SchOS/docs/architecture/filesystem.md

# The VFS

The layer between `open(2)` and a filesystem. SchFS is described separately in
[schfs.md](schfs.md); this is the machinery it plugs into, plus the buffer cache,
mounts, and procfs.

---

## `struct fs_ops`

A filesystem is a vtable, chosen by `iget()` from the device number and never
null once an inode is held.

```c
struct fs_ops {
  const char *name;

  void (*iload)   (struct inode *);          // fill type/size/nlink from disk
  void (*istore)  (struct inode *);          // write them back
  void (*ifree)   (struct inode *);          // last link gone, release it

  int  (*read)    (struct inode *, int user_dst, uint64 dst, uint64 off, uint n);
  int  (*write)   (struct inode *, int user_src, uint64 src, uint64 off, uint n);
  void (*truncate)(struct inode *, uint64 newsize);

  struct inode *(*dirlookup)(struct inode *dp, char *name, uint *poff);
  int  (*dirlink)  (struct inode *dp, char *name, uint64 inum);
  int  (*dirunlink)(struct inode *dp, char *name, uint off);
  int  (*dirempty) (struct inode *dp);

  struct inode *(*ialloc)(uint dev, short type);
  int  (*getdents)(struct inode *dp, char *out, int cap, uint64 *off);

  uint max_write;          // largest write in one call; 0 for no limit
  uint refuse_types;       // 1 << T_* the filesystem cannot store: EPERM
  int  no_hardlinks;       // link() fails with EPERM
};
```

The implementations: `schfs_ops` (`kernel/fs/schfs/vfs.c`), `fat_ops`
(`kernel/fs/fat/fat.c`, see [fat.md](fat.md)), `procfs_ops`
(`kernel/fs/procfs.c`) and `devfs_ops` (`kernel/fs/devfs.c`).

**Creation can fail, and is undone when it does.** `ialloc` and `dirlink` used
to panic on failure, because on v1 and SchFS they cannot fail for a reason the
caller could have avoided. FAT can: the volume fills, the fixed root directory
of FAT16 holds 512 names, a name holds a character FAT does not allow. So
`createat_x()` gives the object back and reports `ENOSPC`, `EINVAL`, `EPERM` or
`EROFS`, and `open`, `mkdir`, `mknod` and `symlink` return that errno instead of
`ENOENT`.

**A short write is a full disk.** `filewrite()` panicked on one, which held as
long as no filesystem ever wrote short. Now `write(2)` returns what went in, and
the next call gets the error.

`poff` is deliberately **an opaque cookie** rather than a byte offset —
meaningful only to the same filesystem's `dirunlink`. A directory does not have
to be an array of fixed-size records for this interface to work.

`ialloc` creates an object **with no name**. The caller sets `nlink` and calls
`dirlink`; until it does, nothing can reach the object. If the caller dies in
between, the result is an unreferenced record that is harmless and, in SchFS,
never reaches the disk anyway — nothing is durable until the transaction commits.

### The one place the models disagree

The VFS inherited xv6's shape: allocate an inode, fill it in, link a name to it.
SchFS has no free-standing inode — an object exists because a record with its id
exists.

`nlink` is the sharper disagreement. **The VFS adjusts it itself** and then calls
`iupdate`, so SchFS's own link accounting is deliberately bypassed in the VFS
path: `schfs_dir_add`/`schfs_dir_del` are used rather than `schfs_link`/
`schfs_unlink`, and `iput`'s `ifree` hook releases the object once the VFS has
decided there are no names left.

> Two pieces of code counting links independently is precisely how a filesystem
> loses a file that is still referenced.

`nlink` is an `int`, not a `short`. SchFS stores a `uint32` and the narrower type
truncated it, so a file past 32767 names came back with a *negative* count — and
`iput` frees an object whose count reaches zero. Signed on purpose: `nlink--` on
a count that is somehow already zero should land somewhere a check can see it,
not wrap to four billion.

## Path resolution

`namex()` walks a path one component at a time. It was refactored into
`namex_from(struct inode *start, ...)`, parameterising what used to be a
hardcoded `idup(proc->cwd)`.

