Deaicup/Deshab-OS
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
71 Commits 3 Branches 0 Tags
8e6506721da0a834bf7a2eff8da08683239a004b
Commit Graph
40 Commits
This Branch
This Branch
All Branches
Author SHA1 Message Date
Linehu 8e6506721d feat(ath9k): WPA2-PSK crypto core + connect path (auth/assoc/4-way) 
wpa2_crypto.{c,h} (new): self-contained, verified crypto base for the
ath9k STA connect path. No heap, no SSE, DKM-freestanding.
- SHA1 (RFC 3174), HMAC-SHA1 (RFC 2202), PBKDF2-HMAC-SHA1 (RFC 2898)
- PRF-512 (802.11i H.4) for PMK -> PTK
- AES-128 encrypt/decrypt (FIPS-197)
- RFC 3394 AES key unwrap (KEK -> GTK)
- CCMP (RFC 3610 + 802.11i): MIC verify + CTR decrypt, and encrypt for TX
- wpa2_kat(): 7 in-driver known-answer/reference cases (SHA1/HMAC/PBKDF2/
  AES x2/CCMP round-trip+tamper/unwrap). Panics ATH-E02 if any case fails.

ath9k.c: connect path built on the crypto base.
- TX descriptor ring + ath9k_tx_frame (AR9002 minimal field set; bit
  layout to be reconciled with HAL initvals during hardware bring-up)
- management frames: auth (open) / assoc request with WPA2-PSK RSN IE
- EAPOL-Key builder with HMAC-SHA1-128 MIC; M2 and M4 senders
- 4-way state machine driven from RX: M1 (ANonce) -> derive PTK ->
  M2 -> M3 (MIC verify, GTK unwrap) -> M4 -> CONNECTED
- ath9k_tx_impl for data frames (CCMP encrypt); data TX before key
  install panics ATH-E04 (zero-degradation rule)
- netdev now registers tx/connect/disconnect callbacks

build.ps1: Build-DkmDriverMulti to compile+ld -r ath9k.c with wpa2_crypto.c.

