docs/plan/hybrid.md mechanism 3. A process asks for a device's interrupts with
PCIIOC_IRQ and then reads them, which is how Linux's uio does it and for the
same reason: a descriptor composes with poll, and a driver that stops reading
falls behind rather than losing events. read() returns a count rather than a
flag, so a driver that was busy sees it missed three and acts once.
Proven against an e1000e by asserting a cause through its ICS register and
reading the interrupt back out in userspace:
interrupts routed here (line 10)
interrupt delivered, count = 1
ICR = 0x80000004
0x80000004 is the cause the test set plus INT_ASSERTED -- the device's own
register, read by userspace as its acknowledgement.
**The masking is the design, not a detail.** A PCI line is level-triggered and
stays asserted until the cause is cleared at the device, which is a different
register on every chip and so is something only the driver can do. The kernel
masks the line on arrival and userspace's write() unmasks it. Signalling EOI on
a still-asserted line re-enters the handler immediately and for ever; this
kernel did exactly that by accident at 72,000 interrupts a second, and what it
looked like from outside was a clock running seven times slow (§13).
Which decides what happens when a driver dies mid-interrupt: the line is left
**masked**. The instinct is to unmask -- do not leave the device dead -- and it
is wrong, because unmasking an asserted line with no handler is a crashed
driver taking the machine down with it. Masked is safe and not permanent: the
next PCIIOC_IRQ arms it. Both halves verified by crashing a driver on purpose
and then starting another one.
A line already carrying kernel handlers is refused outright. Masking is
per-line, so handing out a shared one would silence whichever kernel driver is
also there, with nothing logged. MSI is the real answer and pci_msi_enable() is
already in the tree waiting for a caller.
Also fixed, found by running the crash twice: DMA buffers were never returned
when a process exited. The arena is 4 MB and contiguous, so a driver that
restarts a few times exhausted it -- "out of DMA memory, 3584 free" -- which is
a poor property for an interface whose whole promise is that a driver can crash
and come back.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
docs/plan/hybrid.md mechanisms 2 and 4: MMIO and DMA to a process.
A process can enumerate the bus, claim a device exclusively, map its BARs and
get memory the device can be pointed at. Proven end to end against an e1000e
-- the exact device class we cannot drive and would want a foreign driver for:
device 7: 8086:10d3 (network)
claimed
BAR0 at 0x81060000, 128 KB
regs[0..3] = 0x00140261 0x00000000 0x00080283 0x00000000
MMIO reaches the device
DMA 8 KB at virtual 0x40001000, physical 0x3f8eb000
Those are the e1000e's CTRL and STATUS with full-duplex and link-up set, which
is how you tell a real mapping from one reading 0xFFFFFFFF off an unmapped
address. Claiming a device the kernel drives is refused -- two writers of one
register set looks like hardware that has started lying -- and everything a
process held is released when it exits, because a driver that crashed must not
keep its hardware away from its replacement.
mmap's offset is the BAR's physical address, checked against the BARs of
devices *this process has claimed*. Without that check the device is /dev/mem
with extra steps and a process could map the kernel's own memory. Mappings are
write-through: a cached write to a doorbell arrives whenever the cache feels
like it, which from a driver's side is indistinguishable from hardware that
ignores you.
The interface is about PCI devices, not about any driver model, and its first
users will be SchOS's own drivers when virtio-gpu is evicted. That ordering is
the design rather than an accident: an interface whose first consumers are
native drivers is a general mechanism something foreign can use later, instead
of a hole shaped like something foreign.
Also: PROT_* and MAP_* moved to include/posix.h. They were defined only in
sys_mem.c, so cmd/fbinfo carried a private copy and this wanted a third --
two copies of a number the kernel compares against, where a mismatch shows up
as a mapping with the wrong permissions rather than an error.
No IOMMU, so this isolates a driver against its own bugs and not against its
own DMA. The plan says so at length and it is still true.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
docs/plan/init.md milestones I2 and I3. The rule change.
**I2 -- the heartbeat.** The failure the two-init scheme could not see is that
an init usually does not die, it *hangs*: wait() never returns for a process
that is merely stuck, so a supervisor watching a wedged init reports
everything normal for ever. PID 1 now pings every two seconds and expects an
answer within ten; silence means stuck, and stuck is worth acting on.
Exercised rather than reasoned about, as the plan asks: `sysrq w` wedges
schinitd by making the scheduler pass over it for ever. It stays alive, keeps
its descriptors and appears in every listing -- which is the point.
Two things that only showed up by running it:
- kill() sets a flag the target checks when it next runs, and the premise
here is that it is not running. Waiting for a corpse that may never be
produced turns the recovery path into a second hang. PID 1 replaces the
supervisor *without waiting*; the old one's socket is closed, so the moment
it runs again its first write fails and it exits.
- the replaced supervisor must notice the hangup. poll() reports POLLHUP and
never POLLIN again, so a loop that only looks for POLLIN sits there
complaining that PID 1 has stopped pinging it.
**I3 -- proc_reincarnate().** exit() on initproc panicked. Now the kernel
tears down nothing, keeps pid 1, and re-execs /bin/schinit in place with a
generation counter. The pid stays 1, so every orphan the kernel reparented is
still correctly parented; the children survive; the open files survive, which
is why this is the first line of exit() and not the last; and the job table
survives in /proc/schinit, which is what it is there for.
The escalation ladder: generations 2-3 carry on, generation 4 ignores
/etc/schinit and starts a console shell only -- a manifest that kills init
must not kill it for ever -- and generation 5 stops, with PSOGM rendering
INIT_RESURRECTION_FAILED, a call chain and a QR code.
Three bugs found by running it, each a lesson about exec():
- exec() returns 0 on success; it does not fail to return. The new image
starts when the kernel returns to userspace through the rewritten
trapframe. Returning from exit() is the success path.
- exit() was reached because `killed` was set, and the trap path checks it
again on the way out. Leaving it set exits the new image before its first
instruction: five generations in a few microseconds, a resurrection loop
that looks exactly like the crash it was meant to survive.
- exec() keeps open files, which is what gives the new image a console --
and also keeps the previous life's socket, holding the old supervisor's
channel up so it never sees a hangup.
Verified: `sysrq i`, and the shell being typed into keeps its pid across the
death and rebirth of PID 1, with /proc/schinit byte-identical either side.
schfs-check passes.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
docs/plan/init.md milestones I0 and I1.
**The split.** schinit is PID 1 and does four things: bootstrap the
filesystems, fork schinitd, fork jobs on schinitd's behalf, reap children. No
manifest parsing, no filesystem walking after boot, no allocation. It is the
one program on the machine with no supervisor, so it is written to be too
boring to fail.
schinitd is PID 2 and holds everything else -- /etc/schinit, respawn policy,
throttling -- and is restartable, which is where the bugs are allowed to be.
**Jobs are forked by PID 1, not by schinitd.** That is the load-bearing
detail. schinitd decides and asks; PID 1 acts. So every job is a child of PID
1 at all times, and a supervisor that dies orphans nothing, reparents nothing
and restarts nothing.
**/proc/schinit** is the job table, written by PID 1 whenever it changes. The
kernel holds the text and does not interpret it: a procfs generator runs in
the kernel, and walking PID 1's memory to answer a read would put a userspace
program's format into the kernel, where it would rot.
Verified the way the plan asks for -- `kill 3`, then look:
schinitd: adopting os.schos.console-shell
schinitd: adopting os.schos.tty2-shell
/proc/schinit byte-identical across the restart, same pids, same cookies, same
start times, and the shell being typed into never noticed.
The first attempt got I1 wrong in the instructive way: PID 1 published the
table and schinitd never read it back, so the restarted supervisor started
everything a second time -- two console shells on one terminal, which is the
split-brain this plan rejects the two-init scheme for, reached by a different
road. Publishing state is half of it; coming back and believing it is the
other half.
The protocol is include/schinit.h: fixed-size messages, versioned in the
hello, entirely internal. Fixed-size so PID 1 never allocates, and a
supervisor that has gone mad cannot make it run a parser.
Still to come: I2 (heartbeat both ways, so a hang is as detectable as a death)
and I3 (proc_reincarnate, the actual law change).
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
A timer interrupt that arrives while interrupts are disabled is not delayed,
it is lost -- the LAPIC in periodic mode holds one pending interrupt and drops
the rest. So anything the kernel does with interrupts off for longer than
10 ms makes `ticks` run slow, and it never catches up.
That has now presented twice as "uptime is wrong": once as an interrupt storm
taking 83% of the CPU, once as heavy console output on a display that is
expensive to draw to. Both times the tick was the messenger, and both times it
looked like a timekeeping bug.
monotonic() reads the TSC instead. It is a counter in the silicon that
advances whether or not this kernel is paying attention, it is already
measured against the PIT at boot, and reading it is one instruction and no
lock. wallclock(), /proc/uptime, sysinfo, getrusage and times() all go through
it now.
`ticks` keeps the job it is good at -- the scheduler preempts on it, sleep()
counts it -- where losing one means a slightly longer sleep rather than a
wrong answer about what time it is.
Lost ticks are no longer harmful, which makes them invisible, which is worse.
So /proc/interrupts now says
timer: 3383 ticks taken, 3387 expected
and names the gap when it grows, because a machine dropping timer interrupts
is spending too long with interrupts disabled, and that is worth knowing
before it turns into a clock that lies.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
The sweet spot was available the whole time and I ruled it out by writing
"AltGr+anything" when the true statement is "AltGr+*letter*".
A Polish layout wants AltGr with a letter -- ą ć ę ł ń ó ś ź ż. AltGr with a
function key produces nothing on any layout, no window manager binds it, and
no application asks for it. It is the one combination the host, applications
and the keyboard layout all leave alone, and the previous three commits talked
themselves out of it on a technicality that was mine, not the hardware's.
Verified rather than argued, with QEMU sendkey and screendump hashes:
AltGr+F2 -> tty2
AltGr+F1 -> back, pixel-identical to the start
AltGr+a -> reaches zsh, so ą still types
That distinction is only expressible because Right Alt became its own modifier
bit two commits ago. Sharing ALT is harmless for typing -- neither is used to
look up a character -- and made this impossible to say.
Also added, since it was asked for in the same breath: minus and equals step
to the previous and next terminal, on whatever the chord's modifiers are. They
skip terminals that are not in use -- the same `opened` count VT_GETSTATE
reports -- because six terminals with two shells on them means four blank
screens in between, and cycling past those is worse than not having the key.
AltGr+= -> tty2, again -> wraps to tty1 (only two in use)
AltGr+- -> tty2
chvt -c still moves all of it, and `chvt` reports both halves.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
Third guess at a chord, and the third one would have been wrong too: a display
manager that grabs Ctrl+Alt+F<n> matches loosely, so Ctrl+Alt+Shift+F2 never
reaches the guest either. Extra modifiers do not get past a loose grab.
The real problem is that this is not knowable from inside. A key the host
swallowed is indistinguishable from a key nobody pressed, so the guest cannot
probe for a free combination, and there is no combination that is free
everywhere:
the HOST takes Alt+F4, Alt+F2, Ctrl+Alt+F1..F6 (loosely matched),
Ctrl+Alt+digit (QEMU's own console tabs)
APPLICATIONS want Alt+anything (Meta -- zsh binds Meta-digit as
digit-argument), Ctrl+letter, Ctrl+Shift+digit
(Ctrl+Shift+2 types ^@), bare F-keys
the LAYOUT wants AltGr+letter
So it is a setting. VT_SETCHORD/VT_GETCHORD, and in userspace:
chvt -c ctrl+shift+F Ctrl+Shift+F1..F6 (the new default)
chvt -c ctrl+alt+shift+F
chvt -c alt+1
chvt -c off no chord; chvt N is then the only way
chvt report the active terminal and the chord
`off` is a legitimate setting rather than a failure mode. On a machine whose
host grabs everything worth binding, `chvt N` is the honest answer.