That refactor is what made the `*at()` family work. Every one of them — `openat`,
`fstatat`, `unlinkat`, `renameat`, `linkat`, `symlinkat`, `mkdirat`, `mknodat`,
`readlinkat`, `faccessat` — previously handled only two cases: `AT_FDCWD`, or an
absolute path with `dirfd` ignored. A **real** dirfd with a **relative** path is
the standard, efficient way any directory-tree walker avoids rebuilding full
path strings, and it was either rejected outright or — worse, in `openat` —
silently resolved against `proc->cwd`, which can open the wrong file entirely.

`sys_fs.c` now has `at_dir(dirfd)`, `at_namei()` and `at_nameiparent()`, and
every `*at()` call uses them.

## The buffer cache

`kernel/dev/bio.c`, xv6's interface: `bread()` returns a locked buffer,
`bwrite()` writes it, `bdirty()` marks it for later, `brelse()` releases. One
sleeplock per buffer and an LRU list for eviction.

**It grows with the machine.** It starts as the 256 buffers in the kernel
image (1 MB), and a miss with every buffer busy makes a new one instead of
evicting, until the cache is 1/32 of RAM, at most 256 MB: 32 MB on 1 GB, 256 MB
from 8 GB up. It never shrinks, so the fraction is kept small. Lookups go
through a hash on `(dev, blockno)`, not the LRU walk, so a 64k-buffer cache
costs no more per `bread()` than a 256-buffer one. `/proc/fsstat` shows
`buf_cached` as now / most.

On a 2 GB machine under QEMU, reading a 40 MB file back went from 1.83 s to
0.89 s, and `ls -R /` and the big binaries are each served from the cache on the
second run instead of from the disk.

**Sleeping is the point.** A buffer lock is held across a disk read, which is
why every path that touches a filesystem must be able to sleep — and why the
root mount happens in `forkret()` rather than in `main()`.

## Mounts

`mount()` and `umount2()` at their Linux numbers, with a mount table consulted
during path resolution. `/proc` is mounted by init:

```c
mount("proc", "/proc", "proc", 0, 0);
```

The filesystem types are `proc`, `devtmpfs`, `sysfs`, `schfs` and `vfat` (also
accepted as `msdos` and `fat`). Unmounting a FAT volume while anything on it is open, or
is some process's working directory, fails with `EBUSY`.

## procfs

`kernel/fs/procfs.c`. A real filesystem behind the same vtable, not a set of
character devices.

**The reason it had to become one is `ps`.** /proc used to be `mknod`'d nodes —
init ran `mknod("/proc/uptime", 6, 2)` and friends, so every file was a real
inode on the root filesystem with a major/minor behind it. That is fine for a
fixed list and cannot work at all for the thing /proc is mostly for: **one
directory per process**. You cannot `mknod` a directory for a process that does
not exist yet and will be gone in a minute. `ps` listed `/proc`, got no numbers,
and reported nothing. It was never a `ps` problem.

Nothing is stored. Directory listings are generated from the process table when
asked; file contents when read.

Three things worth knowing:

- **The split is deliberate.** `kernel/fs/procfs.c` owns the *namespace* — what
  exists. `kernel/dev/procfs.c` keeps the *generators* — what the global files
  say. They change for different reasons, and `proc_gen_global()` is the only
  thing that crosses between them.
- **procfs never walks the process table.** `proc_snapshot()` copies what is
  needed out from under `ptable.lock`, so a process that exits between the
  readdir that listed it and the open that reads it produces a stale line rather
  than a use-after-free.
- **Sizes are reported as zero**, exactly as Linux does. A synthetic file has no
  length until it is generated, and generating every one on every `stat()` to
  fill in a number almost nothing reads would be absurd.

`/proc/mounts`, `/proc/filesystems` and `/proc/partitions` are in Linux's
formats to the column, since what reads them (`df`, `mount`, `fdisk -l`,
installer scripts) parses them. Every disk and partition in `/proc/partitions`
has major 8 and its device number as the minor, matching `/dev`.

## devtmpfs

`/dev` is generated, not stored. `kernel/fs/devfs.c`, behind the same `fs_ops`
vtable, mounted by init with `mount("dev", "/dev", "devtmpfs", 0, 0)`.

It used to be twenty `mknod()` calls in `cmd/schinit`, which was wrong twice
over:

**The cost.** Every node was a real inode on the root filesystem, and SchFS
commits each mutation as its own transaction group — a copy-on-write rewrite of
the path to the root, then an uberblock behind a cache flush. Measured at
**2600 ms of a first boot**, spent before init printed its first line. On later
boots the nodes already existed and the same code took 0 ms, so the cost was
invisible to anyone not building fresh images — which is to say, invisible to
everyone except whoever works on the system.

