);
 
-__EXTERN void shell_write_string(const char* s, size_t len);
+__EXTERN size_t shell_write_string(const char* s, size_t len);
 
 __EXTERN void shell_cls();
+
 #endif

programs/develop/ktcc/libc.obj/include/stdio.h

@@ -34,6 +34,8 @@
 
 #include <stdarg.h>
 #include <stddef.h>
+#include <stdint.h>
+#include <stdbool.h>
 
 DLLAPI int puts(const char* str);
 DLLAPI int printf(const char* format, ...);
@@ -50,15 +52,23 @@ typedef size_t fpos_t;
 #define _FILEMODE_R 1 << 0    // Read
 #define _FILEMODE_W 1 << 1    // Write
 #define _FILEMODE_A 1 << 2    // Append
-#define _FILEMODE_PLUS 1 << 3 // Plus
+#define _FILEFLAG_ERR 1 << 3 // Error
+#define _FILEFLAG_EOF 1 << 4 // EOF
 
 typedef struct FILE_s {
     char* name;
     fpos_t position;
-    int error;
-    int eof;
-    int mode;              // flags _FILEMODE_*
     int __ungetc_emu_buff; // Uses __ungetc_emu (temporary solution!)
+    union {
+        struct {
+            bool read : 1;
+            bool write : 1;
+            bool append : 1;
+            bool error : 1;
+            bool eof : 1;
+        };
+        uint32_t val; // exist just because tcc won't optimize bitfields
+    } flags;
 } FILE;
 
 #define _IOFBF 0
@@ -82,7 +92,7 @@ typedef struct FILE_s {
 #define TMP_MAX FOPEN_MAX
 
 #define stderr (FILE*)3
-#define stdin (FILE*)1
+#define stdin  (FILE*)1
 #define stdout (FILE*)2
 
 DLLAPI int fgetc(FILE*);

programs/develop/ktcc/libc.obj/include/time.h

@@ -1,11 +1,10 @@
 #ifndef _TIME_H_
 #define _TIME_H_
 
-#include <sys/ksys.h>
+#include <stddef.h>
 
 typedef unsigned long int clock_t;
 typedef unsigned long int time_t;
-#define clock()        _ksys_get_clock()
 #define CLOCKS_PER_SEC 100
 
 #pragma pack(push, 1)
@@ -22,6 +21,7 @@ struct tm {
 };
 #pragma pack(pop)
 
+DLLAPI clock_t clock(void);
 DLLAPI time_t mktime(struct tm* timeptr);
 DLLAPI time_t time(time_t* timer);
 DLLAPI struct tm* localtime(const time_t* timer);

programs/develop/ktcc/libc.obj/source/libc.c

@@ -61,6 +61,8 @@
 /* strntold() is used by vsscanf() below; include its definition first */
 #include "stdlib/strntold.c"
 #include "stdio/vsscanf.c"
+#include "stdio/fclose_r.c"
+#include "stdio/fopen_r.c"
 
 #include "string/memccpy.c"
 #include "string/memchr.c"
@@ -119,6 +121,7 @@
 #include "math/sinh.c"
 #include "math/tanh.c"
 
+#include "time/clock.c"
 #include "time/asctime.c"
 #include "time/difftime.c"
 #include "time/localtime.c"
@@ -338,6 +341,7 @@ ksys_dll_t EXPORTS[] = {
     { "con_cls", &con_cls },
     { "con_get_cursor_pos", &con_get_cursor_pos },
     { "con_set_cursor_pos", &con_set_cursor_pos },
+    { "clock", &clock },
     { "mktime", &mktime },
     { "time", &time },
     { "localtime", &localtime },

programs/develop/ktcc/libc.obj/source/stdio/_stdio.h

@@ -0,0 +1,9 @@
+#ifndef __LIBC_STDIO_H
+#define __LIBC_STDIO_H
+
+#include <stdio.h>
+
+int fclose_r(FILE* stream);
+FILE* fopen_r(const char* restrict name, const char* restrict mode, FILE* restrict stream);
+
+#endif // __LIBC_STDIO_H

programs/develop/ktcc/libc.obj/source/stdio/clearerr.c

@@ -1,6 +1,6 @@
 #include <stdio.h>
 
-void clearerr(FILE *stream) {
-	stream->error = 0;
-	stream->eof = 0;
+void clearerr(FILE* stream)
+{
+	stream->flags.val &= ~(_FILEFLAG_ERR | _FILEFLAG_EOF);
 }

programs/develop/ktcc/libc.obj/source/stdio/fclose.c

@@ -1,8 +1,15 @@
-#include "stddef.h"
+#include "_stdio.h"
 #include <stdio.h>
 #include <stdlib.h>
 