The default moved to Ctrl+Shift+F1..F6: display managers bind Ctrl+Alt+F<n>,
not Ctrl+Shift+F<n>, and Ctrl+Shift with a *function* key is not a character
on any layout -- unlike Ctrl+Shift+digit, which is.
Caught while writing it: VTC_RALT was 0x08 while kbd.h's RALT is (1<<7), since
(1<<3) is CAPSLOCK. A chord set from userspace with AltGr would simply never
have matched, and the only symptom would have been a key that does nothing.
A _Static_assert now holds the two headers together.
Verified: Ctrl+Alt+F2 leaves the screen byte-identical, Ctrl+Shift+F2 switches,
Ctrl+Shift+F1 returns to a pixel-identical tty1, and chvt -c round-trips
through alt+1 and off and back. schfs-check passes.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
Alt+1..6 was a worse choice than the Alt+F-keys it replaced, and the reason
is the one that matters: anything the kernel swallows is taken from every
program that will ever run here, permanently.
Alt+<digit> is Meta-digit. zsh's own line editor binds it as digit-argument --
Alt+1 Alt+2 x repeats x twelve times -- and every readline editor does the
same. Yesterday's chord broke that system-wide, which is exactly the failure
mode of picking a chord applications want.
Working the whole space instead of one axis of it:
the HOST takes Alt+F4, Alt+F2, Ctrl+Alt+F1..F6
APPLICATIONS want Alt+anything (Meta), Ctrl+letter, Ctrl+Shift+digit
(Ctrl+Shift+2 types ^@), bare F-keys
the LAYOUT wants AltGr+letter -- Polish diacritics
Ctrl+Alt+Shift+F1..F6 is in none of those sets. Display managers bind the same
chord without Shift, so adding it steps out of the host's way, and no terminal
application binds a triple-modified function key.
Which chords actually reach a guest was measured rather than assumed: a
temporary probe in kbdgetc logged every modified keypress and QEMU's sendkey
fired each candidate at it. Everything reached the guest, Ctrl+Alt+F2
included -- so that conflict is the host's display manager, not QEMU. The same
probe corrected an earlier claim: `sendkey altgr-2` produces no event at all,
so the previous commit's "AltGr+2 did not fire" was never actually tested.
And the escape hatch is real now. VT_SETMODE was accepted and ignored, so a
program could ask to own the chord, be told yes, and still not get it.
VT_PROCESS genuinely stops the kernel switching. Together with the EVIOCGRAB
path -- which kbdgetc already checks before the chord is recognised -- there
are three honest layers: nothing, VT_PROCESS, exclusive grab.
What VT_PROCESS still does not do is the VT_RELDISP handshake; struct vt_mode
carries relsig/acqsig for when Schcom needs to release the GPU before the
screen changes hands.
Verified by screendump hash: Alt+2 reaches zsh as Meta, Ctrl+Alt+F2 and
Ctrl+Shift+F2 leave the screen untouched, Ctrl+Alt+Shift+F2 switches.
schfs-check passes.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
Alt+F1..F6 was wrong in three separate ways, and the third is the one that
matters most here.
Alt+F4 the host window manager closes the QEMU window
Alt+F2 GNOME and KDE open a run dialog
Ctrl+Alt+F1..F6 the host switches its *own* virtual terminals
AltGr+anything Polish diacritics -- a c e l n o s x z with their marks --
and on some setups AltGr is delivered as Ctrl+Alt
Ctrl+Shift+key widely bound inside applications
Scroll Lock not present on many keyboards
Alt+1..6 survives all of it: the number row, so no function key is involved,
and left Alt, so AltGr is untouched.
Right Alt is tracked as its own modifier bit now. It used to share ALT, which
is harmless for typing -- neither is used to look up a character -- and wrong
for chords, because it made AltGr indistinguishable from Alt.
And the test is an equality rather than a mask:
if ((shift & (CTL | SHIFT | ALT | RALT)) == ALT && ...)
`shift & ALT` fired on Ctrl+Alt+1 and AltGr+1 as well, so a chord aimed at the
host -- or an ordinary Polish keystroke -- would have switched terminals
behind your back. That was a real bug in what shipped yesterday.
Verified with QEMU's sendkey, comparing screendump hashes:
start (tty1) 86cee2284878
Ctrl+Alt+2 86cee2284878 identical -- did not fire
AltGr+2 86cee2284878 identical -- did not fire
Alt+2 114562804a78 switched
Alt+1 86cee2284878 back, pixel-identical to the start
schfs-check passes.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
Three symptoms, two causes, both in the terminal layer.
**The grid and TIOCGWINSZ disagreed about the size.** vt_init() sizes every
grid from gfx_cols()/gfx_rows(), and it runs at main.c:75 -- while the display
driver attaches its scanout from the PCI bus walk at main.c:99. So the grids
were the 80x24 fallback, and TIOCGWINSZ, which computed its own answer live,
said 160x50.
That one disagreement produced all three things visible in the report: text
stopped two thirds of the way down the screen, `ls` laid out 160 columns into
an 80-column grid so words split across the wrap ("vers" / "ion"), and zsh's
cursor tracking was wrong on every line.
Fixed in two places, because either alone would leave the trap set:
gfx_attach_scanout() now calls vt_resize(), so the grids learn the real screen
the moment there is one; and TIOCGWINSZ reports v->cols/v->rows rather than
measuring again. One number, from the terminal being asked.
**Wrapping was eager.** A glyph printed in the last column moved the cursor to
the next line immediately. Real terminals set a pending-wrap flag and wrap when
the *next* glyph arrives -- which is why a line exactly as wide as the screen
takes one row and not two, and why a program can write the bottom-right cell
without scrolling.
zsh's partial-line marker depends on that rule directly: it writes exactly
COLUMNS-1 spaces and expects to still be on the line. With eager wrap it left a
reverse-video '#' above every prompt, which is the hashtag in the report. Both
are gone.
Verified by screendump: `ls /proc` now fits one line with "version" intact, no
markers, and output fills the screen to the last row. schfs-check passes.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
/dev/tty1..tty6, Alt+F1..F6 or `chvt N`. Each is an independent terminal --
its own screen, cursor, line discipline, foreground process group and
session. Two shells on two terminals are unrelated processes that share a
keyboard and a screen.
What made it possible is a backing store. The console drew straight into the
framebuffer and scrolled by moving pixels, which works exactly as long as
nothing else wants the screen: there was nowhere to put terminal 2 while
terminal 1 was visible, and nothing to repaint terminal 1 from. Each terminal
owns a grid of cells now and the framebuffer shows whichever grid is active,
so vt_switch() is set-index, repaint, return -- nothing is saved on the way
out because the outgoing grid has been current all along.
The rule the layer is built on: a terminal does not know whether it is
visible. Writes land in the grid, the cursor moves, the grid scrolls, all
unconditionally; drawing is one question asked at the bottom of vt.c. Every
alternative spreads "am I on screen" through the escape parser, and then a
terminal switched away from mid-sequence comes back wrong.
Everything that was a single global is per-terminal now: tty_fg_pgrp (deleted),
termios, the input ring, the escape parser, the UTF-8 decoder, and the cursor
-- which also stops being pixels in gfx.c and becomes cells in the terminal.
TIOCSCTTY was `return 0` and now records proc->ctty, which is what /dev/tty
resolves through; /dev/tty was an alias for /dev/console with the same minor,
and they are different questions.
kbd.c names Alt+F1..F6 as KEY_ALT_F1+n, above 0xFF on purpose -- every byte
value is already a character or a UTF-8 continuation, and a chord colliding
with one would be typed into the terminal instead of switching away from it.
What the chord means is decided in console.c, where the terminals are.
VT_ACTIVATE / VT_WAITACTIVE / VT_GETSTATE / VT_OPENQRY, Linux's numbers.
The argument to ACTIVATE is by value, not a pointer -- same shape as TCFLSH.
Two bugs found by screenshotting the result rather than reading the log:
- zsh emits 0x0F to leave alt-charset mode, because that is what the `linux`
terminfo says rmacs is, and the console drew CP437 glyph 0x0F: a sun, in
the prompt, on every line. A terminal never draws a C0 control character.
- SGR 7 was unimplemented, so zsh's partial-line marker rendered as a bare
'#' instead of a reversed block. Reverse video swaps the colours at draw
time, not at store time, so turning it off restores what was underneath.
Verified by screendump: tty1 and tty2 show different screens, and switching
back repaints tty1's earlier content from its grid. schfs-check passes.
docs/architecture/terminals.md, including what it does not do -- no
scrollback, no reflow, no VT_PROCESS mode for a display server to veto a
switch, which Schcom will need before it owns Alt+F7.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
Porting zsh found four defects in the kernel and one in the filesystem. None
of them were about zsh; all of them had been there the whole time.
**exec threw the environment away.** kernel/fs/exec.c wrote a single zero for
envp with the comment "(no environment)" -- not yet, but ever, for every
process. musl has been passing a perfectly good envp to execve (Linux 59, which
SchOS answers) and it was never read. So $TERM was unset, $PATH was whatever
each program invented, and a variable exported in a shell did not survive the
exec meant to carry it. exec() takes an envp, sys_exec reads argument three,
and schinit manifests grew an Environment key. The userland two-argument
exec(path, argv) stays as a wrapper that passes null explicitly -- the raw stub
is execve now, so a short caller cannot leave %rdx holding a stray pointer.
**Which overflowed the kernel stack.** Three pointer arrays, ~1.8 KB, on a
4096-byte stack, under exec -> loaduvm -> readi -> SchFS -> btree. Pid 1 went
down on the first exec. The canary caught it and named the process, which is
what it is for. They live in one kalloc'd page now.
**waitpid returned a bare -1** for both "no children" and "interrupted". musl
reads [-4095,-1] as -errno, so userspace saw EPERM; zsh printed "wait failed:
operation not permitted" and abandoned the rest of /etc/zsh/zshrc. -ECHILD and
-EINTR now. Third time a bare -1 has cost real time -- on a system with no
permission model, EPERM should be read as "a syscall returned -1".
**NVMA was 64.** musl's allocator takes a mapping per size class and does not
release them eagerly, so compinit ran out of slots and got -1, which is EPERM,
which malloc turns into NULL, which zsh reports as "fatal error: out of memory"
-- on a machine with 1000 MB free. 256 slots, and -ENOMEM.
**And a small write to SchFS costs 58 ms.** 200 writes to one open file take
11.75 s; the same to /dev/null take 0.14 s. That is the design stated at the
top of kmount.c -- one transaction group per operation, and a commit is two
flushes and an uberblock -- not a bug. But compdump writes its digest with
thousands of separate prints, so it takes minutes and looks like a hang.
compinit -D, which skips the dump, finishes in 3.6 s. This one is a decision
rather than a fix, so it is written up and left: /etc/zsh/zshrc keeps compinit
switched off and explains exactly why, with the measurements.
zsh runs as /bin/zsh and is the console shell. ZLE works -- arrow keys recall
history through terminfo, $COLUMNS reads the real 160 because TIOCGWINSZ was
fixed, colour and prompt expansion work, startup is 0.14 s. mksh keeps /bin/sh,
so schinit's no-jobs fallback is a different program from the one that could
have failed.