**The duplication, which matters more.** Major and minor numbers are the
kernel's: they are what `devsw[]` in `kernel/dev/cdev.c` dispatches on. init
held a second copy of that mapping and nothing checked the two agreed. Adding a
character device meant editing a userspace program, and a wrong minor there
produced a node that opened successfully and did nothing.

The table now sits next to the thing it describes. The namespace is all devfs
provides — reads and writes never reach it, because `fileread`/`filewrite` see
`T_DEV` and dispatch through `devsw` themselves.

It is read-only: `dirlink` and `ialloc` answer `EROFS` rather than pretending.
A node exists because the kernel has the device, or it does not exist.

### What is there in `/dev`

`console`, `null`, `zero`, `full`, `random`, `urandom`, `tty`, `tty0` and
`tty1`..`tty12`, `stdin`, `stdout`, `stderr`, `keyboard`, `mouse`, `fb0`, `dsp`,
`ptmx` and a `pts/` directory of pseudo-terminal slaves, `hda`, `hdb`,
`input/event0`.., and the SchOS-specific `pci` (a device's registers, DMA and
interrupts, for a driver in a process), `netdev` (a network interface for a
process), `dbg` and `dbgro` (the kernel debugger's views).

Every disk and partition the block layer registered is there too, under the
name it was given (`sata0`, `sata0p2`, `nvme0p1`, `usb0p1`): major 8, minor its
block device number, listed only while it is present. Reads and writes go
through `blkdev_xfer` in whole sectors, end at the end of the device, and do
not pass through the buffer cache. A partition need not be a whole number of
4 KB blocks, and the buffer cache's last block of one would run into the next. Minors under major 1 follow Linux's
mem-major layout where it costs nothing.

The raw disks are created whether or not a drive is present: opening one with
nothing behind it fails with `ENXIO`, which is a better answer than a missing
path — and with no node at all, `echo x > /dev/hdb` silently creates an ordinary
file of that name and appears to work.

### What is there in `/proc`

`/proc/self`, and one directory per process holding `stat`, `status`, `cmdline`,
`statm`, `comm`, `schedstat`, `maps`, `exe` (a symlink), `fd/` (a symlink per open
descriptor) and `task/` (one directory per thread). `maps` lists the file
mappings, and a process killed by a fault is told which file and offset it died
in.

The global files: `meminfo`, `uptime`, `cpuinfo`, `version`, `stat`, `loadavg`,
`mounts`, `filesystems`, `partitions`, `schfs`, `pci`, `interrupts`, `devices`,
`netif`, `resolv.conf`, `slab`, `fsstat`, `kmsg`, `strace`, `bootmode`,
`bootdebug`, `quiet`, `verbose`, `schinit` and `sysrq`.

**/proc is read-only except `sysrq`**, which is the one deliberate write target,
and the code says so in one place. Marking every procfs inode read-only once
broke `echo s > /proc/sysrq` and took out three USB checks within a minute —
they drive the disk through sysrq.

`sysrq` keys: `c` crash on purpose, `u` / `s` storage checks, `d` display
selftest, `t` text specimen, `n` the netdev ARP selftest.

---


## sysfs

`/sys`, in Linux's layout for the parts that exist here. `kernel/fs/sysfs.c`,
mounted by init with `mount("sysfs", "/sys", "sysfs", 0, 0)`.

| path | contents |
|------|----------|
| `/sys/devices/pci0000:00/0000:BB:DD.F/` (and `/sys/bus/pci/devices/` links to it) | `vendor device class revision subsystem_vendor subsystem_device irq resource config uevent subsystem driver` |
| `/sys/devices/pci0000:00/<xHCI>/usbB/B-P[.P...]/` (and `/sys/bus/usb/devices/` links) | `idVendor idProduct bcdDevice busnum devnum speed bDeviceClass bDeviceSubClass bDeviceProtocol bNumConfigurations uevent subsystem driver` |
| `/sys/firmware/acpi/tables/` | each ACPI table, raw; a repeated signature is numbered from 1 (`SSDT1`, `SSDT2`) |
| `/sys/firmware/dmi/tables/` | `smbios_entry_point` and `DMI`, raw |
| `/sys/firmware/memmap/N/` | `start end type`, the firmware's map as the loader handed it over |
| `/sys/firmware/efi/` | present when UEFI booted the kernel; empty |
| `/sys/class/dmi/id/` | `sys_vendor product_name product_version board_name bios_version`, and the rest Linux has, except serials and the UUID |
| `/sys/class/power_supply/` | `BAT0` (`capacity`, `status`, `energy_*`/`charge_*`, `voltage_now`, `cycle_count`, ...) and `AC` (`online`), from ACPI — the battery and the adapter are read through the laptop's embedded controller |
| `/sys/bus/pci/devices/…/` | also `boot_vga`, `enable` and `graphics/fb0`, which libpciaccess (and so Xorg) reads |