-int fclose(FILE *stream) {
+int fclose(FILE* stream)
+{
+	int err = fclose_r(stream);
+
+	if (err) {
+		return err;
+	}
+
 	free(stream);
 	stream = NULL;
 	return 0;

programs/develop/ktcc/libc.obj/source/stdio/fclose_r.c

@@ -0,0 +1,10 @@
+#include "_stdio.h"
+
+int fclose_r(FILE* stream)
+{
+	if (stream->name) {
+		free(stream->name);
+	}
+
+	return 0;
+}

programs/develop/ktcc/libc.obj/source/stdio/feof.c

@@ -1,5 +1,6 @@
 #include <stdio.h>
 
-int feof(FILE *stream) {
-	return stream->eof;
+int feof(FILE* stream)
+{
+	return stream->flags.eof;
 }

programs/develop/ktcc/libc.obj/source/stdio/ferror.c

@@ -1,5 +1,6 @@
 #include <stdio.h>
 
-int ferror(FILE *stream) {
-	return stream->error;
+int ferror(FILE* stream)
+{
+	return stream->flags.error;
 }

programs/develop/ktcc/libc.obj/source/stdio/fopen.c

@@ -1,7 +1,22 @@
 #include <stdio.h>
 #include <stdlib.h>
+#include <errno.h>
+#include "_stdio.h"
 
-FILE *fopen(const char *restrict _name, const char *restrict _mode) {
-    FILE *out = malloc(sizeof(FILE));
-    return freopen(_name, _mode, out);
+FILE* fopen(const char* restrict _name, const char* restrict _mode)
+{
+    FILE* stream = malloc(sizeof(FILE));
+
+    if (stream == NULL) {
+        errno = ENOMEM;
+        return NULL;
+    }
+
+    FILE* out = fopen_r(_name, _mode, stream);
+
+    if (out == NULL) {
+        free(stream);
+    }
+
+    return out;
 }

programs/develop/ktcc/libc.obj/source/stdio/fopen_r.c

@@ -0,0 +1,81 @@
+#include "_stdio.h"
+#include <errno.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/ksys.h>
+#include <sys/dirent.h>
+
+/*
+	Open new file
+*/
+FILE* fopen_r(const char* restrict _name, const char* restrict _mode, FILE* restrict out)
+{
+	if (!_name || !_mode || !out) {
+		errno = EINVAL;
+		return NULL;
+	}
+
+	bool file_not_exits = false;
+	ksys_bdfe_t info;
+	{
+		int err = _ksys_file_get_info(_name, &info);
+
+		switch (err) {
+		case KSYS_FS_ERR_SUCCESS:
+
+			if (info.attributes & IS_FOLDER) {
+				errno = EISDIR;
+				return NULL;
+			}
+
+			break;
+		case KSYS_FS_ERR_5:
+			file_not_exits = true;
+			break;
+		default:
+			errno = err;
+			return NULL;
+			break;
+		}
+	}
+
+	out->flags.val = 0;
+	out->position = 0;
+	out->name = strdup(_name);
+
+	if (strchr(_mode, 'r')) {
+		out->flags.read = true;
+
+		if (file_not_exits) {
+			errno = ENOENT;
+			free(out->name);
+			return NULL;
+		}
+	}
+	if (strchr(_mode, 'w')) {
+		out->flags.write = true;
+	}
+	if (strchr(_mode, 'a')) {
+		out->flags.val |= _FILEMODE_W | _FILEMODE_A;
+	}
+	if (strchr(_mode, '+')) {
+		out->flags.val |= _FILEMODE_R | _FILEMODE_W;
+	}
+
+	if (out->flags.write) {
+		if (file_not_exits) {
+			int err = _ksys_file_create(_name);
+			if (err != KSYS_FS_ERR_SUCCESS) {
+				errno = err;
+				free(out->name);
+				return NULL;
+			}
+		}
+	}
+
+	if (out->flags.append) {
+		out->position = info.size;
+	}
+
+	return out;
+}

programs/develop/ktcc/libc.obj/source/stdio/fread.c

@@ -14,7 +14,7 @@ size_t fread(void *restrict ptr, size_t size, size_t nmemb, FILE *restrict strea
 	
 	if(size<=0 || nmemb<=0){
 		errno = EINVAL;
-		stream->error=errno;
+		stream->flags.error = true;
 		return 0;
 	}
 	