Also fixed while here: AT_CLKTCK was the literal 100 and is USER_HZ; the
extract rule in four ports could move a source tree inside itself when make
re-ran it, which is how zsh's configure came to write modfile=zsh-5.9/... into
config.modules and fail three steps later.
schfs-check passes. Full writeup in docs/history/kernel.md section 14.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
fastfetch reported 3 seconds of uptime after 15 real ones, `time` claimed
3.12s for a 20-second command, and ping's one-second interval was closer to
six. One cause, not three: `ticks` was advancing at ~14.5 Hz instead of 100.
It was not the timer. virtio-gpu-pci shares IRQ 11 with the e1000 and
registered no interrupt handler at all. PCI INTx is level-triggered, so the
device held the line asserted, the kernel signalled EOI, and it fired again
immediately -- 72,000 interrupts a second, 83% of CPU0, starving the tick.
The e1000's handler read ICR as zero half a million times, correctly
answering "not mine", which is what made it look like a network-free problem.
Reading virtio's ISR byte clears it and drops the line. That is the fix.
irq_dispatch now measures each line's rate against the TSC -- not ticks,
which is the first casualty of a storm -- and names the attached handlers
once above 20000/s. It reproduces the original fault as a single line.
Found along the way, and fixed because they were real:
- lapicinit loaded TICR with xv6's uncalibrated 10000000, whose own comment
admitted it was never measured. calibrate_timers() now measures the LAPIC
and the TSC against the 8254 PIT. (It came back at 999.9 MHz, so this was
not the bug -- but the next machine is not obliged to be 1 GHz.)
- microdelay() was an empty function body. Every microdelay(1000) in the
xHCI reset path and the USB hub power-on sequence returned instantly.
- HZ is a constant in param.h now, not the literal 100 written out at a
dozen ticks-to-seconds conversions. USER_HZ is separate and deliberately
does not follow it: musl answers sysconf(_SC_CLK_TCK) with 100 always, so
times() and /proc/stat scale at the syscall boundary the way Linux does.
- TIOCGWINSZ answered a hardcoded 24x80. gfx_cols()/gfx_rows() have known
the real geometry all along; at 1024x768 the console is 128x48, so every
program that lays out a screen was formatting for the wrong one.
- TCFLSH fell through to the catch-all `return 0`, claiming to discard
input without discarding any.
uptime now advances 5.01 seconds per 5 real seconds. schfs-check passes.
Full investigation, including the two wrong hypotheses and the measurement
flaw that protected them, in docs/history/kernel.md section 13.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
A fresh boot spent 2600 ms in schinit's bootstrap() before printing its first
line, and the whole of it was device nodes. Measured rather than guessed: an
instrumented bootstrap reported 2600 ms on a fresh image and 0 ms on the next
boot, because the nodes already existed by then and mknod failed fast. The cost
was therefore invisible to anyone not building images -- which is to say,
invisible to everyone except whoever works on this.
Each node was a real inode on the root filesystem, and SchFS commits every
mutation as its own transaction group: a copy-on-write rewrite of the path to
the root, then an uberblock behind a cache flush. Twenty of those is seconds.
The duplication mattered more than the time. Major and minor numbers are the
kernel's -- devsw[] in kernel/dev/cdev.c is what they dispatch on -- and schinit
held a second copy of that mapping with nothing checking the two agreed. Adding
a character device meant editing a userspace program, and a wrong minor produced
a node that opened successfully and did nothing.
kernel/fs/devfs.c puts the table next to the thing it describes and serves /dev
behind the same fs_ops vtable procfs uses. The namespace is all it provides:
fileread and filewrite see T_DEV and go through devsw themselves, so a read
never reaches it. It is read-only, and dirlink and ialloc answer EROFS rather
than pretending -- a node exists because the kernel has the device, or it does
not exist.
This also makes /proc/filesystems true. It has advertised "nodev devtmpfs" for
as long as it has existed, against a kernel that had no such filesystem.
Two things found while measuring what remained:
* The thirteen mknod() calls for /proc ran *after* the procfs mount, so every
one was resolving a path into a read-only synthetic filesystem and failing.
procfs.c's own note called them a fallback to be cleaned up later; they have
been dead for longer than that.
* The four mkdir() calls for mount points were the rest of it. They are made
by the image builder now (sysroot-install), and the ones in init are a
fallback -- a mkdir that fails because the directory exists costs a path
walk and no writes, where creating one costs a transaction.
bootstrap() is now 0 ms on a fresh image, down from 2600.
init opens /dev/console after the mount rather than mknod'ing a node called
"console" into the root directory, which is what used to leave a stray /console
in `ls /`. The old path is kept as a fallback: if the mount fails there is no
/dev/console and no way to report it.
socktest passes, /proc still lists per-process directories, ps -e still
enumerates.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
"net: eth0 10.0.2.15/255.255.255.0 via 10.0.2.2" arrived on the display seconds
after the prompt did, so the first thing anyone saw after booting was the kernel
typing into their session. It lands wherever the cursor happens to be -- in the
middle of a command, a fastfetch, or later a compositor's output.
console_takeover() hands the framebuffer to userspace on the first exec, and
after that the screen belongs to whoever is using the machine. cons_serial_only
already encodes that rule for the boot path -- "the display is the product; the
serial line is the diagnostics" -- and it starts set for exactly this reason.
What was missing is the rule for messages that cannot be timed to happen before
userspace starts, because they are waiting on a network.
klog() is that: the serial port only, whatever cons_serial_only currently says.
A macro rather than a function, because cprintf's varargs cannot be forwarded --
there is no vcprintf, and adding one to duplicate the formatter would cost more
than it buys. cprintf holds cons.lock for the whole message, so the flag cannot
be observed half-changed by another CPU printing at the same time.
Nothing is lost. The line still goes out the UART, which is where bootcheck
reads it from and where you want it when something is wrong, and a person who
wants to know their address reads /proc/netif.
Verified by screendumping the framebuffer after a boot and two fastfetch runs:
no interruption on screen, still one occurrence on serial.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
fastfetch reported "getifaddrs(&ifAddrStruct) failed", and the right answer to
that is not a patch to fastfetch. Asking what interfaces a machine has is
something every networked program does, and on Linux there is exactly one way:
a NETLINK_ROUTE socket. musl implements getifaddrs() over it, so a libc call
that looks like a syscall is really a small protocol conversation -- and without
this side of it every caller fails the same way.
kernel/net/netlink.c is a responder rather than a general socket family. No
event stream, no multicast groups, nothing subscribes to link state: a request
arrives, the whole answer is generated into the socket's buffer, and reads drain
it. That shape is chosen, not a shortcut -- getifaddrs sends a dump and reads
until NLMSG_DONE with MSG_DONTWAIT, so generating at send time means a read
never blocks and the non-blocking flag is honoured for free.
RTM_GETLINK returns one RTM_NEWLINK per interface with IFLA_IFNAME and
IFLA_ADDRESS; RTM_GETADDR returns one RTM_NEWADDR per address with IFA_ADDRESS,
IFA_LOCAL and IFA_LABEL. Anything else gets a bare NLMSG_DONE, because the
caller's next move is a read either way and an unexpected error message is worse
than nothing to read.
The data is not netlink's. Interfaces come from the netdev class and addresses
from lwIP's netifs, exactly as /proc/netif reads them; nothing in the file knows
what an e1000 is. Interface indices matter more than they look -- getifaddrs
matches address records to link records by index, so the two dumps have to
agree. Loopback is 1, netdevs follow from 2.
Verified by reading it back through musl rather than by inspecting the bytes:
lo flags=0x49 fam=-1
eth0 flags=0x1043 fam=17 (AF_PACKET, the MAC)
lo flags=0x49 fam=2 addr=127.0.0.1 mask=255.0.0.0
eth0 flags=0x1043 fam=2 addr=10.0.2.15 mask=255.255.255.0
Correct flags, addresses and netmasks, and the link records pair with the
address records.
One thing this does not close: fastfetch's Local IP module defaults to
default-route-only, which calls ffNetifGetDefaultRouteIfName() and reads
/proc/net/route. SchOS has no routing table and no such file, so the module
still shows nothing -- now for a reason that is about routing rather than about
enumeration. That is the next file to provide.
socktest still passes; the AF_UNIX and AF_INET paths are untouched.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
Shift is the logo, Guru is the icon, SchCotie is the mascot, and the difference
between them is not stylistic. An icon works once, at one size, in one state. A
mascot has to be able to *do* things -- confused, pleased, asleep, seen from
behind -- which is why it costs more and is drawn last. Written down because
this is exactly the kind of decision that gets re-derived badly.
Assets/ becomes docs/branding/, which also fixes two things:
* Filenames had spaces in them, so every reference needed quoting. They are
lowercase and hyphenated now.
* Assets/ held hand-written SVG and generated PNG in one directory, which is
the one rule docs/guides/layout.md states.
The guide records what was previously only in people's heads:
* The constraint everything answers to -- include/logo.h is a 128x128
single-channel coverage mask, so a mark has to survive as a solid shape in
one colour and be recognisable at 24 px. Colour is the last decision, never
the first.
* Guru's purple is load-bearing. PSOGM_BG #4B1D9A is deliberately not the
accent #7C5CFF, and theme.h's rule is that the full-bleed purple belongs
entirely to failure. If Guru appears in accent violet on a sticker, the deep
purple stops meaning "the machine stopped" -- and it is the only colour in
the system that means anything.
* Why shift-mask.svg is a separate file from shift.svg. mklogo.py reads ink as
(255 - luminance) * alpha, so tones survive as coverage and the relationship
inverts: darkest here is brightest on screen. Copying the violet fills would
light the object from below.
* Why a raster is committed beside its SVG. inkscape is not a build
prerequisite, so requiring a vector rasteriser would break the build on most
hosts for a file that changes about once a year.
* That the BSD daemon is copyrighted and SchCotie must be drawn from the
concept rather than from Beastie. A daemon is a Greek daimon and a Unix
background process; neither is ownable. The pitchfork, the sneakers and that
face are.
Format rule: SVG, or PNG at high resolution with no lossy compression, nothing
else. SVG is the source of truth for every mark and a raster is generated from
it, never edited.
include/logo.h regenerates byte-identical from the new path -- only the source
comment differs. The two old wordmark SVGs were deleted outside this session and
are left deleted; their PNGs are kept as superseded-*.png for reference.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
A leaf's value area holds variable-length records end to end, so a record begins
wherever the one before it finished -- a multiple of 8 only by luck. The code
reads them by casting a pointer into the block, which tells the compiler they
are aligned when they are not.
That works on x86, which is why it has never shown as a bug, and it is undefined
in C. A compiler entitled to assume 8-byte alignment may emit an instruction
that requires it, and the failure would arrive at a different optimisation level
or on a different architecture looking like filesystem corruption rather than
like codegen. UndefinedBehaviorSanitizer reported 31 of them across obj.c,
objw.c, attr.c and space.c.
Seven types now carry SCHFS_PACKED: dirent_rec, extent, extref, index_rec,
attr_rec, snap and inode. Two of those -- attr_rec and inode -- were never
reported, because schfs_objtest does not reach an unaligned instance of either.
They were latent rather than absent, and casting into a leaf is the same lie
whether or not a test has caught it yet.
NO ON-DISK LAYOUT CHANGED, and that was measured before the annotation was
applied rather than assumed: every one of these already carries explicit `pad`
fields and therefore has no implicit padding, so sizes and member offsets are
identical with and without it. Seven _Static_asserts now pin the sizes, because
a record that silently grows misreads every existing volume from that point on
and looks like corruption rather than like a struct that changed.
schfs_node_hdr, schfs_item, schfs_kptr and schfs_uberblock are deliberately left
alone. The node header is exactly 32 bytes, so the items and key pointers after
it are already 8-aligned, and annotating them would cost byte-wise access on the
tree's hottest paths to fix a problem they do not have.