Verified on host: all wpa2_kat cases pass (AES-128 = FIPS-197 C.1
69c4e0d8..., RFC 3394 C.1 unwrap = 0011..eeff, CCMP round-trip + tamper).
Hardware bring-up (AR9285 initvals, TX/RX descriptor bit layout, key
install into the MAC) still requires real-hardware iteration.
 2026-10-06 15:11:22 +08:00
Linehu b780a6b868 feat(desktop): taskbar wifi quick card - scan list / password / connect via net api 
- taskbar.c: wifi popup card anchored above tray net icon; net api slots read
  via kernel_api+0x48 hard offsets (UTSM net.h layout, connect@0x48)
- flow: icon click -> scan_start -> per-frame scan_count/result pull -> SSID
  list (security tag + rssi bars) -> open net direct connect / secured net
  password textbox (scan set 1 -> ascii, enter=connect) -> net_api.connect
- no wireless device in QEMU: graceful "not detected" status, no panic (by
  design: wireless dev absence is attest-exempt, not a stub)
- shell.c: handle_key routes to wifi_key when card open (modal)
- sdk kernel_api.h: mirror UTSM dkm_net_api layout (append connect/disconnect
  before device_stats, fixes slot offset mismatch)
- QEMU verified: card opens, renders, graceful no-device path (serial
  "[wifi] no wireless device")
 2026-10-06 13:51:00 +08:00
Linehu bd098f51ec feat(wlan): DKM wlan driver framework + net connect/disconnect ABI extension 
- wlan.drv: multi-vendor PCI probe (class 02:80; Intel AX2xx/Realtek/Atheros ath9k/Broadcom), registers wireless netdev when real hw present (attest net_wlan)
- scan/connect/disconnect: strict panic skeletons (NETWL-E10/E11/E12) until native chip backends (ath9k first, no fw blob) or linuxguest backend land
- net.h/dkm_shared.h: connect/disconnect appended to dkm_net_api + device_desc (ABI-compatible tail append); net.c dispatch with zero-degrade panic
- QEMU (no wlan card): no netdev registered, attest MISSING exempted by FUCK [attest] as designed
 2026-10-06 12:13:59 +08:00
Linehu 14015b78a4 chore: drop realtest image variant (dev/release only) 2026-10-05 21:30:51 +08:00
Linehu ace5711a1e fix(login): atomic ld.lld output (tmp+move) - stop intermittent make failure under Defender lock 2026-10-05 21:24:48 +08:00
Linehu 6caf1aa322 feat: login wallpaper blur backdrop + no logo intro + login spinner; firstInit wallpaper backdrop 
- login: drop ohMyLogo intro, blurred back.rgba as full-screen backdrop (built-in data)
- login: spinner (int 16-seg table, no SSE) below card during desktop load, wallpatch clear
- firstInit: same blurred wallpaper backdrop behind wizard cards
- note: 1.8MB wallpaper data also absorbs the DSK-E70 write-zero minefield -> integrity OK
 2026-10-05 20:12:51 +08:00
LinehuandGitHub 10599b62c4 Merge pull request #4 from Linehu-Studio/fix/bootable-image-and-dkm-build 
fix(build): make the packed image bootable, restore 9 missing DKM drivers
 2026-10-05 15:13:41 +08:00
Linehu ee8b607f9a fix: login cursor ghost after double save-bg + integrity GOT immunity (DSK-E70 workaround) 
- login: remove duplicate cursor_save_bg after draw_login_card (ghost residue)
- login: hidden visibility on __text_*/__expected_hash -> lea rip, len calc no longer depends on GOT slot
- login: dsk_e70_layout_guard keeps .text out of the DSK write-zero minefield (whitelisted, see STATUS.md)
- patch_login_hash: hash first PT_LOAD filesz (matches loader/runtime range) + e_shstrndx section names
- STATUS.md: register DSK-E70 (unresolved ibuf-tail write-zero, root cause open)
 2026-10-05 02:34:15 +08:00
LinehuandGitHub 3aebe962be Merge pull request #2 from Linehu-Studio/build/fix-texture-stamp-guard 
build: fix desktop texture regeneration crash on clean tree
 2026-10-05 01:44:25 +08:00
Linehu a570fe09b5 feat: white cursor with black outline (desktop/login/firstInit) + fix hash patcher param 
- outline = empty pixel with any 8-neighbour filled, stays inside cursor save region

- patch_login_hash.ps1: typed byte[] params + ASCII comments, default path fallback (was silently misbinding under some hosts)
 2026-10-04 22:43:18 +08:00
Linehu b4d9c8e7ed feat: desktop wallpaper from material pack back.rgba (889x500) instead of system/wallpaper.rgba 2026-10-04 22:21:14 +08:00
Linehu 87f8ce11da feat: all-white mouse cursor across desktop/login/firstInit 2026-10-04 21:43:43 +08:00
Linehu 87d532df01 feat: SAS-R0 boot-time capability attestation - unimplemented feature blocks boot 
- attest.c: subsystems report readiness, attest_check() before DSK handoff

- FUCK [attest] require_<name>=1 (default) + not-ready -> lotus panic with feature errcode

- wired: sealed_mem/pckc/sched/drr/fs_fat32_rw/net_wired + placeholders (pkg/brand/dkm/nvme/virtio/wlan/vmx/drr_real_mac)

- negative test: require_pkg=1 -> boot-time PKG-E10 panic (verified); dev/release FUCK whitelist = STATUS.md undone items

- CLAUDE.md whitelist rule updated
 2026-10-04 21:34:50 +08:00
LinehuandGitHub 25fbabe368 Merge pull request #1 from Linehu-Studio/fix/del_o_and_fix_build_in_linux 
build: untrack generated .o/.d and allow CC/LD override
 2026-10-04 20:47:21 +08:00
Linehu 055c521ceb chore: gitignore only blocks regenerable binaries - demo video via LFS 
- ignore: *.o/*.elf/*.drv/*.img/*.rlib/*.rmeta/*.pdb/*.dbf/*.rgba/target//.build_tmp (all regenerable)

- keep: source assets (png/ttf/webp), vendored payload (SYSTEM/lib dll, SYSTEM/bin exe), demo video (TESTSHOW/*.mp4 via Git LFS), text build logs (.d/.timestamp)

