| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Root cause: on real hardware SetMode can invalidate the pre-captured
ExitBootServices map key, so EBS fails and the retry refreshed the map
via capture_memory_map -> init_heap, which REBUILT the free list from
the raw map and wiped the allocation state of every pre-EBS live object
(the display shadow and the vt grids). Their spans became free while
still live, later allocations carved inside the shadow, and the display
writes clobbered them — zeroed back-pointers (family A ptr=0x404443e0
block=0x0) and gray glyph pixels (family B 0xaaaaaa00aaaaaa = 0x00AAAAAA
= PALETTE[7] fg-7 text colour, not a firmware scrub).
Fix: refresh_and_exit now re-captures the map and returns a fresh EBS
key without rebuilding the free list (refresh_map_key); the free list is
the authoritative live-allocation state. Verified in QEMU under a forced
EBS failure (SIMULATE_EBS_FAILURE test hook): verify ok, probe NOT free,
flhead outside the shadow — identical to a clean boot, and the full
shell sequence is crash-free.
Also: removed the crash-family debug markers from the screen (boot-time
heapdbg lines, vt verify marker, shadow marker) — the vt grid self-check
stays serial-only. Version 0.5.8 -> 0.5.9.
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
shadow marker
The klog -t 2 crash drops a 126-byte String (the vt verify klog line)
whose data ptr 0x40444970 points into the first-heap-block region whose
content is the firmware 0x00AAAAAA scrub — a stale pointer, not a real
allocation (back-pointer slot never written). Upgrades so the next
real-hardware crash shows the full picture:
- guard: explicit candidate block bases (b=ptr-24 / b16=ptr-32) with
their size/free fields, plus a 16-word window (mem..mem4) from
ptr-0x30 instead of the old 8 words
- vt::debug_verify: content probe of the cells grid (must be all-BLANK
at shell entry) — catches a stale in-heap cells pointer that passes
the range checks
- gop: on-screen marker for shadow carved vs NO-SHADOW fallback
(the scrub content at the crash ptr suggests the shadow was rejected
and its block stayed free + firmware-scrubbed)
|
| |
|
|
|
|
|
|
| |
The OptiPlex crashes free garbage (0xaaaaaa, 0xaaaaaa00aaaaaa) whose
value is the firmware's 0x00AAAAAA scrub pattern left in stack/heap.
The drop size/align (the dealloc layout) now prints on the band, so the
next repro shows both the garbage ptr AND the size it was freed with —
cap×elem_size pinpoints which struct's fields were clobbered.
|
| |
|
|
|
|
|
|
| |
The OptiPlex repro copies are truncated at the ~110-char band width, so
the old single-line 'stk:' (12 words, >200 chars) lost the return chain
after a few words. Split mem/stk dumps into 4-words-per-line blocks:
mem+mem2 (8 words around the corruption) and stk..stk4 (16 words) —
the drop caller chain now survives hand transcription.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The OptiPlex #GP (vec=13 in __rust_dealloc) frees a garbage pointer: the
back-pointer at [ptr-8] is clobbered by the time the shell drops its
Parsed struct. A blind store to the bogus block header was an opaque
#GP — now it is a readable report:
- dealloc validates the back-pointer (in-heap block, sane size, in-use)
before trusting it; on failure it prints ptr/block/heap range, the
memory around the block header, and the live stack chain, writes the
crash band + crash.log, and halts (memory.rs: heap_guard_report).
- fault dumps now carry the full register set (rax..rdi, rbx, r8..r15)
so the next repro shows the exact freed ptr and clobbered value.
- new 'corrupt' shell hook deliberately clobbers a live block's
back-pointer to prove the guard fires (test-fault-style halt).
Verified: test-msg/test-disk/test-fault pass; corrupt hook reports
ptr=.. block=0xdeedbeef... with mem dump instead of #GP. rs-dos.img
rebuilt with the release kernel.
|
| |
|
|
|
|
|
|
|
|
|
| |
- fault_dump prints img= (kernel image base) on the serial line and
the on-screen band, so a real-hardware repro can be resolved against
the PDB without serial
- backtrace: rbp-chain walk then a raw scan of the stack above the
interrupted rsp; both filter to the .text range (PE-header parse via
memory::text_range) so only code addresses are collected
- on-screen boot marker shows the kernel image range at every boot
- crash.log now records img= + bt= lines alongside rip/rsp
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The fault handler now writes crash.log to the volume labeled CRASHLOG
(label-selected via fat::crash_vol), falling back to slot 0. On real
hardware the boot USB may be unwritable, so a second stick labeled
CRASHLOG survives the write; the shell auto-prints the previous crash
report from it at next boot (read_crash_log resolves /disk/<uid>/crash.log
through the FAT path, since Disk nodes hold no cached children).