Verified by reading an image written before the change: schfs_fsck --check
passes on structure, checksums, key ordering, data blocks and derived free
space, and the kernel mounts its own root unchanged.
Also here: mkschfs passed a null pointer to qsort for an empty directory. Every
implementation ignores it when the count is zero; the standard does not say they
have to.
make schfs-check now runs with no sanitizer output at all.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
The network thread waited on the timer and nothing else, so a frame that landed
just after a poll sat in the ring for up to 10 ms -- the largest fixed latency in
the system. That was deliberate when the stack was new (a receive path woken by
an interrupt is much harder to debug than one that is asked) and there is now a
working stack to compare against, which is exactly when it becomes safe to do.
The frame is still *pulled* by the thread through .receive. It has to be: a call
into lwIP can reach the driver, the driver can sleep on the buffer cache, and an
interrupt handler may do neither. netdev_rx_signal() delivers nothing and reads
no ring -- it only ends the wait early.
The wait stays on the timer tick rather than moving to a channel of its own, and
that is the design. lwIP's timers must still run when the link is silent, so the
clock has to remain a floor under the wait, and sleep() cannot wait on two things
at once. Joining the tick's existing wakeup costs nothing and leaves one wait
with one predicate.
rx_signal lives under tickslock for the same reason it gets no lock of its own:
sleep() releases tickslock atomically, so a driver setting the flag and waking
under that lock cannot lose the wakeup by arriving between the predicate test and
the sleep. A private lock would not be held across the sleep, and that window
would be back. Calling it from interrupt context is safe on the same terms the
timer handler already relies on -- acquire() disables interrupts, so a CPU
holding tickslock cannot be interrupted into it.
The e1000 signals on RXT0 (a frame or burst has landed) and RXDMT0 (the ring has
fallen below its threshold, which is when waiting for the clock hurts most).
Measured rather than asserted: /proc/netif reports wake_rx and wake_tick, and
across three pings rx_frames and wake_rx both went 2 -> 6 -- every received frame
woke the thread through the interrupt. wake_tick keeps advancing as the idle
floor, which is what lwIP's timers need. socktest still passes and ping is still
0% loss.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
The old logo was a thin monoline wordmark in a ring, with a grey gradient. The
kernel draws include/logo.h as a 128x128 four-bit coverage mask, so hairlines at
that size go ragged, the ring spent most of the area on nothing, and at small
sizes the S read as a 5.
Three marks, because each depicts the thing it belongs to and so they stop
competing:
Shift three slabs, displaced. The layer diagram in
docs/architecture/README.md as a solid object, and the kernel's
name as a shape. Dimensional rather than flat -- the tradition
Paul Rand's NeXT cube belongs to, which is what the architecture
already claims as its model.
Uberblock the SchFS ring: twelve slots, one of them newest. Twelve so the
slots divide 360 exactly; an arbitrary count leaves a part slot
crammed against its neighbour where the ring closes. The newest
is marked by REACHING FURTHER OUT as well as by colour, because
colour is exactly what the coverage mask discards.
Guru the PSOGM face, drawn for reference. Not wired in: panic.c argues
for the ":(" on the grounds that it is "legible as a mood without
being a picture of anything", and Guru spends that. Worth deciding
on purpose.
mklogo.py reads ink as (255 - luminance) * alpha, so tones survive as *coverage*
rather than being flattened -- which means the old grey gradient was not
discarded, it was turned into partial transparency. It also inverts: darkest is
most ink, which the kernel draws brightest. The mask source is therefore a
separate greyscale file with the top slab darkest, or the object would be lit
from below. Two files sharing one geometry is a drift hazard and the desc in
both says so.
The splash gets it for free: ui_logo() shades per row with metal() and the
mask's own tones reinforce that gradient rather than fighting it, so the slabs
come out looking machined.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
setsockopt was `return 0;` and getsockopt wrote a zero for whatever was asked.
The comment said it was so a server setting SO_REUSEADDR and treating failure as
fatal would not fail. That trade is the wrong way round: a caller told "no"
adapts, a caller told "yes" changes its behaviour on the strength of it -- sets
SO_RCVTIMEO and then blocks forever in a read it believes will time out, sets
TCP_NODELAY and reports latency numbers that measure Nagle. The failure is
silent, remote from the call, and indistinguishable from a bug in the stack.
Six options are honoured now and everything else is ENOPROTOOPT. SO_REUSEADDR,
SO_KEEPALIVE, SO_BROADCAST and TCP_NODELAY reach lwIP's pcb; SO_SNDBUF/SO_RCVBUF
are accepted and report the real size on read. SO_REUSEADDR is implemented rather
than refused precisely because of the original comment's concern -- it is the one
option a server really does set and really does check.
accept4 becomes its own table entry rather than an alias for accept. It has to
be: accept takes three arguments, so a shared handler reading a fourth register
reads whatever the caller last left there and invents a SOCK_NONBLOCK out of
stack litter. SOCK_CLOEXEC and SOCK_NONBLOCK are now honoured by socket(),
socketpair() and accept4() -- a dropped CLOEXEC leaks a listening socket across
every exec, and a dropped NONBLOCK has the caller block forever in what it
believes is a poll loop.
ICMP ping sockets: socket(AF_INET, SOCK_DGRAM, IPPROTO_ICMP), on lwIP's raw API,
with the callback stripping the variable-length IP header. Only echo replies are
eaten, and that restriction is load-bearing -- raw_input runs before lwIP's own
icmp_input, so consuming every ICMP message would swallow the echo *requests* and
this machine would silently stop answering pings from outside.
Loopback, so 127.0.0.1 is a real destination. LWIP_NETIF_LOOPBACK is the half
that is easy to miss: without it a packet addressed to an interface's own address
goes out of the wire and never comes back. With NO_SYS=1 nothing drains the
loopback queue, so the network thread calls netif_poll_all().
cmd/ping is first-party rather than toybox's, and the reason is a limitation not
a preference: toybox reads replies with recvmsg(2) and wants the TTL through an
IP_RECVTTL control message, and SchOS has neither. recvmsg is item 1 of
docs/plan/foundations.md, where it is wanted for SCM_RIGHTS anyway, and should
not be rushed to make one applet run.
Two bugs found on the way. Every AF_INET socket failed with ENFILE until DHCP
bound, because inet_new() demanded an *address* before handing out a socket and
reported it as "file table overflow"; net_is_ready() (lwIP up) is now distinct
from net_is_up() (DHCP bound). And that bug was hiding a bad test: socktest
asserted "AF_INET is refused, not faked", true when written, false since AF_INET
landed, and green the whole time because the suite runs before DHCP binds and the
ENFILE made the refusal look real. Fixing the kernel is what turned it red.
Verified: ping to 127.0.0.1, localhost, the gateway, 1.1.1.1 and 9.9.9.9, all
0% loss; socktest ALL CHECKS PASSED with ten new checks, including a setsockopt
round trip and an unknown option that must be refused -- a permissive stub cannot
pass both at once.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
Everything created after boot was stamped 1970. know() in SchFS and now_sec() in
the v1 filesystem both returned uptime(), under comments saying the kernel had no
wall clock -- which stopped being true when time_init() started reading the RTC.
The comments outlived the condition they described.
Two errors, and the second is the less obvious: uptime() counts 10 ms *ticks* and
they were being used as *seconds*, so a node created 0.6 s into boot read back as
00:01:00 and the clocks ran a hundred times too fast. Both now go through
wallclock(), which exists precisely so nothing can disagree with gettimeofday
about the time. Confirmed against the RTC line from the same boot.
The occupancy counters are the instrumentation docs/plan/roadmap.md asked for
while hunting the ls -lR hang: inode, file and pinned-buffer counts with a
high-water mark each, maintained at the 0->1 and 1->0 transitions rather than by
scanning, so they cost one increment on a path that already holds the table's
lock.
The peaks are the diagnostic. A live count only describes the moment you looked,
and a reference leak is invisible in it right up until the table is full -- at
which point iget() panics and filealloc() returns 0, which is where the trail
used to go cold.
What they measured: across ten iterations of ls -lR with writes interleaved, 33
execs and 13378 tree descents, inode_peak stayed at 9/200, file_peak at 9/512 and
buf_pinned_peak at 1/256, with the SchFS bracket unlocked throughout. That
eliminates the roadmap's own hypothesis -- "a lock or reference-count leak
somewhere in the *at() paths" -- and the hang did not reproduce. It is very
likely fixed in passing by the *at() dirfd rewrite, the getdents cookie rewrite
or SchFS becoming the root, all of which landed after it was reported.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
uber_valid() answered one question -- valid or not -- and a version mismatch
returned the same as a failed checksum. Those two states need opposite handling:
A damaged uberblock means STEP OVER IT. That is what the 128-slot ring is for,
and the older root is a consistent filesystem. This is crash recovery.
A sound uberblock from a newer format means STOP. Falling back mounts a STALE
filesystem: every commit made since the volume was upgraded is discarded, the
mount reports success, and nobody finds out.
Demonstrated rather than asserted. A forged image with two uberblocks --
(version 1, generation 1, root 129) and (version 2, generation 2, root 130) --
made the old code mount root 129 and exit 0, silently rolling the filesystem
back to before the upgrade.
uber_check() returns BAD / OK / RO / TOO_NEW, and asks its questions in order:
is it ours (magic), is it intact (crc), can we read it (version, features), does
it make sense (root in range). Compatibility is asked before sanity on purpose --
the fields a sanity check reads are only meaningfully placed if the format is one
we know.
Plus feature flags in three classes, which is what stops a version number being a
blunt instrument: compat (ignore unknown bits), ro_compat (mount read-only),
incompat (refuse). The test for which class is one question -- what happens if a
driver that does not know this feature writes to the volume? Nothing -> compat.
Something goes stale -> ro_compat. Something is unreadable -> incompat.
They occupy what was reserved0, so every later field keeps its offset and
existing images -- which have zeros there -- read unchanged. schfs_vol_commit
memsets the uberblock before filling it in, which is why v1 volumes already say
"no features" without having been written by code that knew the field existed.
That trick is available exactly once. Four _Static_asserts now pin the layout.
Verified on forged images: version bump refuses, incompat refuses, blocksize
mismatch refuses, ro_compat mounts the NEWEST root read-only rather than falling
back.
Also here, because they are the same file: SchFS's transaction owner is a pid
alongside the pointer, since a bare struct proc * does not identify a process --
ptable slots are recycled, and an exited owner's pointer compares equal to
whichever unrelated process lands in that slot next, at which point klock() skips
the lock entirely and two callers share one volume's scratch pages. And
schfs_lockinfo(), so /proc/fsstat can report a wedged volume instead of leaving
every path asleep in acquiresleep with no way to ask why.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
docs/ was ten files where a plan, a description, a guide and a bug report all
sat next to each other. It is now architecture/ (what SchOS is), plan/ (what it
will be), guides/ (how to work on it) and history/ (why it is like this), with
an index that states the one rule holding it together: when something is
finished it moves from plan/ to architecture/ and is rewritten as a description
of the thing rather than a record of building it.
A ticked checkbox says the work happened. It does not say what the work
*produced*, and six months later the checkbox is all that is left. The network
stack proved the point: finished, verified end to end, and for a month the only
description of it anywhere was a checkbox four levels of indentation deep.