- untrack 290+ build artifacts from index
 2026-10-04 20:03:06 +08:00
Linehu 5f01904be1 feat: white-tone desktop + login crossfade + BUG-031 FAT chain fix + DESK-E texture panics 
- BUG-031: fat32_io.h on-demand FAT sector read (FAT beyond 128KB cache walked OOB, font/wallpaper corrupted past first clusters); host unit test diff=0

- material pack: gen_textures.py wired into build.ps1, startMenuLogo.rgba packaged; missing wallpaper/logo/font now fatal (DESK-E02/E03/E04 red fatal screen)

- login->desktop crossfade 16 steps, boot typewriter screen removed

- IDE exit: Esc detaches and returns to desktop page

- white theme: KS_* light palette, g_wf_light taskbar, hardcoded dark fills replaced

- QEMU verified: labels/clock render, wallpaper + start logo loaded, zero spurious panic
 2026-10-04 15:53:04 +08:00
Linehu 174af71b39 feat: exhaustive STUB instrumentation - 241 panic sites across VMM/kernel + whitelist alignment 
- VMM cluster: XJ3-E/LIN-E/VMM-E/VMX-E/EPT-E/IPC-E/VVB-E (124 sites)

- kernel cluster: TST-E/MM-E/UTRW-E/LAPIC-E/IDT-E/SAS-E/PRB-E/NETST-E/DKM-E (117 sites)

- whitelist fixes: vmm_init env probes E10-E14 fallback, disk_log read-only block path, DSL-E05 non-FAT32 device 0 probe

- QEMU smoke: wizard -> USER.CONF -> login -> desktop ready, zero spurious panic
 2026-10-03 21:26:04 +08:00
Linehu 506d1b18c6 feat: SAS-R0 signature-feature attestation - not-ready means very-early panic 
- SAS-E01: scheduler bootstrap failure was "non-fatal" -> panic (O(1)
  scheduler must stand up or nothing runs)
- PCKC-E01: boot-time key-derive roundtrip in utsm_init (get_or_derive
  probe segment, all-zero key or error -> panic; probe slot invalidated)
- DKM-E04..E07: manifest required-driver failures (module missing /
  ELF invalid / symbol scan fail / load fail) -> panic instead of
  "abort stage" silent continue
- QEMU verify: pckc derive roundtrip ok, sealed MAC established,
  selftest OK, boot reaches wizard, zero spurious panics

[STUB] P0/UTSM SAS-R0 attestation | 剩余STUB命中=101（违规=0） | “特色不是招牌，是性命” | 已提交=<this>
 2026-10-03 15:49:19 +08:00
Linehu dda0cc9dd0 feat: sealed memory early init + verify - UTSM-E01..E04 very-early panic 
- NEW core/sealed.c: utsm_sealed_early_init() runs right after utsm_init:
  empty .sealed section -> UTSM-E01 panic; no root key -> UTSM-E02;
  establishes BLAKE2b(root_key) MAC over [__sealed_start,__sealed_end).
- utsm_sealed_verify(): MAC mismatch -> UTSM-E03; verify-before-init ->
  UTSM-E04. Wired into utsm_selftest (U3+ step).
- .sealed region now non-empty by construction: sealed brand record
  "DESHAB-SEALED-MEMORY-V1-DEAICUP" lives in section(".sealed").
- Before this commit nothing ever referenced __sealed_start/__sealed_end:
  the sealed section was never registered, MACed or verified.
- QEMU verify: [SEAL] region=0x20 bytes, MAC established, selftest OK,
  boot reaches wizard, zero spurious panics.

