SchOS/include/fs/file.h

struct file {
  // FD_MEMFD: an anonymous shared memory object, the thing memfd_create(2)
  // returns. It is a descriptor with no name and no inode behind it, which is
  // the entire reason memfd exists -- two processes share memory by passing
  // the descriptor, not by agreeing on a path or a System V key.
  //
  // FD_EVENT: an eventfd, timerfd, signalfd or epoll set -- see
  // kernel/sys/sys_event.c, which is the only file that looks inside `ev`.
  enum { FD_NONE, FD_PIPE, FD_INODE, FD_SOCKET, FD_MEMFD, FD_EVENT } type;
  int ref; // reference count
  char readable;
  char writable;
  // O_NONBLOCK. A property of the open file description, not of the
  // descriptor, which is why it lives here and not in myproc()->ofile_flags --
  // dup() shares it and F_SETFL on either fd changes it for both, exactly as
  // POSIX says.
  char nonblock;
  struct pipe *pipe;
  struct inode *ip;
  struct socket *sock;
  // The shared segment behind an FD_MEMFD, as an index into the table in
  // kernel/sys/sys_mem.c. Not a pointer, because the segment table is where
  // the reference counting lives and handing out pointers into it would
  // invite somebody to keep one.
  int memfd_seg;
  struct evobj *ev;
  uint64 off;
  // What /proc/<pid>/fd shows. `ino` names the open thing (one per socket,
  // one per pipe); `path` is the absolute path an inode was opened by, as
  // written -- symlinks along it are not resolved. kmalloc'd, or 0.
  uint64 ino;
  char *path;
  struct file *fnext;   // file.c's free list, while ref is 0
};

// ---------------------------------------------------------------------------
// Filesystem operations
//
// One table per filesystem, chosen by iget() from the device number and hung
// off the inode. Everything above this line -- the syscall layer, exec, the
// page cache -- calls the wrappers in fs.c and never learns which filesystem it
// is talking to.
//
// A vtable rather than a chain of `if (dev == SCHFS_DEV)` because there are two
// filesystems today and there will be more: the whole point of the exercise is
// that adding one is a new table, not another branch in fifteen functions.
//
// Every hook takes an inode that is already locked, except ialloc (which has no
// inode yet) and iload (which is what makes one valid in the first place).
// ---------------------------------------------------------------------------

struct stat;

struct fs_ops {
  const char *name;

  // Fill in the type/size/link fields of a freshly iget()'d inode.
  void (*iload)(struct inode *);
  // Write those fields back to disk.
  void (*istore)(struct inode *);
  // The last link is gone and nobody has it open: release the object.
  void (*ifree)(struct inode *);

  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);
  // Set the length. Growing is allowed only where the filesystem can represent
  // the gap; v1 cannot and ignores it.
  void (*truncate)(struct inode *, uint64 newsize);

  // Directory operations. `poff` is an opaque cookie identifying where the
  // entry lives, meaningful only to the same filesystem's dirunlink.
  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);

  // Create an object with no name yet. The caller sets nlink and calls
  // dirlink; until it does, the object is unreachable.
  struct inode *(*ialloc)(uint dev, short type);

  // Read directory entries into a getdents64-shaped stream. `off` is an opaque
  // position that starts at 0 and is echoed back; returns bytes written, 0 at
  // the end of the directory, or negative on error.
  int  (*getdents)(struct inode *dp, char *out, int cap, uint64 *off);

  // The largest write this filesystem wants in one call, or 0 for "no limit".
  //
  // v1 needs one: a write has to fit in the on-disk log, so filewrite splits
  // anything larger and each piece is its own transaction. That limit is a
  // property of the journal, not of writing, and applying it to SchFS was
  // actively harmful -- SchFS makes one transaction out of whatever it is
  // handed, so chopping a 1 MB write into 6 KB pieces bought 170 commits and
  // the flushes that go with them for nothing.
  uint max_write;

  // Object types this filesystem has no way to store, as a mask of 1 << T_*.
  // Creating one fails with EPERM, which is what Linux answers for mknod or a
  // symlink on vfat. 0 for "anything".
  uint refuse_types;
  // One name per file: link() fails with EPERM. Rename still works -- it is
  // a second name for as long as it takes to remove the first.
  int  no_hardlinks;
};

// In-memory inode.
//
// `inum` and `size` are 64-bit because the filesystem underneath may be: SchFS
// object ids and file sizes are both uint64 on disk, and a 32-bit copy here
// would silently truncate them into a different file. The v1 filesystem uses a
// small fraction of the range and is unaffected.
struct inode {
  uint dev;
  uint64 inum;
  int ref;
  struct sleeplock lock;
  int valid;
  const struct fs_ops *ops;   // chosen by iget from dev; never null once held
  struct inode *hnext;        // fs.c: hash chain while held, free list after
  struct inode *anext;        // fs.c: every inode there is, never reordered

  short type;
  short major;
  short minor;
  // int, not short. SchFS stores a uint32 link count and this truncated it, so
  // a file that passed 32767 names came back with a negative one -- and iput
  // frees an object whose count reaches zero. Signed rather than unsigned on
  // purpose: `nlink--` on a count that is somehow already zero should land
  // somewhere a check can see it, not wrap to four billion.
  int nlink;
  uint64 size;

  uint rwmode;
  // POSIX permission bits from a filesystem that keeps them, or 0 for one
  // that does not, in which case stat makes some up from the type. Bit 18 set
  // alone means a mode of 0000, which chmod can ask for.
  uint mode;
  // Owner. Whatever the filesystem keeps: SchFS has stored them all along,
  // FAT and the synthetic filesystems report root.
  uint uid, gid;
  uint supermode;
  uint showmode;
  uint atime;
  uint mtime;
  uint ctime;
  uint addrs[NDIRECT+3];
};
#define I_VALID 0x2