So roadmap.md went from 534 lines to 181 -- nothing deleted, the done work
graduated. Seven subsystems that were finished and undescribed now have
descriptions: boot, kernel, syscalls, filesystem, networking, graphics, usb,
init. Every one ends with "What it does not do", which is the section that pays;
a document listing only strengths is marketing, and the reader who needs it most
is the one about to trip over a limitation nobody wrote down.
schfs.md still opened with "phases 1 through 3 1/2 are done, phase 4 is next"
long after all five had landed. It is a description now, with the phase record
and the hard-won bugs moved to history/.
Three corrections found on the way, all load-bearing:
* projects/schcom/README.md claimed SCM_RIGHTS was done. There is no
sendmsg/recvmsg in the kernel at all, and an entire GUI plan rested on it.
* The root README said "No TCP/IP" months after the stack landed.
* history/kernel.md sec.7 was marked "in progress" on a problem that process
groups solved.
Also adds plan/1.0.md, which draws a finish line so the open list stops being
infinite, and marks schwm/ and schde/ superseded: in Wayland the compositor is
the window manager, so that split dissolves rather than being abandoned.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
Same disease as the network stack, one subsystem over. kernel/dev/vga.c
was three things wearing one name: the owner of the framebuffer, a
bitmap drawing library, and the text console. Only the first is a
driver, and because they shared a file they shared a prefix -- so
panic.c drew the crash screen with vga_fill_rect, console.c printed with
vga_putc, /dev/fb0 handed out vga_fb_phys, and every one of those files
named a display driver it had no business knowing about.
Split three ways:
- kernel/drivers/gpu/vga/vga_fb.c owns the VBE framebuffer -- the mode
the loader programmed and handed over, the aperture, and the display
class vtable. It is a driver and now lives with the driver that
finds the card on the bus.
- kernel/dev/gfx.c draws: pixels, shapes, glyphs, the cursor, scroll,
the ANSI erase operations. It names no driver. The framebuffer
arrives through gfx_attach_scanout(), so a driver publishes what it
owns and this file draws on it without knowing what it is.
- include/gfx.h is the interface, and says why it is deliberately not
the display class: that one is about frames -- acquire, present,
vblank -- and the crash screen has to work when nothing else does.
main() no longer calls a display driver. platform_register_devices()
takes the framebuffer address and brings up whatever owns it, which is
where driver names belong; kernel/drivers/registry.c is the file whose
entire job is naming every driver in the kernel.
virtio-gpu got the e1000 treatment as well: instances in an array, the
instance hung on dev->priv by an `attached` hook, and every op reading
it rather than a file-static. It refused the second adapter outright
before, which is not hypothetical -- integrated plus discrete is an
ordinary laptop. The VGA driver deliberately has no such hook, and says
so: there is one framebuffer the loader programmed however many
VGA-class controllers the bus turns up, and a second one publishing the
same scanout is the truth rather than a bug.
Private headers moved next to their drivers: include/virtio.h and
include/xhci.h each had exactly one includer. include/ is for headers
the rest of the kernel needs.
Deleted rather than moved: consoleinit() programmed the sixteen ANSI
colours into the VGA DAC and disabled text-mode blink, by writing ports
0x3C0-0x3C9 from the console. The comment three lines above the colour
table already said this does nothing in a 32bpp linear mode -- there is
no palette to index -- so it was both dead and in the wrong file.
Verified, and this is a refactor so the evidence is that nothing moved:
- bootcheck -Fb -Net: 59 [ok], no failures. That covers the console
through every command, /dev/fb0's geometry, writing pixels through
write() and mapping the framebuffer through mmap.
- make crashshot: the crash screen renders exactly as before --
shapes, glyphs, logo and QR code -- which is the whole of gfx.c
exercised at once, in a panic, with the scanout attached by a driver.
- make schcom-test: five pixel assertions from the host oracle still
pass, so the display class and the compositor above it are unmoved.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The TCP/IP stack reached straight into the first driver anyone wrote.
net.c called e1000_present(), the lwIP glue was named netif_e1000.c and
called e1000_transmit(), and include/defs.h declared the driver's
internals to the whole kernel. The second NIC driver would therefore
have arrived with a near-identical copy of the glue file and a patch to
the protocol stack, which is the arrangement include/device.h was
written to prevent -- the e1000 even said so, with `.ops = 0, // no
class vtable yet; there is no net class`.
There is now. A driver publishes a struct netdev_ops -- transmit,
receive, mac, link_up -- and is finished: it never mentions lwIP, and
nothing above it learns its name. The display class was already this
shape; this is the same shape for networking, and the two now read the
same way in kernel/dev/device.c.
What that made possible immediately:
- One netif per bound device instead of one per driver. lwIP's glue is
netif_eth.c, it takes a struct device, and it holds up to NETIF_MAX
interfaces. net.c attaches every network device it finds and starts
DHCP on each, because which interface the network answers on is not
ours to decide.
- /proc/netif lists all of them, one block of fields each, with the
interface name and the driver that claimed it.
- The ARP self-test moved from the e1000 to the class. Nothing in it
was chip-specific, and it is the first thing anyone will want to run
against the next driver. Still `echo n > /proc/sysrq`.
- Private headers live with their driver: e1000.h was in include/,
where exactly one file could use it. The driver build now adds the
driver's own directory to the include path.
The e1000 also drives more than one card. Its probe() refused the second
outright, because there was a single static instance -- so a machine
with two of them drove one and silently ignored the other. Instances are
an array now, and the instance is hung on dev->priv by a new optional
`attached` hook that the bus calls once the device exists. probe() runs
before device_add(), so there was previously nowhere to put it; that is
the whole reason the hook exists rather than a line inside probe.
QEMU gets an explicit NIC on every target (QEMU_NET). The default device
happens to be an e1000 on this machine type today, which is exactly the
sort of thing that stays true until a QEMU release changes it and the
network stops being exercised without anyone noticing.
Verified:
- bootcheck -Net passes, 48 [ok], no failures -- including the moved
self-test and two new checks that /proc/netif names both the
interface and the driver behind it, which together say the stack
went through the class rather than through a driver it was compiled
to know.
- Two NICs attached by hand (-device e1000 twice, second with its own
MAC): eth0 and eth1 both attach, each reports its own MAC in
/proc/netif, and each gets its own DHCP lease. Before this commit
the same machine produced one interface.
- make schfs-check still passes in full.
One trap worth recording: moving a header broke the build for anyone
with stale .d files -- make stopped with "No rule to make target
'include/e1000.h'". CFLAGS gains -MP, which emits a phony target per
header so a moved or deleted one can never wedge the build again.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The second of the two things the stack owed. recvfrom accepted the
address arguments and ignored them, so every caller was told nothing
about where a datagram came from -- which for UDP is most of the point:
a server that cannot see the sender cannot reply to it.
Fixing it properly meant fixing something worse that was sitting
underneath. UDP payloads went into the same byte ring TCP uses, with no
record of where one datagram ended and the next began, so two datagrams
that arrived together came back from one read as one. That is a bug
whose appearance depends on timing and whose blame lands on the sender.
Datagrams now carry a header alongside the bytes: source address, source
port, length. Sixteen of them per socket, in a ring beside the existing
one. Whole buffers would have been 16 sockets x 16 datagrams x an MTU of
static memory for something almost always empty; only the boundaries need
keeping, and those are eight bytes each. A datagram too big for the
buffer it is read into is truncated and the remainder discarded, rather
than left in the ring to be delivered as the head of the next one. One
that arrives with the queue full is dropped whole -- that is UDP's own
contract, and half a datagram in the ring would corrupt every one after.
An empty queue and a zero-length datagram are kept apart, because a
sender really can send zero bytes and the right answer to that is 0
rather than "nothing has arrived yet".
sendto grew the mirror image: a destination address on a datagram
socket, which is what a server replying to whoever just wrote to it
needs. On TCP a destination is ignored rather than refused -- the
connection decides where the bytes go, and programs pass one out of
habit. For a stream, recvfrom's source is read from the pcb: it is the
peer, and it is the same answer for every byte.
Verified: bootcheck -Net now covers UDP too, 45 [ok] and no failures.
The address check is real rather than cosmetic -- nettest deliberately
leaves its sockaddr uninitialised, so a kernel that filled in nothing
would print stack rubbish instead of 10.0.2.2 and the port the responder
is actually on. The boundaries are checked by byte count: the responder
sends 20 bytes and then 16, and the guest reports reading 20 and then
16. Treated as a stream the first read would have taken all 36.
Left for later, and deliberately: accept() still does not report the
peer's address, and getsockname/getpeername do not exist.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The AF_INET blocking paths -- connect, accept, read and write -- woke on
every timer tick and looked to see whether anything had changed. That is
a wakeup per blocked socket per 10 ms whether or not the network did
anything, and it answers late even when the data was already sitting in
the buffer.
What stood in the way was a lock mismatch, and it is written down in the
file now because it is the reason this is not simply sleep on the socket:
sleep() wants a spinlock, and the lwIP lock is a sleeplock -- it must be,
since a call into lwIP can reach the driver and the driver can sleep.
Nor can a waiter hold the lwIP lock while sleeping, or the network thread
could never deliver the thing it is waiting for.
So the wait is on a counter with a spinlock of its own. A waiter reads
the counter while it still holds the lwIP lock, right after checking its
condition, then releases and sleeps only if the counter has not moved.
The lwIP callbacks bump it and wake the channel. No wakeup can be lost:
anything that changes a condition does so after the reader's snapshot,
and therefore bumps the counter after it too, and the sleeper re-checks
under the spinlock before committing.
One counter and one channel for all sixteen sockets rather than one
each. Waking a socket with nothing to do costs a few instructions, and a
per-socket channel would have to be torn down in close() against
callbacks that can still arrive -- which is a use-after-free waiting to
be written.
tcp_sent is registered now, on both connected and accepted pcbs. It was
not needed while writes woke on the clock; with this change a write
blocked on a full send buffer would otherwise sleep with nobody left to
wake it, because an acknowledgement is the only thing that frees space.
Verified: bootcheck -Net passes, 38 [ok], no failures. It exercises the
path that matters -- nettest blocks in read() until the host answers, and
the host answers only after reading what the guest sent, so the wakeup
comes from a callback and not from a tick that would have fired anyway.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The TCP/IP stack landed with its end-to-end test performed by hand: start
a listener, boot, type nettest, read the output. That is evidence once
and nothing afterwards, and it is exactly the kind of check that quietly
stops being run.
bootcheck -Net does the whole exchange. A listener runs on this host on a
port picked by binding one and letting go, the guest is told to connect
out to 10.0.2.2 at that port, and both ends are asserted: the guest
reports what it received, the listener reports what arrived. One side
alone proves much less than it looks -- a stack that completes a
handshake and then drops every byte satisfies any check made from inside
the guest.
The listener is a background job, a separate process, because this
harness has one hard rule: nothing may block the writer feeding QEMU's
stdin. A socket accept is precisely the wait that deadlocked two earlier
attempts at reading QEMU's output. It answers the protocol nettest
speaks and then closes -- the close is load-bearing, since nettest reads
until end of file and a listener that stayed open would hang the guest
rather than fail it.
The NIC is spelled out (-netdev user + -device e1000) rather than left to
QEMU's default. Other runs can take the default and still answer "did a
driver bind"; a run that asserts on an address should be reading its own
configuration.