[STUB] P0/UTSM/core/sealed.c | 剩余STUB命中=101（违规=0） | “封缄即落款，空区即自爆” | 已提交=<this>
 2026-10-03 15:40:00 +08:00
Linehu 301006d611 feat: re-instrument per STUB rules - all silent placeholders now panic 
- PE-E03 pe_loader.c: unsupported base-reloc type silently skipped -> panic
- PE-E04 pe_shim.c ReadFile: fake TRUE+0-bytes ("pretend EOF") -> panic
- PE-E05 pe_shim.c CreateFileA: no file backend -> panic (no honest-failure excuse)
- PE-E06/E07 x86emu32.c: emu IAT unimplemented API / shim index OOB -> panic
- NETST-E01 net.c: netdev stats all-zero placeholder ("lying zero traffic") -> panic
- apic.c comment: SIPI reserved now points to DKM-E03 skeleton
- pe_dll_manager comments: circular-dep stub call now documented as PE-E01
- virtio_blk.c: capacity declaration marked as instrumentation whitelist
- QEMU verify: wizard -> Save -> verify ok -> login success -> desktop ready,
  zero spurious panics

[STUB] P1/UTSM/pe+net re-instrument sweep | 剩余STUB命中=101 (全部为 panic 坐标注释 /
IDT halt-stub 防御术语 / 白名单) | “未成之物，不留活口”
 2026-10-03 12:44:57 +08:00
Linehu cd39caa10c docs: CLAUDE.md add STUB instrumentation rules - grep-verified zero-fallback contract 2026-10-03 12:31:03 +08:00
Linehu 583262f5f0 feat: SKEL skeleton instrumentation - designed-but-unimplemented paths panic 
- hypercall.c: 5 Phase 1.3 stubs -> HCALL-E01..E05 panic (no NOSYS return)
- drr_stub.c: snapshot alloc fail -> DRR-E01; new skeletons DRR-E02 root key
  seal, DRR-E03 task kill crypto erase, DRR-E04/E05 segment/system rollback,
  DRR-E06/E07 driver recovery ops dispatch
- sched.c: enable fail -> SCH-E02 panic (no silent demo skip)
- dkm/kernel_api.c: D1 hot-unload DKM-E01, D2 bare code loader DKM-E02,
  SIPI AP wake DKM-E03 skeletons
- virtio_net tx/rx stubs -> VNET-E01/E02; ath9k tx stub -> ATH-E01;
  xhci port reset -> XHC-E01; nvme MSI/MSI-X fail -> NVME-E01..E03 (no
  silent polling downgrade); e1000 MAC fallback -> E1K-E01
- pe_shim unimpl API -> PE-E01; PE ordinal import -> PE-E02
- kernel/main.c: lib sync / vscode install fail -> LNX-E01/E02
- NEW brand/brand.c (E1 boot anim BRAND-E01, E2 time egg BRAND-E02) and
  pkg/pkg.c (pacman skeleton PKG-E01..03), wired into UTSM MAKEFILE
- Deliberate design-internal fallbacks kept and documented (VMX skip,
  bzImage skip, KDE fallback, boot-module fallback, apic->PIC route)
 2026-10-03 00:30:54 +08:00
Linehu 38ff3cf93c feat: FS-E strict filesystem mode - ext4 rw driver, FAT32 big-disk fix, forced re-login 
- CODE/dsk/ext4.c: minimal full ext4 read/write (GPT probe, feature gate,
  extent depth 0/1 rw, block/inode alloc-free, dir insert, write-verify);
  boot selftest PASS on p2. Unsupported features panic (FS-E2x), zero fallback.
- DSK fat32 write: on-demand FAT entry RMW + chunked free-cluster scan
  (fixes "no free clusters" on 5.3G ESP); rec_len shrink offset (+4) and
  chunk alignment fixes. USER.CONF write-verify ok, firstInit flip ok.
- First-boot: wizard -> persist -> forced login (FS-E06 if skipped/failed)
  -> desktop. login.elf loaded via system/user/use fallback.
- FirstInit network page: Connect now panics NET-E01 (no wlan stack);
  Save (config only) is the only path forward.
- FirstInit/DSK fatal: ohMyLogo lotus fatal screen (NET-E01 / FS-E series).
- Boot screen: Logo.png image removed, "DESHAB P01" text centered only;
  dcp/xj logos regenerated from TEXTURES/.
- run_qemu.bat: main GPT image on SATA ide.0 (AHCI provider must see ext4).