@@ -23,7 +23,7 @@ size_t fread(void *restrict ptr, size_t size, size_t nmemb, FILE *restrict strea
 		return nmemb;
 	}
 
-    if(stream->mode != _FILEMODE_W && stream->mode != _FILEMODE_A){
+    if(stream->flags.read){
         if(!stream->__ungetc_emu_buff){
 			((char*) ptr)[0]=(char)stream->__ungetc_emu_buff;
 			//debug_printf("Ungetc: %x\n", ((char*) ptr)[0]);
@@ -31,10 +31,10 @@ size_t fread(void *restrict ptr, size_t size, size_t nmemb, FILE *restrict strea
 		unsigned status = _ksys_file_read_file(stream->name, stream->position, bytes_count, ptr , &bytes_read);
 		if (status != KSYS_FS_ERR_SUCCESS) {
 			if(status == KSYS_FS_ERR_EOF){
-				stream->eof=1;
+				stream->flags.eof = true;
 			}else{
 				errno = EIO;
-				stream->error = errno;
+				stream->flags.error = true;
 				return 0;
 			}
 		}

programs/develop/ktcc/libc.obj/source/stdio/freopen.c

@@ -1,62 +1,14 @@
-#include <stddef.h>
-#include <sys/ksys.h>
-#include <errno.h>
 #include <stdio.h>
 #include <stdlib.h>
-#include <string.h>
-#include <sys/dirent.h>
+#include "_stdio.h"
 
+FILE* freopen(const char* restrict _name, const char* restrict _mode, FILE* restrict out)
+{
+    fclose_r(out);
 
-static FILE* _set_errno(FILE *out, int err){
-    errno = err;
-    if(out){
-        free(out->name);
-        free(out);
+    if (_name == NULL) {
+        _name = out->name;
     }
-    out = NULL;
-    return out;
+
+    return fopen_r(_name, _mode, out);
 }
-
-static void _create_file(char *name, FILE *out){
-    if(_ksys_file_create(name)){
-       _set_errno(out, EIO);
-    }
-}
-
-FILE *freopen(const char *restrict _name, const char *restrict _mode, FILE *restrict out) {
-    if(!_name || !_mode || !out){
-        return _set_errno(out, EINVAL);
-    }
-
-    ksys_bdfe_t info;
-    int no_file = _ksys_file_get_info(_name, &info);
-    if(!no_file && info.attributes & IS_FOLDER){
-        return _set_errno(out, EISDIR);
-    }
-    
-    out->eof=0;
-    out->error=0;
-    out->position=0;
-    out->name = strdup(_name);
-    
-    if (strchr(_mode, 'r')) { out->mode = _FILEMODE_R; }
-    if (strchr(_mode, 'w')) { out->mode = _FILEMODE_W; }
-    if (strchr(_mode, 'a')) { out->mode = _FILEMODE_A; }
-    if (strchr(_mode, '+')) { out->mode |= _FILEMODE_PLUS; }
-
-    if(out->mode & _FILEMODE_A){
-        if(no_file){
-            _create_file(out->name, out);
-        }
-        out->position = info.size;
-    }else if(out->mode & _FILEMODE_W){
-        _create_file(out->name, out);
-    }else if((out->mode & _FILEMODE_R)){
-        if(no_file){
-            _set_errno(out, ENOENT);
-        }
-    }else{
-        _set_errno(out, EINVAL);
-    }
-    return out;
-}