/proc/netif is new, and the address is why. Until now the only report was
a line on the console at boot, which userland cannot read and a test can
only find by scraping the transcript. It gives interface, MAC, state,
address, netmask, gateway, MTU and frame counts. Named netif rather than
net on purpose: /proc/net is a *directory* on Linux, and a plain file
there would give ENOTDIR to anything reaching for /proc/net/dev, which is
worse than not having it. The counters are kept at the lwIP seam rather
than in the driver, because a frame the driver received and the netif
dropped never reached the stack.
Verified: 38 [ok], no failures. Eleven of them are new -- lwIP up, DHCP
obtaining 10.0.2.15 from SLIRP (a number nothing in the guest knows in
advance, so it is the exchange completing rather than a guess), the four
/proc/netif fields, frames counted in both directions, the handshake, the
reply read back, and the listener confirming HELLO-FROM-SCHOS arrived.
One bug found by writing it: $netPort was chosen after the command list
was built, so the guest ran `nettest 10.0.2.2` with no port and printed
its usage. An unset PowerShell variable interpolates to nothing rather
than failing, which is worth remembering next time a command in this
file looks right and behaves oddly.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
lwIP 2.2.0, fetched by ports/lwip and compiled straight into the kernel.
Not written from scratch, and the reason is not the packet formats: it is
congestion control, retransmission, window management and twenty years of
corner cases that only show up against a real peer. Not FreeBSD's
netinet either, though the licence would have allowed it -- that stack is
not a library, it reaches into mbufs, epoch(9), netisr, VNET and uma(9),
so adopting it means porting FreeBSD's kernel infrastructure first. lwIP
was written for exactly this situation: a small kernel with a raw
Ethernet driver and a port layer of a few hundred lines.
It is kernel code, so nothing is installed into the sysroot and there is
no second build system: the top-level Makefile compiles the files it
wants against our port headers. The file list is explicit rather than a
wildcard, because lwIP ships IPv6, 6LoWPAN, PPP, SLIP, DNS-SD and a dozen
apps in the same tree, and building what is not configured yields either
dead code or a link error naming a port function nobody meant to write.
Its sources are unmodified and its warnings are silenced for its files
only -- patching them would mean maintaining a fork.
The port layer is ours: a lock, a clock, the netif that hands frames to
e1000, lwipopts.h, and a shim/ directory of tiny headers so lwIP's
<string.h> and friends resolve to the kernel's functions instead of the
host's, whose prototypes disagree and would otherwise be found on the
default include path.
NO_SYS=1, so lwIP has no threads of its own and one kernel thread feeds
it. A thread rather than the interrupt handler, because a call into lwIP
can reach the driver, the driver can sleep, and an interrupt handler
cannot -- so the interrupt still only clears the cause register and the
thread polls. kthread_create is new and generally useful: a process with
a kernel-only page directory, no files, no cwd, no parent, dispatched
from forkret, which can therefore sleep and be preempted like anything
else. DHCP obtains an address at boot and the console says which one.
AF_INET went in exactly where phase 2 said it would. socknew dispatches
on the domain, kernel/net/inet.c implements the internet side on lwIP's
raw API, and not one syscall changed shape: bind, connect, listen,
accept, read and write reach the same descriptors they always did. TCP
and UDP both work. cmd/nettest opens a connection, writes, and reads the
answer back.
Two things are owed and named here so they are not discovered later:
recvfrom does not yet report the source address, and the socket paths
poll on the timer tick instead of sleeping on the socket. Both are
follow-ups, not redesigns.
NOTICE carries lwIP's BSD-3 text. Clause 3 is no-endorsement, not the
old advertising clause, so it sits with Apache-2.0 without difficulty.
Verified: the tree builds clean, and the headless boot check passes with
27 [ok] and no failures with the stack compiled in and the thread
running -- which is the part that would break first if netinit deadlocked
or the thread starved anything. An end-to-end connection test belongs in
that suite and is the next commit.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The last three phases of the filesystem, which were written together and
touch the same files throughout -- kmount.c, objw.c and vfs.c are each
rewritten by all three -- so they land as one commit rather than three
that could not be built separately.
Durability. A transaction that reaches the platter out of order is worse
than one that never reached it: the uberblock can name a tree whose
blocks are still in the drive's write cache. bdirty/bflush give the
buffer layer an explicit barrier, ideflush issues ATA FLUSH CACHE and
usb_storage_flush the SCSI SYNCHRONIZE CACHE that means the same thing,
and a commit now waits for the data before it writes the root that
points at it. The uberblock groups are read newest-first with fallback,
so a volume whose newest roots are torn mounts at an older one and says
which generations were lost.
Reflink. Extents carry a refcount, so two objects can name the same
blocks and neither frees them while the other holds them. Exposed as
FICLONE -- Linux's ioctl value, so `cp --reflink` works unmodified --
and as cmd/reflink for the case where there is no cp to hand.
Attributes, and the one genuinely new idea here. attr.c stores named
values on an object, reachable through setxattr/getxattr/removexattr at
Linux's syscall numbers. On top of that, schfs_index declares an index
over an attribute name and schfs_query answers "which objects carry this
value" from the B-tree rather than by walking. A query against an
unindexed attribute is refused with an error, never silently answered by
a full scan: a query that quietly becomes a scan is worse than one that
admits it cannot help. cmd/schattr drives both.
Three bugs fixed on the way, each found by the numbers being wrong
rather than by inspection:
- statfs's f_fsid was declared int64[2]. Linux's fsid_t is two *ints*,
so every field after it landed one slot late: userland read f_frsize
out of f_namelen, got 255, and df reported a 32 MB root filesystem
as 1992 KB. It hid for as long as it did because the fields used to
be invented constants, so wrong numbers looked like the usual ones.
- struct inode's nlink was a short against SchFS's uint32, so a file
past 32767 names came back negative -- and iput frees an object
whose count reaches zero. It is an int now, signed on purpose, so an
underflow lands somewhere a check can see rather than wrapping.
- filewrite split every write to fit v1's log. That limit belongs to
the journal, not to writing, and SchFS makes one transaction of
whatever it is handed: a 1 MB write was becoming 170 commits and
170 flushes for nothing. fs_ops carries max_write now, and SchFS
sets no limit.
Also here: a real mount table, so `mount -t schfs /dev/hdb /mnt` shows
up in /proc/mounts -- df looks a path up there before it calls statfs at
all, so a mount missing from that list was invisible to it -- several
volumes mounted at once, snapshots mounted by name, and /proc/fsstat
with an rdtsc-based counter, which is the only instrument fine enough to
see inside a single syscall.
Verified: `make schfs-check` passes in full, including the sweep that
crashes the volume at each of 119 points mid-transaction and requires
every one to mount with exactly its committed data, and the case where
both newest uberblocks are rubbish and the volume still comes back. The
headless boot check passes with 27 [ok] and no failures, on a SchFS root.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
struct file already carried a uint64 offset and bread already took a
uint64 block number. The narrowing was one boundary in the middle:
devsw's read and write took a 32-bit uint, so fileread passed a 64-bit
f->off into a 32-bit parameter and the top half was discarded.
The effect was not a refusal at 4 GB but silent wraparound past it. A
read at 5 GB returned the bytes at 1 GB, reporting success. disk.c had a
DISK_MAX_BYTES constant documenting the limit, and nothing anywhere
referenced it -- the cap was described but never enforced, which is the
worst of the three options.
The signature is now uint64 across all nine files that carry it, and
diskread/diskwrite no longer narrow the block number on the way to
bread. Widening this is a prerequisite for NVMe and for any real disk:
namespaces start well above 4 GB and the failure mode reads valid data
from the wrong place.
Verified end to end rather than by inspection. A 6 GB sparse image with
"PAST4GB-MARKER-OK" written at byte 5368709120, attached as /dev/hdb to a
ramdisk-root kernel so device 1 was free:
# dd if=/dev/hdb bs=4096 skip=1310720 count=1 | od -c | head -2
0000000 P A S T 4 G B - M A R K E R - O
Under the old code that offset truncates to 1 GB, where the image is
zeroes. The whole path is exercised: f->off, devsw, diskread, bread and
ide.c's LBA48.
Worth recording for next time: `make ramroot` stages the ESP itself, and
running `make esp` afterwards rebuilds the kernel without RAMROOT and
silently undoes it. The tell is a 855 KB kernel where a 25 MB one is
expected.
110 [ok], no failures, on the default build.
Two bugs on the ExitBootServices path, both invisible under OVMF and both
aimed squarely at real firmware.
The retry did not re-translate. If ExitBootServices returns an error the
loader re-reads the memory map and tries again -- correct, and the
specification expects it -- but the e820 table handed to the kernel was
still the one built from the *previous* map. The retry exists precisely
because the map changed, so the kernel was being given a description of
memory the firmware had already rearranged. QEMU never takes this path,
which is why it survived. Translation is now a function called on both.
Truncation was silent. phystop comes from the largest usable region, so
dropping the tail of the map can leave the kernel believing a large
machine has a fraction of its RAM -- and nothing said so, because the
loop simply stopped at the cap. It now warns while there is still a
console to warn on; after ExitBootServices there is nowhere to report
anything.
SCHBOOT_E820_MAX goes from 32 to 40. The map sits at BASE+0x208 and the
next field is at BASE+0x600, so 1016 bytes are reserved and 42 entries of
24 bytes fit. Real firmware emits far more descriptors than QEMU's; the
headroom was already allocated and unused.
Coalescing adjacent same-type runs is what keeps any of this in budget,
and it works because boot services memory becomes usable at this point --
that is most of the fragmentation in a UEFI map.
110 [ok], no failures, and the map still reads 7 regions under OVMF.
Groundwork for the unified loader: before the two paths can share code,
they have to agree on what they are producing, and until now they only
agreed by coincidence.
Every field of the loader/kernel hand-off was #defined in whichever file
happened to need it -- kernel/core/main.c, kernel/dev/vga.c,
kernel/arch/x86_64/e820.c, boot/bios/cdboot.S, boot/bios/stage2main.c and
boot/uefi/main.c, six copies of the same addresses. schboot.h existed for
exactly this and held only the entry magic and the RSDP slot, with a
comment describing the rest of the block it did not define.
The whole block lives there now: the VBE info fields, the verbose and
media flags (with the real-mode segment:offset spelling cdboot.S needs),
the e820 count, map and cap, and the RSDP. Every one of those files
includes it instead of restating it.
This is not a tidy-up. It is the same hazard as defining the entry magic
twice, only quieter -- nothing breaks until one copy moves, and the two
loaders diverging is precisely what produced the EFER.SCE and RSDP bugs,
where a fix on one path silently did not exist on the other.
Verified on both firmwares: BIOS boots to schinit with the same e820,
ACPI and VGA lines as before; UEFI reaches cpu0 with 1018 MB in 7 regions
and its own 1280x800 mode. `make iso` still builds. bootcheck 108 [ok].
Next: the loaders become starters that gather firmware-specific facts and
hand to one common stage. That needs the BIOS side to climb to long mode
first, since common code can only be compiled for one CPU mode -- which
is the risky part, and is why the contract went first.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Two loose ends from the mount-by-name commit, both closed.
A slot is a device number, so supporting more than one snapshot is a
small array rather than a redesign: SCHFS_SNAP_DEV through
SCHFS_SNAP_DEV + SCHFS_SNAP_SLOTS - 1, each with its own volume and its
own scratch pages, allocated on first use. Four of them -- "one at a
time" is an odd thing to have to explain, and the cost is two pages per
slot. Not many, because each one is a device number and those are a
small fixed space.
The leak mattered more. umount removed the mount table entry and left the
slot marked in use, so a snapshot could be unmounted and then never
mounted again until reboot. sys_umount2 releases it now, keyed off the
device it already has in hand.