// Device numbers for filesystems that are not a raw disk. Real disks are 0..3,
// and these are deliberately far from them so a stray value never looks like a
// SchFS inode. Nothing on any disk records them -- they are runtime identities
// for the inode cache, so the numbering can be rearranged freely.
//
// Mounted SchFS volumes: SCHFS_DEV through SCHFS_DEV + SCHFS_VOL_SLOTS - 1.
//
// This was a single number, because there was a single volume: kmount.c held
// one `kvol` and every write entry point named it directly, so `mount -t schfs
// /dev/hdb /mnt` ignored its source argument entirely and handed back whichever
// disk the kernel had found at boot. Slot 0 is that boot volume, which is why
// SCHFS_DEV alone still means "the root" everywhere that only ever wanted it.
// Sixteen, which fills 64..79 up to PROCFS_DEV.
#define SCHFS_VOL_SLOTS 16
#define SCHFS_DEV       64
#define SCHFS_IS_VOL_DEV(d) \
    ((d) >= SCHFS_DEV && (d) < SCHFS_DEV + SCHFS_VOL_SLOTS)

// Mounted snapshots. Same block device, an older tree root, read-only.
//
// Each needs its own device number because an inode is identified by
// (dev, inum) and a snapshot's object ids are the *same numbers* as the live
// tree's -- one number would make the views alias in the inode cache and hand
// a reader whichever was cached first. So a slot is a device: SCHFS_SNAP_DEV
// through SCHFS_SNAP_DEV + SCHFS_SNAP_SLOTS - 1.
//
// More than one because "one at a time" is an odd thing to explain and
// the cost is two scratch pages per slot, allocated on first use. Not many
// because each is a device number and they are a fixed, small space.
#define SCHFS_SNAP_SLOTS 8   // 96..103, up to FAT_DEV
#define SCHFS_SNAP_DEV   96
#define SCHFS_IS_SNAP_DEV(d) \
    ((d) >= SCHFS_SNAP_DEV && (d) < SCHFS_SNAP_DEV + SCHFS_SNAP_SLOTS)

// Anything SchFS answers for, live volume or snapshot.
#define SCHFS_IS_ANY_DEV(d) (SCHFS_IS_VOL_DEV(d) || SCHFS_IS_SNAP_DEV(d))

// The synthetic /proc.
#define PROCFS_DEV 80

// The synthetic /dev. See kernel/fs/devfs.c.
#define DEVFS_DEV  81

// The synthetic /sys. See kernel/fs/sysfs.c.
#define SYSFS_DEV  82
// Graft it over `target`; 0 or a negative errno. In sysfs.c.
int sysfs_mount(const char *target);

// The buffer /proc renders a file into, and how big it is; one at a time, so
// every get is followed by a put. In kernel/dev/procfs.c.
char *proc_genbuf_get(int *cap);
void  proc_genbuf_put(void);

// Mounted FAT volumes, FAT_DEV onward. See include/fs/fat.h.

// Linux's limits, for every filesystem behind the VFS. A path component can be
// NAME_MAX bytes and a whole path PATH_MAX including its NUL. Each filesystem
// may refuse less: SchFS and FAT take 255, the v1 format DIRSIZ.
//
// The VFS used to copy every component into a DIRSIZ buffer and every path
// into a MAXPATH one, which cut names at 29 bytes and paths at 127 without a
// word -- so a longer name opened whichever file shared its first 29 bytes.
#define VFS_NAME_MAX 255
#define VFS_PATH_MAX 4096           // one page: namex_from() kalloc()s it

// 1 if some inode on dev with this number is referenced. In fs.c.
int iheld(uint dev, uint64 inum);
// 1 if any inode on dev but `except` is referenced (0: no exception). In fs.c.
int ibusy(uint dev, uint64 except);
// How many in-memory inodes exist; the table grows past NINODE. In fs.c.
uint inode_slots(void);
// How many open file descriptions exist; the table grows past NFILE. file.c.
uint file_slots(void);
// Path lookup that follows a final symlink too; 0 with *err set on failure.
// `start` 0 means the cwd. In fs.c.
struct inode *nameifollow(struct inode *start, char *path, int *err);
// execve(2) with the argv and envp arrays still in the caller's memory. In
// exec.c; exec() is the same for arrays the kernel holds.
int exec_user(char *path, addr_t uargv, addr_t uenvp);

// table mapping major device number to
// device functions
struct devsw {
  int (*read)(struct inode*, uint64, char*, int);
  int (*write)(struct inode*, uint64, char*, int);
  // Optional. Returns 0 when a read would block, which is the only thing
  // fileread needs to know to honour O_NONBLOCK without every device driver
  // growing a flags argument. A device that omits this is treated as always
  // ready, which is the right answer for a disk and for /dev/zero.
  int (*ready)(struct inode*);
  // Optional. The last descriptor for this open file is going away. A device
  // that hands a program exclusive-feeling control of the hardware needs this
  // to put the hardware back -- /dev/fb0 uses it to give the text console its
  // screen back when whatever was drawing on it exits.
  void (*close)(struct inode*);
  // Optional. A descriptor has just been opened on this device. /dev/input
  // uses it to discard events that arrived while nobody was listening -- see
  // inputopen().
  void (*open)(struct inode*);
};

extern struct devsw devsw[];

#define CONSOLE 1
// Major number of /dev/input/eventN. The minor is the device index.
#define INPUT_MAJOR 7
// Major number of /dev/dsp. See include/dev/audio.h.
#define AUDIO_MAJOR 12