schdbgd bridges TCP to the debugger, which is the one way in that does not
need you to be sitting at the machine. It is a bridge and nothing else: a line
in becomes a line written to /dev/dbgro and the answer comes back, so `nc` is a
complete client and there is no clever client to be broken on the day you need
this.
# schdbgd &
schdbgd: debugger on port 4444, loopback only
$ nc 127.0.0.1 4444
dma
arena 0xffff80003da41000-0xffff80003ea41000 16384 KB total, ...
A kernel debugger plus a TCP socket is a remote root hole, so three things were
decided rather than discovered:
- It does not start by itself. schdbgd.job ships Disabled. SchOS has no
kernel command line, so the boot flag the plan asked for is the job
manifest, which is the same thing in this system's own idiom.
- It binds loopback; -a names an address and says so in a paragraph.
- It cannot reach a view that changes state, enforced in the kernel.
That last one is /dev/dbgro: /dev/dbg on minor 1. The minor is the one piece of
per-open state a character device already sees, so which node you opened is the
permission -- nothing to latch, nothing to remember between the write and the
read, no way across. Which views are refused comes from dbg_register_rw() on
the registration itself, because a central list of dangerous views goes stale
in the direction where a new view is quietly treated as safe. Today that is
`watch <addr>`; `watch` with no argument is still the listing and still
answered.
Also: socket() now accepts IPPROTO_TCP with SOCK_STREAM and IPPROTO_UDP with
SOCK_DGRAM. They name the default rather than departing from it, and toybox's
nc writes them out -- so TCP was "Protocol not supported" on a system whose
whole network is TCP.
docs/plan/debugger.md retires into docs/architecture/debugging.md; 16 files
repointed. Every plan in docs/plan/ that was open is now closed, so the next
thing is Gate 0 itself: the survey, with the instrument built.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
docs/plan/ldls.md L3a -- the class bridge, and the first of the three pieces
L3 turned out to be.
/dev/netdev is /dev/pci with the arrows reversed. That hands a device to a
process; this lets a process be one. The kernel side is netdev_ops and nothing
else -- transmit, receive, mac, link_up, the same four entries e1000
publishes -- so a process claims the device, supplies a MAC, and from then on
read() gives it frames the stack wants sent and write() hands it frames that
arrived.
cmd/netecho proves it and is deliberately not a driver: no hardware, none
claimed, every frame answered by the program itself. DHCP so the interface
gets an address at all, ARP so the stack can find it, ICMP so it can be
pinged. All built by hand from RFCs 826, 791, 792 and 2131.
# netecho &
netuser: ldls0 is a process now, 2:4c:44:4c:53:1
netecho: I am 10.9.0.1 and will offer 10.9.0.2
netecho: answered DHCP, offering an address
# cat /proc/netif
interface ldls0 driver netuser address 10.9.0.2 gateway 10.9.0.1
# ping 10.9.0.1 3
netecho: answered an ARP
netecho: answered a ping
3 packets transmitted, 3 received, 0% packet loss
The interface got its address by DHCP from a process and then answered ping
from one. lwIP never learns that the thing behind netdev_ops is not a driver,
which is the whole property the bridge exists to have.
It dies the way it should, too. kill the process and link_up() answers false
immediately, so the stack refuses to send -- "ping: sendto: -5" -- rather than
swallowing packets into a driver that is not there. Start another and the
*same* interface comes back, keeping its address. That came free:
netuser_release_all() sits beside pciuser_release_all() in exit() for exactly
the same reason, and the reason is that a process holding a device is one that
may crash.
One thing the kernel needed: netinit() starts DHCP on the devices that exist
when it runs, so an interface created later -- which is how an LDLS-hosted
driver arrives -- would have come up with no address and no way to get one.
net_attach_late() fixes that, and lives in kernel/net/net.c because that is
the only file allowed to include lwIP's headers.
Recorded as not done, because it is not: this proves the bridge, not a driver.