Verified in the running system: two snapshots mounted at once, both
readable, both unmounted, and a third mount afterwards succeeds and reads
correctly -- which is the case that failed before.
schfs-check passes; bootcheck 108 [ok].
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
`mount -t schfs @golden /mnt` mounts a snapshot. The source argument
finally means something: sys_mount ignored it outright, because there was
one volume and nothing to choose between. `@name` rather than a path
because a snapshot is not a device -- there is no node in /dev for it and
there should not be, since it lives inside the volume already mounted.
A second volume rather than a parameter threaded through all 23 schfs_k*
entry points. Only the read side can ever address it, so the write path
keeps talking to the live volume and needed no change; the four read
entry points grew _dev variants and pick the volume from the device the
inode already carries. One fs_ops table serves both, because every entry
takes an inode and dispatches on it -- a second table would be the same
pointers under a different name.
It needs its own device number: a snapshot's object ids are the *same
numbers* as the live tree's, so one device would make the two views alias
in the inode cache and hand a reader whichever was cached first.
Three things this turned up, none of them new but all of them reachable
for the first time:
- sys_mount refused with `dev == SCHFS_DEV`, meaning "do not mount SchFS
onto SchFS". A snapshot's mount point is necessarily on the live root,
so that refused the entire feature. The real condition is mounting a
device onto *itself*, which is `dev == fsdev`.
- createat allocates an inode and then panics if dirlink fails --
reasonable about a writable filesystem, a crash on a read-only one.
The refusal now happens before anything is allocated. Unlink and
rename already handled a refusal properly.
- a refused create reported ENOENT, so writing to a snapshot said "No
such file or directory" -- the sentence a shell prints for every
create failure. It reports EROFS now.
Verified in the running system against an image built with
--snapshot golden: the mount succeeds, ls lists the snapshot's tree,
/mnt2/etc/motd reads back through it, and a write is refused with
"Read-only file system".
One at a time, and no umount path for it yet: mount_del_by_root removes
the mount table entry, but ksnap_unmount is only called on the failure
paths. schfs-check passes; bootcheck 108 [ok].
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A snapshot is a tree root that is kept. The B-tree is copy-on-write, so
the blocks a committed root reaches are already immutable -- nothing
overwrites a block a committed tree can reach -- and retaining one costs
a record: (SNAP_OBJECTID, KEY_SNAP, id) -> struct schfs_snap.
The record names the root as it was *before* the record was added.
Inserting it rewrites the path under copy-on-write and produces a new
root, so a snapshot that included its own creation would name a block
nobody had written yet.
What makes it survive is the derivation. There is no free-space map on
disk; it is rebuilt at mount by walking the tree, and that walk now
covers every snapshot root as well as the live one. Without it a block
reachable only from a snapshot looks free, gets handed out, and
overwrites the data the snapshot exists to keep -- silently, surfacing
only when the snapshot is read. A snapshot's own tree may hold older
snapshot records; those are deliberately not followed, or deleting a
snapshot could be undone by an older one's memory of it.
Runtime frees are suspended while a snapshot exists. Knowing which freed
blocks a snapshot still needs requires per-extent reference counting,
which is a design and not a flag; until that exists the safe answer is to
free nothing and let the next mount's derivation reclaim what is
genuinely unreachable. Allocation is unaffected and stays correct -- it
works from the in-use bitmap the derivation filled. The running
free-blocks *counter* does drift from the truth while a snapshot is
present, which is cosmetic and is reported by the test rather than
asserted away.
Tested in schfs_objtest: after a snapshot, the live tree shows new
contents while the snapshot still shows the old ones; a file deleted from
the live tree is still readable through the snapshot; deleting the
snapshot removes the record and the space comes back at the next mount.
schfs-check passes including the round-trip; bootcheck 108 [ok].
Not yet: snapshots are not reachable by name from userspace -- there is
no mount-a-snapshot path, so this is the format and the mechanism, not
the feature a user sees.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
SchFS is a 64-bit format addressing 64 ZB. The stack under it was not,
and the lowest limit was three orders of magnitude below the one anybody
would have guessed:
ide.c panicked on any block >= FSSIZE -- 150000 blocks, 614 MB
ide.c computed the sector in an int, and addressed it with LBA28
bio.c took a 32-bit block number, capping any device at 16 TB
schfs refused blocks past 0xffffffff, mirroring the cache
The FSSIZE check is the one that mattered. FSSIZE describes the size of a
v1 filesystem *image*; it had no business bounding a physical device, and
it meant a disk larger than half a gigabyte panicked the kernel rather
than being read.
struct buf's blockno is uint64 now, and so is bread's. idestart computes
the sector in 64 bits and picks LBA48 when it does not fit in 28 -- per
request, so everything that fits stays on the LBA28 path every existing
image has been read through. The 48-bit registers are written high half
first, because each is a two-deep FIFO and the other order leaves the
drive addressing whatever the low bytes alone describe.
Verified with a 2 GB disk carrying a marker at 1 GiB -- block 262144,
well past the old panic at 150000. The guest reads it back:
"MARKER-PAST-THE-OLD-LIMIT". schfs-check passes including the round-trip;
bootcheck 108 [ok].
Two limits deliberately left, both noted rather than guessed at: raw
/dev/hdN access still goes through devsw's 32-bit offset, so a raw device
is addressable to 4 GB; and usb_storage computes a 32-bit LBA, which is
what SCSI READ(10) carries -- READ(16) is a different command, not a
wider variable.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Groundwork for usb-tablet, which is the fix for the guest cursor and the
host's never being in the same place: a PS/2 mouse reports how far it
moved, two pointers accumulating deltas independently drift apart, and no
amount of care in the compositor can make them agree. An absolute device
reports where it is, so they coincide by construction.
- the HID driver no longer matches the boot subclass only. usb-tablet
declares subclass 0 and was refused at the door; hid_probe decides
what an interface actually is, and still declines what it cannot make
sense of.
- a non-boot HID interface is taken to be an absolute pointer and is
not asked for boot protocol, which it does not have -- that request
failing is what stopped it binding.
- tablet_report decodes the fixed 6-byte layout (buttons, 16-bit X,
16-bit Y, wheel) and scales the logical range to pixels there,
because it is the only place that knows the range.
- input_mouse_abs reports EV_ABS/ABS_X/ABS_Y so a reader can tell an
absolute position from a delta, and mouse_set_abs places the pointer
rather than nudging it.
- Schcom treats EV_ABS as authoritative and EV_REL as a nudge.
The device binds and identifies correctly: "usb-hid: tablet (absolute) on
slot 1, endpoint 0x81 every 4 ms".
It is NOT finished, and QEMU_INPUT is empty by default because of it. No
reports have yet been observed arriving on the tablet's interrupt
endpoint, so making it the default pointer would trade a cursor that
drifts for one that does not move at all. Opt in with
make qemu QEMU_INPUT='-device qemu-xhci,id=xhci -device usb-tablet,bus=xhci.0'
Next step is to find out why nothing arrives: whether the first transfer
is queued but never completes, whether a completion needs re-arming for
this endpoint, or whether QEMU is not routing motion to the tablet at all
(`info mice`).
bootcheck 108 [ok], schcom-test passes.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
kernel/dev/mouse.c held the pointer inside literal 800x600 bounds. That
was correct while 800x600 was the only mode the loader could set, which
it was until the resolution became a boot-time choice. At 1280x800 the
pointer simply stopped two thirds of the way across and would not go
further -- which reads as the cursor drifting away from the host's rather
than as a clamp.
vga.c grows vga_get_width/vga_get_height for it. The mode is chosen at
boot now, so nothing above the framebuffer may hardcode its size, and
these are the first two callers that needed to ask.
Verified that the system still builds and boots at 1280x800, and that
bootcheck is 108 [ok] and schcom-test passes. I did not manage to
photograph the pointer past the old limit in the time available -- there
is no reader for /dev/mouse in the image and the screendump attempt did
not resolve, so the fix rests on inspection rather than on a picture.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Nothing outside the kernel could touch a disk. sys_mount does not even
look at the device it is handed -- it mounts the one SchFS volume found
at boot and ignores `source` -- so a program had no way to read a
partition table, format a volume or write a boot sector. An installer is
a program that needs all three, so this is the first piece of it.
A character device in devsw rather than a separate block layer, because
that is where the offset already lives: fileread/filewrite pass f->off
and advance it by what comes back, which is everything a seekable device
needs. The minor is the IDE device number, the same numbering iderw()
uses; two schemes for the same disks would be a bug waiting to happen.
Writes bypass every filesystem above them, which is inherent to a raw
device and why an installer must target a disk it has not mounted.
The nodes are created unconditionally, like every other entry in /dev.
That matters more than it looks: with no node, `echo x > /dev/hdb`
silently creates an ordinary file of that name and appears to work. It
did, and it agreed with itself on read-back; only the host's copy of the
disk showed nothing had been written.
st_rdev was never filled, so every node in /dev reported "0, 0" under
ls -l. Harmless while nothing read it, wrong as soon as something has to
tell two disks apart. Filled with the full Linux dev_t encoding rather
than (major << 8 | minor), which only works while both stay small.
Verified against the host as an independent oracle, both directions: the
guest wrote to /dev/hdb and the host saw those bytes at offset 0; the
host wrote at offset 4096 and the guest read them back with dd skip=.
ls -l now shows 1,1 for console, 2,0 for fb0 and 8,1 for hdb.
Known limit, commented in disk.c: devsw passes the offset as a 32-bit
uint, so a raw disk is addressable to 4 GB. That must be widened before
anyone installs onto a real disk.
bootcheck 108 [ok] with -Fb -Tty -Usb -Sock -Evdev -Ps.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
find_rsdp() scanned the EBDA and 0xE0000-0x100000. That is a BIOS
convention, and UEFI firmware is under no obligation to leave a copy
there -- OVMF does not. So the kernel found no RSDP, no MADT and no
IOAPIC, and nothing routed IRQ 4. Serial output is polled, so the machine
booted all the way to a shell prompt and then ignored every key, which
reads as a broken console rather than a missing table.
The UEFI loader reads the RSDP out of the EFI configuration table,
preferring ACPI 2.0 for its XSDT, and leaves it at SCHBOOT_RSDP_ADDR --
the same low-memory hand-off block the e820 map already uses, now
described in include/schboot.h so the two sides cannot disagree.
The kernel validates it rather than trusting it: the BIOS path never
writes that slot, so there it holds whatever was in memory, and following
a wild pointer faults before there is anything to report the fault with.
Signature and checksum do that. It deliberately does not require 16-byte
alignment -- that rule is about searching the BIOS area, and OVMF's RSDP
sits at 0x3f77e014, so demanding it threw away the one address this code
exists to use.
UEFI now reports "acpi: 1 cpu, lapic at ..., ioapic id 0" and the ISO
boots to an interactive shell: uname, ls /bin (241), writing and reading
a file, ps, /proc/schfs -- all driven over serial.
BIOS unchanged: bootcheck 108 [ok], same ACPI line as before.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Every driver reached its registers with P2V(bar.base), which assumes the
BAR sits inside the 0-4 GB the direct map covers. Under SeaBIOS it always
does. Under OVMF it does not: firmware may place a 64-bit BAR anywhere,
and it puts virtio-gpu's at about 768 GB. P2V produced an address with no
translation behind it, so the driver's first register write took a page
fault, then a double fault, then a triple fault and a reboot loop --
which, coming from a driver bind, reads as broken hardware rather than a
missing mapping.
ioremap() carries each region on a 1 GB page in PDPT entries 4-511, which
were unused: the direct map occupies 0-3, and kpml4[0], kpml4[KERNBASE]
and every process's PML4 all share that one table, so a mapping appears
in every address space at once with no page table to allocate. The pages
are cache-disabled and write-through for the same reason the top GB of
the direct map is -- these are registers, and a write sitting in a cache
line is a write the device never sees.