 2026-10-02 23:49:05 +08:00
Linehu c4a5295a81 feat: UI strict-error mode - ohMyLogo login intro, typewriter removed, fatal instrumentation 
- desktop boot screen: typewriter effect removed, whole screen drawn in one pass
- login: ohMyLogo (DEAICUP lotus, 288px embedded) intro screen before password card
- strict error policy instrumentation (zero degradation): login LOGIN-E01..E05 (integrity/badctx/noconf/nouser/nohash) and desktop DSK-E01 now full-screen fatal with symbol + halt instead of silent skip
- root-cause fixed: missing .build_tmp/mkfat32.ps1 silently left SATA sidecar stale (login.elf 'not found' degradation) - replaced with CODE/linux/make_sata_sidecar.sh (auto adaptive FAT32 rebuild each build)
- recreated lost CODE/login/patch_login_hash.ps1 (.text SHA256 -> .hashseg)
- dev config now full flow: show_logo=1, wizard+login enabled, dev shell off
 2026-10-02 21:07:05 +08:00
Linehu 3aa15e7b21 feat: D4 DKM colored log prefixes - every driver has its own temper 
- log.c: color context (log_set_color / log_color_enable); ANSI only goes to serial, BOOTLOG.TXT buffer stays clean plain text
- kernel_api.c: dkm_log_set_driver() — driver-name FNV-1a hash into an 8-color bright ANSI palette, deterministic per driver
- load sites (dkm_core builtin + boot_modules external) set color around driver_init and clear after
- FUCK [debug] color_log gate (dev=1, release=0)
- QEMU verified: 1150 colored lines, per-driver stable colors (.build_tmp/qemu_serial_d4color.log)
 2026-10-02 19:48:54 +08:00
Linehu 068e0e37c3 docs: README 项目特质章 - explain what makes Deshab Deshab 
- new traits section: 疯但好使 / 毛坯神殿非赛博保险柜 / 仪式感是功能(@sealed旋钮+莲花崩溃屏) / 诚实比体面重要 / 错也错得有章法
- refresh stale status refs (UTRW slow path + KAT, phase 1-2 done) and build section (cross-platform Makefile)
 2026-10-02 19:37:07 +08:00
Linehu ba16217afe feat: U3 @sealed optional segment marking - phase 2 complete 
- UTSM_SEG_F_UNSEALED: plaintext direct-store segments (zero crypto overhead) while keeping capability checks, dirty tracking, intent/commit log and seqlock protocol; default remains sealed
- linker.ld: .sealed output section with __sealed_start/__sealed_end symbols for future static sealed data
- selftest: unsealed segment stores plaintext verifiably + read-back, full SELFTEST PASS (.build_tmp/qemu_serial_u3sealed.log)
- phase 2 (UTSM/UTRW) now complete in ROADMAP_INSANE
 2026-10-02 15:30:25 +08:00
Linehu 1e74ee9fab feat: U5 land deterministic chacha20 KAT (RFC 8439 2.4.2) 
- test/gen_chacha_kat.py emits KAT from RFC 8439 section 2.4.2 mapped to the impl's 64+64 counter/nonce layout; expected bytes cross-checked against pycryptodome (independent impl) and the official vector: 10f1e7e4d13b5915... byte-exact
- verdict: the 'intermittent chacha20 KAT FAIL' referenced in FUCK comments was a phantom — that test was never merged into this tree; the implementation itself matches RFC byte-for-byte
- selftest now runs KAT + page-MAC determinism checks at selftest=2; full SELFTEST PASS in QEMU (.build_tmp/qemu_serial_u5kat.log)
 2026-10-02 15:24:15 +08:00
Linehu 7925c09677 feat: U4 ciphertext page dump tool 
- utrw/debug.c: utrw_debug_dump_page(slot,page,offset) — kernel debug-privileged dump of 64B ciphertext + decrypted plaintext hex to serial (PCKC key derive, no capability path)
- selftest demo: dump at offset 128 shows plaintext 'deshab utsm sealed memory selftest' vs unrelated ciphertext; SELFTEST PASS (.build_tmp/qemu_serial_u4dump.log)