Nothing in that test loads a module. That is why it was done first -- if ping
had not worked here the bridge would be wrong, and now a failure with a real
driver can only be the driver's side. L3b is the SchOS-side host, L3c is the
fifteen symbols for real, and L3 is the two joined.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
docs/plan/debugger.md, milestones D0, D1 and D2. What they replace is the
ninety-second loop -- add a cprintf, rebuild, boot, read the serial log --
which answers exactly the one question you thought of before the rebuild.
The rule the shape follows: **it observes, it does not control.** No
stepping, no software breakpoints, no stopping the world. Half the bugs
this kernel produces are timing -- an interrupt storm, a level-triggered
line nobody acknowledges, a driver that faults only on its second start --
and every one of them disappears when you halt the machine to look. And
the layering rule, which is the circularity rule the display already has:
a debugger that needs lwIP cannot debug lwIP.
D0 -- symbols. tools/ksyms.pl turns the first link pass into a table the
second pass carries. Two passes is enough because .ksym is placed after
.text and .rodata, so carrying the table cannot move anything recorded in
it; the build lists both passes' text symbols and refuses to link if they
ever disagree, because a symbol table that is subtly wrong gives confident
false answers. ksym.c reads it through weak symbols, so pass one -- which
has no table -- still links. A crash chain now reads panic+0x5f <-
proc_sysrq_cmd+0x99 <- ... <- syscall+0xf6.
D1 -- a log that outlives the kernel. Every character through consputc()
lands in a byte ring with a timestamp per line, readable as /proc/kmsg.
It sits in a region carved before the page allocator gets anything, with a
magic header, so a kernel finds the previous kernel's log still there --
including the whole PSOGM report. panic.c's own comment said "there is no
core dump and no log to go back to"; that is no longer true.
The obvious place for it is the top of RAM beside the DMA arena. That
was the first attempt and the magic came back as ASCII on the reboot,
because the top of usable memory is exactly where the firmware makes its
own allocations. It lives above the kernel image instead, which is a
stable address across boots of the same build.
D2 -- the introspection registry. A subsystem publishes its state with one
line at the bottom of the file that owns the data:
dbg_register("dma", dma_dump, "the DMA arena's free extents");
Seven views -- dma, mem, proc, vma, claims, pci, sock -- reachable three
ways: sysrq 'g' (always, no process and no filesystem), /dev/dbg (echo a
line, cat the answer), and schdbg. The protocol is lines of text because
the same bytes have to work where nothing can allocate or frame a record,
and because a protocol you cannot drive by hand is one you cannot debug.
struct sbuf moved out of kernel/dev/procfs.c into include/sbuf.h rather
than being written a second time.
Several views exist because of a specific bug. `vma` is the table that
decides whether exit() frees a page, and the DMA ownership bug that made
this worth building was a block that was never in it.
Also fixed, found by building with clang for the first time in a while:
struct iovec was defined in both posix.h and socket.h, and every file that
included both was a redefinition. GCC 15 defaults to C23, which permits an
identical struct redeclaration, so the gcc path never noticed -- while
clang defaults to gnu17 and rejected it, breaking the one toolchain
shell.nix exists to provide and the UEFI loader requires. Both are now
guarded.
D3 (`whose <addr>`, the cross-subsystem ownership query) and D4 (hardware
watchpoints on DR0-DR3) are not built and are described honestly as
missing in the architecture doc's "What it does not do".
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VQ7Qx8E1q1M1bbCq8QLHS2
sendmsg/recvmsg with SCM_RIGHTS. Item 1 of docs/plan/foundations.md and
mechanism 1 of docs/plan/hybrid.md, which are the same primitive seen from two
directions: Wayland's wl_shm_create_pool passes a file descriptor with no
integer-shaped alternative, and handing a driver process the device it is
allowed to touch is the same act. Both plans stopped here.
Linux's structures byte for byte -- msghdr, cmsghdr, the CMSG_ macros -- because
the programs that need this are not ours. libwayland and toybox both call
recvmsg with Linux's headers and would not notice a friendlier door.