- fat.rs: volume_label() + crash_vol() (no allocation)
- fs.rs: write_crash_log targets crash_vol(); read_crash_log prefers the
CRASHLOG disk node; drop now-unused first_disk()
- mkdisk.sh: --label option
- test-fault.sh: boots with a dedicated CRASHLOG stick (mkfs -F 32 -n
CRASHLOG), verifies crash.log lands there and NOT on the ESP
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
User's machine: boot reached 'rsnix: boot done — shell' then panicked
'len is 0, index 0' at the first cell write — CELLS (the vt grid) was
valid at vt::init and empty at the first shell paint, i.e. something
stomped the heap between vt::init and the shell banner. The only
change in that window vs the working v0.5.3 was the v0.5.4 move of the
1.9 MB display shadow from the heap to the fb-aware page allocator:
on this machine's memory layout the page-allocator shadow's fill loop
(480K words) clobbered the heap's vt vectors.
- Shadow is again the *first* heap allocation, in gop::init before
vt::init builds its grids (the v0.5.3 layout that provably boots on
the user's hardware) — nothing else in the heap can collide with it.
- init_heap now splits the free list at the GOP framebuffer hole
instead of subtracting its size from one block that still *spanned*
the hole — no heap allocation of any size can land on the display.
- Kept the panic-proof try_reserve + fb-overlap rejection + no-shadow
warning band; dropped the late init_shadow() call in main.rs.
- Verified on QEMU: shadow @ 0x1980020 origin=heap, allocated before
vt::init, fb clean at 0x80000000; all 5 suites PASS; 0 warnings.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The real-hardware scroll crash (white bands + instant freeze, QEMU-clean)
was invisible because the only trace was a single serial line on machines
with no serial console. This makes every crash self-identifying and
hardens the real-hardware failure surfaces:
- Every CPU exception vector 0-31 now has an IDT entry (macro-generated
stubs; 8/14 keep their IST stacks). A fault in the shell prints a
white-on-red band to the framebuffer via gop::dump_lines (raw fb
writes bypassing the vt, fully clamped) plus a parseable serial line:
'!!! <kind> vec=N err=… rip=… cr2=… rsp=… rflags=… task=N "name"',
then halts — instead of converging to a cryptic double fault.
Background-task faults abort the task as before.
- Boot log now reports the authoritative post-SetMode geometry:
'gop: fb base=… size=… pitch=… WxH bpp=… bgr=…', so the next real
machine's report identifies whether pitch*height exceeds the mapping.
- Framebuffer writes (clear, push_rect, shift_up_n bulk push, cursor
overlay) are clamped to the firmware's framebuffer_size: a pitch-
padded mode degrades to a truncated view instead of faulting past
the mapping.
- The GOP framebuffer region is excluded from the page allocator (and
subtracted from the heap's conventional-region pick) even when the
firmware lists it as ConventionalMemory — kernel allocations can
never overwrite the display again. The GOP protocol is located and
peeked before the memory map is captured so the exclusion is known
at heap-init time.
- Shell gains a hidden 'fault' command (deliberate unmapped write) and
'make test-fault' verifies the dump: vec=14, cr2=0x400000000000,
task=0, halt — no triple fault. All other suites re-pass.
- CONTEXT.md: fault-dump format documented for real-hardware reports.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
(fmtbuf/scripts/hexdump/cat/ahci/pci)
- memory.rs: self-test Layout::from_size_align no longer panics — the
test fails cleanly on an invalid layout
- fat.rs: v() no longer unwraps the active-volume slot — invariant
violation logs + halts instead of panicking mid-disk-io
- strbuf: shared FmtBuf replaces the duplicated PanicBuf/FmtBuf
(main.rs + sched.rs)
- test-disk.sh: SC2046 — conditional --fresh arg no longer relies on
word splitting
- scripts: boot.sh/boot-image.sh share lib-boot-args.sh (dup args loop)
- shell: cat/hexdump stdin pairs unified with Option<&str> marker;
hexdump format_row folded into format_partial
- ahci: read/write_sectors unified behind dma_sectors
- pci: config_read/write share config_addr
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
- edition = "2024" in the workspace manifest and rsdosedit
- extern blocks must be unsafe: interrupts.rs ISR stubs -> unsafe extern "C"
- #[no_mangle] -> #[unsafe(no_mangle)] (efi_main, isr_dispatch)
- One missed static-mut read in free_list_insert wrapped in an explicit
unsafe block (edition 2024 removed the implicit unsafe context inside
unsafe fns)
- rustfmt reflows for 2024: nested if-lets become let-chains, imports
re-sorted, rsdosedit help-screen box rows wrapped
No codegen-affecting changes. Validated: cargo check/clippy/fmt clean
(host + uefi target), make test-disk PASS, editor round-trip PASS.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Warning cleanup (cargo check + clippy now 0 across the workspace):
- Migrate all 21 static_mut_refs sites to &raw mut accessors (fat, gop,
ps2, vt, ahci, interrupts, memory) so edition 2024 won't deny them
- panic = "abort" profiles: bare cargo check/build now works without
nightly/-Zbuild-std
- Wire TICKS in the timer ISR (it was never incremented, so the vt scroll
batching trigger was dead code)
- Clippy: efi_main becomes unsafe extern "efiapi" with a # Safety doc,
KeyBuf::new uses inline-const array repeat (no interior-mutable const),
editor main loop is a while-let, PIT control byte regrouped
- [workspace.lints.clippy] allows result_unit_err + needless_range_loop
(deliberate kernel style); large_enum_variant allowed on LineInput
- Drop now-stale static_mut_refs allow from the rsdos bin
FAT32 correctness fixes (both pre-existing, exposed by the SFN fix):
- Unique 8.3 short names: to_8_3() never disambiguated collisions, so
files sharing a long-name prefix got identical SFNs and fsck reported
duplicate directory entries. Add dir_short_names() + unique_short_name()
(~N tails, Windows-style) and force LFN when the SFN changes.