programs/develop/ktcc/libc.obj/source/stdio/fseek.c

@@ -1,18 +1,19 @@
 #include <stdio.h>
 #include <sys/ksys.h>
 
-int fseek(FILE *stream, long int offset, int whence) {
-	if (whence == SEEK_SET) {
-		stream->position = offset;
-	} else if (whence == SEEK_CUR) {
-		stream->position += offset;
-	} else if (whence == SEEK_END) {
-		ksys_bdfe_t info;
-	    if (_ksys_file_get_info(stream->name, &info)) {
-	        return -1;
-	    }
-	    stream->position = info.size + offset;
-	}
-	stream->eof = 0;
+int fseek(FILE* stream, long int offset, int whence)
+{
+    if (whence == SEEK_SET) {
+        stream->position = offset;
+    } else if (whence == SEEK_CUR) {
+        stream->position += offset;
+    } else if (whence == SEEK_END) {
+        ksys_bdfe_t info;
+        if (_ksys_file_get_info(stream->name, &info)) {
+            return -1;
+        }
+        stream->position = info.size + offset;
+    }
+    stream->flags.eof = false;
     return 0;
 }

programs/develop/ktcc/libc.obj/source/stdio/fsetpos.c

@@ -1,7 +1,8 @@
 #include <stdio.h>
 
-int fsetpos(FILE *stream, const fpos_t *pos) {
+int fsetpos(FILE* stream, const fpos_t* pos)
+{
 	stream->position = *pos;
-	stream->eof = 0;
+	stream->flags.eof = false;
 	return 0;
 }

programs/develop/ktcc/libc.obj/source/stdio/fwrite.c

@@ -16,7 +16,7 @@ size_t fwrite(const void* restrict ptr, size_t size, size_t nmemb, FILE* restric
 
     if (size <= 0 || nmemb <= 0) {
         errno = EINVAL;
-        stream->error = errno;
+        stream->flags.error = true;
         return 0;
     }
 
@@ -33,11 +33,11 @@ size_t fwrite(const void* restrict ptr, size_t size, size_t nmemb, FILE* restric
         return nmemb;
     }
 
-    if (stream->mode != _FILEMODE_R) {
+    if (stream->flags.write) {
         unsigned status = _ksys_file_write_file(stream->name, stream->position, bytes_count, ptr, &bytes_written);
         if (status != KSYS_FS_ERR_SUCCESS) {
             errno = EIO;
-            stream->error = errno;
+            stream->flags.error = true;
             return 0;
         }
         stream->position += bytes_written;

programs/develop/ktcc/libc.obj/source/stdio/ungetc.c

@@ -6,22 +6,21 @@ int ungetc(int c, FILE* file)
 {
 	int res;
 
-    if(!file){
-        errno = EBADF;
-        return EOF;
-    }
-
-	if (file->mode != _FILEMODE_R){
-        errno = EACCES;
-        return EOF;
-    }
-
-	if (file->position == 0 || c == EOF)
-	{
-	    errno = EOF;
+	if (!file) {
+		errno = EBADF;
 		return EOF;
 	}
-	
+
+	if (!file->flags.read) {
+		errno = EACCES;
+		return EOF;
+	}
+
+	if (file->position == 0 || c == EOF) {
+		errno = EOF;
+		return EOF;
+	}
+
 	file->__ungetc_emu_buff = c;
 	file->position--;
 	return c;

programs/develop/ktcc/libc.obj/source/time/clock.c

@@ -0,0 +1,7 @@
+#include <time.h>
+#include <sys/ksys.h>
+
+clock_t clock(void)
+{
+    return _ksys_get_tick_count();
+}

programs/develop/libraries/http/http.asm

@@ -1618,21 +1618,23 @@ endl
 
 ; Open a new TCP socket
         mcall   socket, AF_INET4, SOCK_STREAM, 0
-        test    eax, eax
-        jz      .error3
+        cmp     eax, -1
+        je      .error3
         mov     [socketnum], eax
         DEBUGF  1, "Socket: 0x%x\n", eax
 
 ; Connect to the server
         mcall   connect, [socketnum], [sockaddr], 18
         test    eax, eax
-        jnz     .error3
+        jnz     .error4
         DEBUGF  1, "Socket is now connected.\n"
 
         invoke  freeaddrinfo            ; Free allocated memory
         mov     eax, [socketnum]
         ret
 
+  .error4:
+        mcall   close, [socketnum]
   .error3:
         DEBUGF  2, "Could not connect to the remote server\n"
         invoke  freeaddrinfo            ; Free allocated memory