**Linux's layout because everything that reads hardware reads it from there.**
toybox's `lspci` and `lsusb` walk `/sys/bus/*/devices` and parse each `uevent`
(`DRIVER=`, `PCI_ID=`, `PRODUCT=`, `BUSNUM=`), so with these files they work
unmodified, with names from `/usr/share/hwdata/pci.ids` and `usb.ids`. Formats
follow Linux to the character: `0x8086` in `vendor`, `%04x` in `idVendor`,
`PCI_CLASS` in upper case.

Like `/proc`, nothing is stored: a listing comes from the live PCI list, the
USB core's table of attached devices, the ACPI root table and the SMBIOS table
at the moment it is read, so a stick shows up in `/sys/bus/usb/devices` as soon
as it is attached. Text files report size 0, as `/proc`'s do; binary ones (the
tables, `config`) report their real length, since tools copy them by it.

The kernel's own devices (input, tty, mem, graphics, block) have the Linux tree:
`/sys/devices/virtual/<class>/<name>` with `dev`, `uevent` and a `subsystem`
link, `/sys/class/<class>/<name>` and `/sys/dev/{char,block}/MAJ:MIN` as symlinks
into it, and input capabilities; `uevent` is writable, and a write is announced
on the uevent netlink socket ([udev.md](udev.md)).

PCI and USB devices are in that tree too, under `/sys/devices/pci0000:00`, with
the `bus/*/devices` entries as links into it and `add`/`remove` events for USB
hot-plug ([udev.md](udev.md)).

What is missing, compared with Linux: PCI bridges as parents, USB interfaces,
`/sys/class` entries for most devices, `/sys/block`, and most of what is writable. PCI `config` is
256 bytes, because config space is read through port `0xCF8`, which reaches
no further.

The machine's identity comes from SMBIOS (`kernel/dev/dmi.c`): the UEFI loader
hands over the entry point from the firmware's configuration table, and the
kernel falls back to scanning `0xF0000`-`0xFFFFF`. The boot log carries Linux's
line for it, `DMI: <vendor> <product>/<board>, BIOS <version> <date>`.

## Errno, and why `return -1` is a bug

A syscall that fails returns a negative errno. `-1` is therefore **`-EPERM`**,
"operation not permitted" — and about ninety error paths in `kernel/sys/sys_fs.c`
still say `return -1` because that is what xv6 did, where the convention was
just "negative means no".

It is not a cosmetic difference. Callers branch on `errno`, and the standard
shape for removing a file you are not sure about is:

```c
if (unlink(p) < 0 && errno != ENOENT)
    complain();
```

Against `-1` every such test takes the complaining branch. That is how it was
found — zsh, saving its history, reported failures for things that had
succeeded:

```
zsh: locking failed for /root/.zsh_history: operation not permitted
zsh: failed to write history file /root/.zsh_history.new: operation not permitted
```

`link`, `unlink`, `rename` and `renameat` return real values now — `ENOENT`,
`EEXIST`, `EXDEV`, `EISDIR`, `ENOTEMPTY`, `EFAULT`. **The rest of the file does
not**, and that is the honest state: these four were corrected because these
four were the ones a program tripped over.

`symlink()` joined them later, and for the same reason: `symlink()` is the
classic atomic lock — create the link, and if it exists someone else holds it —
so zsh distinguishes `EEXIST` ("contended, retry") from anything else ("this
cannot work, give up"). A bare `-1` is `-EPERM`, which put every shell on the
machine in the second bucket on a filesystem that supports symlinks perfectly
well, and printed the first line above on every boot.