The delivery rule is the whole design. A batch of descriptors is pinned to the
stream position of the first byte of the message that carried it, and that has
two halves: a reader arriving at that position is handed them, and a reader that
has not reached it is stopped SHORT of it. Leave out the second half and a
program asking for 4096 bytes swallows three messages and is handed one
message's descriptors, with nothing in the result saying which. socktest checks
exactly that: two one-byte messages, an eight-byte read, one byte back.
The kernel holds a reference to every descriptor in flight, so a sender may
close its own copy the moment sendmsg returns. Giving that reference back is the
part that leaks if it is not written down, and there are three ways a message
ends:
- received with room installed in the receiver's table
- received without room, or by
plain read(2) closed; MSG_CTRUNC for the first case
- never received, socket closed closed by sockclose
The middle row looks like a choice and is not: holding kernel references for a
receiver that has proved it is not looking is a leak with extra steps.
No garbage collector. A socket descriptor sent over that same socket and never
received holds a reference to the socket that holds the reference to it, and
neither close reaches zero. Linux carries several hundred lines for this.
Nothing on either plan passes a socket over a socket -- Wayland passes memory,
the driver work passes devices -- so it is a bounded leak that is documented
rather than a silent one.
Fourteen checks in cmd/socktest, and every one moves data through the descriptor
rather than inspecting it. The descriptor passed is always a pipe end, because a
pipe announces its own lifetime: a read of a pipe whose writers are all gone
returns 0, which turns "was the kernel's reference released" -- otherwise
invisible from userspace -- into a value a test can compare.
Two bugs fell out, neither in the new code:
- sockpair() never set the socket's domain, so every socketpair() socket
claimed domain 0. Invisible while the data paths asked only whether a socket
was AF_INET or AF_NETLINK, since zero answers "neither" correctly.
SCM_RIGHTS asks "is this AF_UNIX", and zero is wrong for that one.
- pipe_poll() reports POLLHUP but not POLLIN at end of file, so a non-blocking
read of a pipe whose writers had all gone returned EAGAIN forever instead of
0 -- a program that polls and reads spins waiting for an end that already
happened. sockreadable() always had the equivalent condition; the pipe path
did not. Found because the test helper for "is this descriptor closed" is a
non-blocking read of a pipe.
projects/schcom/README.md said SCM_RIGHTS was done for a long time while nothing
in the kernel implemented it, and an entire GUI plan was drawn on that line.
It is true now, which is not the same as it having been right, and the README
says so.
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
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
/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
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
Started as "fix the ASAN leak in schfs_objtest" and the test turned out to be
the messenger. The allocation map was taken through a palloc callback and there
was no way to return it: no schfs_space_release, and a comment in
schfs_space_init reading "caller frees what was taken" for an API the caller
did not have.
Three separate leaks from that one gap:
* The kernel leaked the entire map on every unmount -- up to 2 MB, at 8 KiB
per GiB of volume -- because kmount_unmount flushed, cleared `mounted` and
walked away. This is the one that matters; the others are test-only.
* schfs_space_init stranded a page on its own error path. npages was raised
only after *both* of a pair of allocations succeeded, so a failure on the
second left the first outside the count -- unreachable to any cleanup that
could later exist.
* schfs_objtest leaked 2121728 bytes in 518 allocations, which is what made
the whole SchFS suite exit non-zero regardless of its assertions.
The map now carries its own pfree rather than taking one at release time, so it
can only ever be freed by the allocator that filled it -- the kernel's pages
come from kalloc and a host tool's from calloc, and the two are not
interchangeable. schfs_vol_open and schfs_vol_format release any map the volume
is already holding before building another, which is what stops a re-open
leaking; struct schfs_vol must therefore be zeroed before first use, and the
header now says so.