- Multi-cluster directories: scan_dir/dir_short_names stopped at the first
0x00 slot, hiding every entry written into a chained cluster (writes
failed with 'cannot write', duplicate SFNs accumulated); add_dir_entry
now packs the cluster (0xE5 deleted-markers) before extending so other
FAT readers (mtools, Windows) see the continuation too.
Other: delete dead fields (Disk.fis, Volume.fat_count), drop timer_irq/
keyboard_irq/rendered, refresh_and_exit is a documented unsafe fn, CONTEXT
.md corrected (stable toolchain, IRQ input, 4-crate layout), .gitignore
excludes .pi-subagents artifacts.
Validated: cargo check/clippy/fmt clean, make test-disk PASS on fresh and
reused images (incl. two directory-chain extensions and ~10 SFN tails),
editor round-trip smoke test PASS.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The cursor was blitted as an inverse-video cell directly onto the
framebuffer, but that inversion was never recorded. RENDERED (the
diff grid) only stored the logical cell, so when scrolling shifted the
raw pixels up (gop::shift_up), the stale inverse block rode along
while the grids said 'blank' — and blank-vs-blank never re-blits.
Result: a cursor-shaped block at the end of every scrolled line,
surviving even 'clear'.
- flush() now mirrors the inverse colours into RENDERED when drawing
the cursor (and restores RENDERED when erasing it), and the diff
loop skips the visible cursor cell (owned by the cursor pass).
Any scroll/clear then sees CELLS != RENDERED at a stale-inverse
position and repaints it.
- memory.rs heap self-test: snapshot heap state before its own
allocations and assert the heap returns to that state, instead of
asserting it coalesces to a single empty block. The vt cell grids
(allocated earlier in boot) made the old 'free_blocks == 1' check
fail; the new check is correct regardless of prior live allocations.
Verified in QEMU: hexdump /dev/mem streams hundreds of lines, ^C,
screendump shows exactly one inverse cell (the live cursor); 'clear'
leaves exactly one (the fresh prompt cursor). heap self-test OK again.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
- uefi.rs: full EfiMemoryType constants (0-14) + memory_type_name(); fix
MEMORY_TYPE_LOADER_DATA from 6 to 2 per the UEFI spec
- memory.rs: replace the bump allocator with a first-fit free-list heap
(splitting, coalescing, 8-byte back-pointer for alignment-safe free);
add a bitmap physical page allocator (heap region reserved, page-0
guard, aligned runs); grow the heap from the page allocator on demand
- paging.rs: identity page-table helpers — adopt_tables() rebuilds the
firmware map into our own tables (OVMF maps its tables read-only, so
the firmware tree can't be modified in place) and switches CR3;
map_page/unmap_page/is_mapped with 2 MB/1 GB large-page splitting
- main.rs: boot self-tests for heap, pages, and paging
|
|
|
Architecture:
rs-dos/ (workspace root, was raisernics/)
├── rsnix/ — kernel lib crate (uefi, serial, gop, ps2, interrupts, memory)
├── shell/ — shell lib crate (prompt, commands, line input)
├── rsdos/ — binary crate (efi_main, init, panic handler)
└── rsboot/ — future git submodule
No functional changes — boots to shell prompt identically.
Version auto-pulled from Cargo.toml via env!("CARGO_PKG_VERSION").
|