### rename() replaces its target, and why it did not

`rename()` used to refuse an existing destination, which broke
write-to-a-temporary-then-rename — how every careful program saves a file, and
how zsh saves history. The second save of anything failed with `EEXIST` and left
a `.zsh_history.new` behind for ever.

The first attempt to fix it was reverted after it lost both files, and the note
left behind blamed HTree slot reuse. That was the wrong suspect. **`sys_rename`
was removing the old name by hand** — `memset` a `struct dirent` to zero and
`writei()` it into the parent directory at the offset `dirlookup` returned. That
is the v1 filesystem's on-disk layout, written directly, bypassing `dp->ops`.
On SchFS a directory is a B-tree with an index over it: "the record at that
offset" is not a thing, and writing zeroes there corrupted the directory while
leaving the name in the index. Adding a second removal on top of that is what
lost both files.

Every other caller in the file already used `dirunlink()`. This one had been
open-coding a filesystem it no longer ran on, and the bug was invisible while
nothing asked it to remove two names instead of one.

The lock discipline is the delicate part and is inherited rather than invented:
`dp_old` and `dp_new` are frequently the *same* directory reached through two
`nameiparent()` calls, which is two references to one inode. No two directory
locks are ever held at once, and the source inode stays unlocked throughout,
held alive by an `nlink` bump.

Renaming a *directory* is still refused with `EISDIR`, and so is replacing one:
both need the `..` fixup and the ancestor-loop check that are not written.

## Names and paths

A name is at most 255 bytes and a path 4096 including its NUL, Linux's
`NAME_MAX` and `PATH_MAX` (`VFS_NAME_MAX` and `VFS_PATH_MAX` in
`include/fs/file.h`). Longer is `ENAMETOOLONG`. A filesystem may hold less: the v1
format stores `DIRSIZ` bytes and refuses longer names rather than matching on
their first 30.

Both limits used to be the copy buffers: every component went through a
`DIRSIZ` array and every path through a `MAXPATH` one, so a name was cut at 29
bytes and a path at 127, silently. Two files sharing their first 29 bytes were
one file. The path buffers are a page each now, from `kalloc()`, since two of
them would be most of a kernel stack.

## Symlinks

Followed the way Linux follows them, in `namex_from()`. A link in the middle of
a path (`/a/link/b`) always is; the last component only for the calls that
follow (`open`, `stat`, `chdir`, `execve`, `access`, `truncate`, `statfs`,
`utimensat`), not for `lstat`, `readlink`, `unlink`, `rename`, `link` and the
create paths. A relative target resolves from the directory holding the link,
not from the caller's cwd, and a 41st link in one lookup is `ELOOP`.

The walk never holds a directory lock while it reads a link, only a reference
to the directory it is in, so following one is restarting the walk from there
with the target spliced in front of what is left of the path.

`open(O_CREAT)` through a link to an existing file opens the file, so
`echo hi > link` writes where the link points. A dangling link is not created
through.

## What it does not do

- **No page cache.** `mmap` of a file reads the whole thing in at map time, and
  writable `MAP_SHARED` of a file is refused rather than faked. This is the
  single largest gap in the layer.
- **`rename()` does not move directories**, and will not replace one. `EISDIR`
  either way.
- **No named pipes.** `mkfifo()` fails with `EPERM`, Linux's answer for a
  filesystem that cannot hold one. It used to *succeed* and make a character
  device, because `mknod()` read a Linux program's `(mode, dev)` as the native
  `(major, minor)`: `ls -l` showed `crw-rw-rw- 438, 0` where a `p` belongs. The
  zsh port was configured with `zsh_cv_sys_fifo=yes` on the strength of that.
- **No `/sys/devices` and no uevents**, so nothing can be told that a device
  appeared; Xorg is given its input devices by a config file instead.
- **No unified inode cache across filesystems** beyond the single `itable`.
- **`sys_delete`/`sys_show`/`sys_showstat`** are orphaned entries from the old
  v1 filesystem's recycle bin. Nothing reaches them and they die with `fs.c`.
- **One long-standing hang was never found.** Heavy `ls -lR` use across many
  commands in one long shell session once made a *later*, unrelated directory
  hang; the identical command succeeded on a fresh boot. It was instrumented and
  hunted and could not be reproduced, which is recorded in
  [../history/kernel.md](../history/kernel.md) rather than forgotten.