The test's own leak was the last one found and the simplest: it does
memset(&VOL, 0, sizeof(VOL)) before reopening, which wiped the map's pointers
before vol_open could release them. Eight of those, each now releasing first.
schfs_objtest asserts the balance -- 520 pages taken, 520 given back -- because
a suite that leaks cannot detect a leak: the filesystem's own allocations would
arrive as noise inside the harness's. The next one fails the run instead.
make schfs-check now exits 0. The 31 UBSAN misalignment reports are still
there and are still a Gate 0 item.
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
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>
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>
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>
The one-entry cache lived inside schfs_file_read, so stat() never touched
it -- which is why `ls` barely moved. Moved into schfs_inode_get, where
every caller benefits, and widened to sixteen entries.
`ls` is the shape that matters: it stats every name in a directory, so a
single entry never hits and each file costs two descents, one to find the
name and one to read the inode. Sixteen keeps hitting through a directory
walk without a second reader evicting everything. Round-robin rather than
LRU: the pattern is a walk, not a working set, so the bookkeeping on every
hit would buy nothing.
Measured in the guest, TCG, against the original baseline:
ls /bin warm 0.70s -> 0.43s
ls /bin cold 2.80s -> 1.32s
read 834 KB 1.73s -> 0.35s
schfs-check passes including the round-trip; bootcheck 108 [ok].
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Every read descended the B-tree twice per block -- once for the inode and
once for the extent -- so an 834 KB file cost 418 descents, each a tree
walk plus a 4 KB copy out of the buffer cache per node visited. A
sequential read asks for the same inode every time and usually stays
inside one extent, so both answers were already known and thrown away.
One entry each, not a table: the pattern being optimised is a single file
read straight through, and one entry catches all of it without needing an
eviction policy. A second reader simply misses.
Correctness rests entirely on invalidation, so it is blunt: all five
write entry points in objw.c call schfs_cache_invalidate() before
touching anything. A cache that is merely usually right would hand back
another file's data. The volume pointer is part of the key because the
host tools can hold two images open at once.
Measured in the guest, bracketed inside one command line, TCG:
read 834 KB 1.73s -> 0.31s (5.6x, together with the BOUNCE fix)
ls /bin cold 2.80s -> 1.40s (2x)
ls /bin warm 0.70s -> 0.70s (unchanged)
Warm ls is untouched on purpose-by-consequence: it is 241 separate
inodes, so a one-entry cache never hits. That cost is the per-entry
lookup path -- dirlookup plus iload, each its own descent -- and it is
where the next order of magnitude is, not here.
schfs-check passes, including the round-trip; bootcheck 108 [ok].
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
BOUNCE was 2048 while the comment directly above it said "one page at a
time", and the buffer it sizes is a single kalloc() -- so half of every
bounce buffer went unused and every read did twice the work it needed to.
The chunk size is not a copy-size tuning knob. Each trip through the loop
takes the volume sleeplock and builds a fresh B-tree cursor to find the
extent for that offset, so it really sets how many tree descents happen
per byte read. An 834 KB file took 417 of them; it now takes 209.
Measured in the guest, bracketed inside one command line so nothing
depends on host timing: reading 834 KB goes from 1.73s to 0.90s under
TCG, cold and warm alike. Warm equalling cold is itself the tell -- the
cost was never the disk, it was the per-chunk work, which is why KVM was
6.4x faster on the same read while caching made no difference at all.
This does not fix `ls`, and the numbers say why: 2.80s -> 2.65s cold.
loaduvm already reads a page at a time, so exec was never the bottleneck
there. At 241 entries in 2.65s that is ~11ms per file, spent in the
per-entry lookup path rather than in reading anything.
schfs-check passes; bootcheck 108 [ok].
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
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.
An x86-64 operating system built from xv6, now largely rewritten: SMP,
a copy-on-write root filesystem (SchFS), PCI/USB/virtio drivers, a TTF
rasteriser, and a Linux-numbered syscall ABI.
No history before this point -- the tree was never under version control.
This commit exists so the rename that follows has something to undo to.
Nexoniarz