A physical address below 4 GB returns the direct-map address unchanged,
so the BIOS path, where every BAR is low, is byte for byte what it was.
virtio-gpu now initialises under OVMF (1280x800, queue 64) instead of
triple faulting. e1000 and xhci had the identical latent bug; xhci was
already refusing to bind above 4 GB with a comment saying the mapping did
not exist yet, and now it does.
BIOS unchanged: bootcheck 108 [ok] with -Fb -Tty -Usb -Sock -Evdev -Ps,
e1000 and xhci both binding.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
e820_init() took the usable run *containing the kernel* and treated its
end as the top of RAM, and kinit1() walked a single range from the end
of the kernel image to there. That is the shape of a BIOS map. It is
not the shape of a UEFI one: OVMF places an ACPI NVS region at 8 MB,
so the kernel's run ended there, the map failed the PHYSTOP_MIN check,
and the kernel fell back to asserting 224 MB it had not measured --
then freed straight through the reserved hole anyway.
Keep the usable regions as a list as well as a top address. kinit1()
now frees each one, clipped to the window main() passes in, so the
holes are skipped rather than either freed or fatal. PHYSTOP still
means what it always did -- the top of what kalloc may hand out, and
the bound kfree() rejects pages above -- so no call site changes.
phystop comes from the largest region rather than the highest. The DMA
arena is carved off the top of PHYSTOP and must be physically
contiguous, and the largest region is the only one certain to have
room; a smaller region above it is left unfreed rather than raising
the bound past a hole, because kfree()'s check is worth more than the
pages it costs. Report the usable total separately from the top
address, since with a fragmented map they are different numbers and
printing one as the other is how 8 MB went unnoticed.
Verified on both paths. BIOS: unchanged at 1023 MB in 2 regions,
bootcheck 27/27. UEFI under OVMF: 1018 MB in 7 regions, up from the
8 MB that triggered the fallback, booting through vga, e1000, the RTC
and on to cpu0.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Loads the kernel from the ESP, places its ELF segments, records the
framebuffer, translates the UEFI memory map into E820 entries, exits boot
services, builds the page tables the kernel expects and jumps. Verified
under OVMF: the kernel starts, brings up PCI, VGA, e1000 and the RTC, and
reaches "cpu0: starting".
UEFI structures are written from the specification rather than importing
gnu-efi or EDK II -- a large dependency for two hundred lines of use, and
a licensing question the project does not need to answer. clang emits
PE/COFF directly, so this needs no new toolchain.
The kernel gained a second entry point. The multiboot entry is 32-bit
code that climbs into long mode itself, and firmware that already put us
there would have to drop back down only to climb out again -- a mode
switch whose failure mode is a triple fault with no console. So entry.S
publishes its 64-bit entry the same way it publishes the multiboot
header: a magic to scan for, SCH64BOO, with include/schboot.h owning the
constants so the two cannot disagree. That header goes *after* the
multiboot header, never before: header_addr says the multiboot header
lives at mboot_load_addr, and the BIOS loader computes the kernel's file
offset from it.
The kernel loads its own GDT at that entry rather than the loader doing
it. Arriving in long mode means arriving on the firmware's descriptors,
and the first thing depending on the kernel's own selectors faulted --
"unexpected trap 13 ... cs=38", a long way from the cause.
KNOWN LIMITATION, diagnosed and not yet fixed: e820_init takes the usable
run *containing the kernel*, which under OVMF ends at 8 MB -- the
firmware places an ACPI NVS region at 0x800000, so the run is genuinely
fragmented. The map is truthful and untruncated (21 entries); the
assumption of one contiguous run is BIOS-shaped. The fix is for the
kernel to free each usable region rather than one range, which touches
kalloc and deserves its own change rather than the tail of this one.
BIOS boot is unaffected: 27 checks, no failures.
LOGO_SRC named "Assets/SchOS Small Logo.png", which no longer exists --
the small logo was replaced by the full logo, the same mark slightly
redrawn. `make logo` had been failing since. Points at the full logo now;
not the text logo, which is the wordmark and would be illegible scaled to
128 px. Verified by photographing the splash: the mark renders as the
circle with Sch over OS.
/etc lived only in sysroot/, which is generated and gitignored, so
/etc/passwd, /etc/group, /etc/profile, /etc/kshrc and /etc/motd would all
have been missing from a fresh clone -- and the machine boots to a shell
that reads them. They are tracked under etc/ now and installed by
etc-install alongside the schinit manifests.
/etc/motd still said "Schunix": the rename swept tracked files, and this
one was excluded from version control at the time. Fixed.
Noted while checking: sysroot/bin/sh still contains the symbol
__schunix_tp, so the staged mksh predates the rename of the musl TLS
patch to __schos_tp. Harmless -- it is an internal symbol -- but it
confirms the staged ports are older than the patches that produce them,
and they cannot be rebuilt on a Windows host. It resolves the next time
ports are built on Linux.
Milestone 1 of three: manifests, RunAtLoad, KeepAlive with throttling,
and reaping. Replaces the hardcoded fork-a-shell loop in cmd/init, which
stays in the tree for now.
Jobs are sectioned key=value files in /etc/schinit/*.job, kept in the
repo under etc/ and installed into the sysroot -- sysroot/ is generated
and gitignored, so anything written only there is lost on a fresh clone,
which is already true of /etc/passwd and should not become true of the
boot configuration. Unknown keys and sections are ignored rather than
fatal: a manifest written for a newer schinit must not stop this one
booting the machine.
Socket sections are parsed and recorded but not yet held. That is
milestone 2 and it is the whole reason for copying launchd rather than
sysvinit -- the supervisor holds the listener and starts the job on first
connect, which inverts the dependency graph and makes runlevels a
non-concept. The key exists now so manifests need not change later.
getdents64 is exposed to userland for this. The kernel has had it since
the Linux ABI landed, but nothing declared it in user.h or lib/usys.S, so
first-party programs had no way to list a directory at all.
Two things found by testing rather than reasoning:
- setsid() gives a job its own session, and a session has no
controlling terminal. The console then delivers input only to the
foreground process group it tracks, so the shell came up, echoed
every keystroke through the line discipline, and executed none of
them -- indistinguishable from a hung shell. ControllingTerminal=yes
claims it with TIOCSCTTY/TIOCSPGRP; off by default, since a daemon
has no business seizing the console.
- the first bootstrap() dropped /dev/input/event0-1 and most of /proc,
because it was transcribed from the part of the old init that had
been read rather than all of it. -Evdev caught it.
build/usbstick.img is built by `all` now: without it QEMU refuses the
-drive and the entire -Usb run produces no output, which reads as a
broken USB driver.
Seven suites green: base 27, Tty 51, Ps 32, Sock 64, Fb 36, Usb 31,
Evdev 29. make schfs-check passes.
boot/bios/ now holds the MBR, the El Torito stub and the real-mode stage
2; boot/uefi/ is empty and documented.
Separate directories rather than #ifdefs, because the two share almost
nothing. A BIOS loader lives in 16-bit real mode, reaches the firmware
through software interrupts, and walks itself up into long mode. A UEFI
loader is a 64-bit PE that the firmware has already placed in long mode
with paging on, calling services through a table of function pointers.
What they do share is the hand-off contract, which boot/README.md now
writes down: ELF segments placed, long mode, a 24-byte-entry memory map
at E820_MAP_ADDR, framebuffer geometry at VBE_INFO_ADDR.
That contract is the point of documenting it now. UEFI's GetMemoryMap
returns EFI_MEMORY_DESCRIPTORs, and the right move is to translate them
into the entries kernel/arch/x86_64/e820.c already parses rather than
teach the kernel a second format.
Build output paths under $(BUILD)/boot are unchanged; only sources moved.
Two bugs reported from actual use, plus a test harness that had stopped
testing.
init could be killed by Ctrl-C, taking the machine with it. SIGINT goes
to the foreground process group, init is pgid 1, and any window where
tty_fg_pgrp is 1 -- before a shell claims the terminal, after one exits,
or between the two while the key is held -- lands it on init. init has
no handler, so killed was set, exit() ran, and exit() panics outright on
initproc. Spamming Ctrl-C at a frozen program is exactly the input that
finds that window. POSIX and Linux both discard a default-action signal
for pid 1; init_ignores() now does the same, SIGKILL included, because
init is not killable from userspace on any Unix.
Everything was slow because BSIZE was 1024 while a SchFS block is 4096,
so every block cost four IDE requests, each with its own interrupt and
sleep/wakeup. exec() of the 800 KB toybox binary meant ~800 round trips.
The giveaway: `seq 1 200`, which touches no file and prints nothing,
cost the same 2.0s as `cat /bin/toybox`, and `ls | wc -l` cost exactly
double -- both dominated by loading the same binary twice.
BSIZE is 4096 now, one request per SchFS block, and NBUF is 256 blocks
(1 MB) rather than 90 KB -- smaller than the binary every command execs,
so nothing was ever still resident. ideinit() issues SET MULTIPLE MODE,
which is what makes a 4096-byte transfer legal: READ/WRITE MULTIPLE move
the drive's multiple count per interrupt and nothing had ever set it,
which is the real reason idestart() carried a sector_per_block > 7 panic.
seq 1 200 2.02s -> 0.06s
ls /bin 2.76s -> 0.68s
ls /bin | wc -l 4.68s -> 0.73s
And the harness. bootcheck called Find-Qemu from a param() default,
which is bound before any function in the body exists -- so it died on
its first line and printed an error containing neither "FAIL" nor
"Exception", the two things being grepped for. It had not run since it
was made host-portable. Worse, an earlier index-based edit had deleted
the Expect blocks for -Tty, -Usb, -Sock, -Ps and -Evdev, so five of the
seven modes asserted nothing at all. Restored, and the USB run gets its
own 16 MB build/usbstick.img: it had been pointed at build/fs.img, which
nothing builds since SchFS became root, so QEMU refused to start.
242 checks across seven modes, all passing, and make schfs-check green.
Removes the "-unix" element and with it any trademark exposure to The
Open Group.
The split follows the ecosystem layout: SchOS is the system, and the
kernel has its own name. Kernel-identity strings are the only places
that take SchIFT -- uname(2) sysname, /proc/version, and the two licence
sentences that say "kernel" outright. That is what every Unix does here:
`uname -s` on Linux reports the kernel, not the distribution. Everything
a user reads as the product -- panic screen, boot menu, error prefixes,
docs, LICENSE, NOTICE -- is SchOS.
schunix.img is schos.img, the logo assets and the musl TLS patch are
renamed, and __SCHUNIX__ is __SCHOS__ (the define and the patch that
consumes it moved together).
Two things the rename broke, both found by the bootcheck suite:
- shmtest wrote "SCHUNIX-MMAP-OK" with a hardcoded length of 15 and
compared it back by hand-counted offsets (m[3]=='U', m[14]=='K').
The shorter marker left the write running two bytes past the literal
and the compare reading past the data, failing a file-backed mmap
that was working correctly. Now written and compared through one
constant with sizeof.
- bootcheck asserted a 31-byte file after appending to
"hello-from-schunix"; the shorter string makes it 29.
Both are the same mistake: a string literal whose length is written out
somewhere else. Full suite green -- base, Tty, Ps, Sock, Fb, Usb, Evdev.
Nexoniarz