 2026-10-02 14:40:26 +08:00
Linehu 42fc06439c feat: R3b wire drr_emergency_pool_kb / drr_recovery_log_size from FUCK 
- drr_apply_config(pool_kb, log_size) before drr_init; pool_limit clamped to compile-time 64KB static pool (min sub-buffer layout enforced, defaults kept on bad values); recovery log size clamped to remaining pool
- verified in QEMU: pool_kb=32 -> pool_limit=0x8000, log_capacity=0x80, boot continues (.build_tmp/qemu_serial_r3b.log)
 2026-10-02 14:34:35 +08:00
Linehu b76e8f1556 feat: U2 UTRW slow path real branches 
- slow_path.c: KEY_MISS -> PCKC re-derive (retry OK); EPOCH -> refresh cap epoch via out_cap (retry OK); STALE_CAP permanent; MAC_FAILED/POISONED -> PCKC invalidate + DRR fault report; RETRY passthrough
- status.h: add UTSM_ERR_MAC_FAILED (-11); utrw.h: slow path takes cap/rights/offset/len/reason + out_cap
- selftest: EPOCH refresh + KEY_MISS re-derive + POISONED passthrough coverage, full SELFTEST PASS in QEMU (.build_tmp/qemu_serial_u2slowpath.log)
 2026-10-02 14:29:28 +08:00
Linehu 1eaf1aedf8 docs: F3 cross-platform build status (Makefile) 2026-10-02 14:22:24 +08:00
Linehu 23c29f7980 feat: F3 cross-platform root Makefile (Linux/macOS make dev|release|realtest) 
- mirrors build.ps1: kernels (UTSM/DSK), user apps, tools, 8 DKM .drv drivers (clang -c + ld.lld -r), config injection from build/configs/<variant>, packing via CODE/linux/pack_system_image.sh
- panic logo regeneration hook (ohMyLogo.png newer -> gen_panic_logo.py)
- Windows still primary via build.ps1; Makefile targets POSIX (pack needs sudo + mkfs)
 2026-10-02 14:22:00 +08:00
Linehu 6e44fa5df7 feat: DEAICUP lotus panic screen + F2 fbcon abstraction 
- panic screen now draws embedded ohMyLogo.png (384px RGBA, gen_panic_logo.py); background color sampled from image corners so screen bg matches logo bg
- F2: include/utsm/fbcon.h unified framebuffer console (pixel/blend/fill/rgba-image/text with FBCON_HAVE_FONT opt-in); panic.c migrated onto it
- build.ps1 regenerates panic_logo_data.inc when ohMyLogo.png changes
 2026-10-01 23:50:04 +08:00
Linehu e4307ff0ec feat: F1/F4/F5 phase-1 skeleton + insane-roadmap docs rework 
- docs: ROADMAP_INSANE (five-phase plan), TASKBOARD, STATUS single source of truth, PHILOSOPHY; rewrite README/ARCHITECTURE non-goals; archive ROADMAP phase 0-9; drop contradictory dev-plan doc
- F1: unified lotus panic screen (panic_full: fb draw + regs dump + DRR crash archive); IDT exception path and dkm_api_panic routed through it
- F4: leveled boot selftest (FUCK [boot] selftest=0|1|2), IDT/CS fatal checks + DMA probe
- F5: [dsk] skip_firstinit direct-to-desktop; build/configs dev+release templates now injected
 2026-10-01 23:29:49 +08:00
Linehu 1074962546 ohhhhhhhhhhh (exclude linux-extra-rootfs.img build artifact) 2026-10-01 23:13:42 +08:00
LinehuandCursor 737a2cb0db fix: let nested-KVM vmlaunch succeed with Virtual NMIs 
Self-test and Linux guest were failing VM-instruction error 7. Share a KVM-style control adjust, request PIN_VIRTUAL_NMIS, and stop intercepting guest #PF. Land isolated XNU/Python/KDE scaffolding off the boot path.

Co-authored-by: Cursor <cursoragent@cursor.com>
 2026-09-06 10:48:07 +08:00
Linehu cc0fcf26e3 FU 2026-08-29 18:47:54 +08:00
Linehu 880906edc9 docs: 添加 Apache 2.0 许可证 (Apache-2.0) 2026-08-10 13:54:51 +08:00
Linehu 2a50fa58c7 FUCKKKK 2026-08-09 22:20:01 +08:00