SchOS/kernel/dev/procfs.c

// /proc — synthesised kernel status files.
//
// Each file is a character device node under major PROCFS_MAJOR, distinguished
// by its minor number, created by init at boot. Reading one renders the whole
// text into a scratch buffer and returns the slice starting at the caller's
// file offset, so a normal read-until-zero loop terminates.
//
// SCOPE: these are the *global* files. Per-process directories (/proc/<pid>/…)
// are not here, because they need something this kernel does not have yet:
// synthetic directories, i.e. a directory whose entries are generated at
// lookup and getdents time rather than read from disk. That is a real VFS
// feature touching iget/ilock/dirlookup/getdents, and it is the natural
// companion to the SchFS work rather than a bolt-on to the current fs.c.
// The namespace (which files exist, including one directory per process) is
// kernel/fs/procfs.c; this file only generates the *content* of the global
// files. `ps` and `top` work through the former.

#include "types.h"
#include "defs.h"
#include "klog.h"
#include "irq.h"
#include "x86.h"
#include "device.h"
#include "display.h"
#include "netdev.h"
#include "param.h"
#include "memlayout.h"
#include "mmu.h"
#include "spinlock.h"
#include "sleeplock.h"
#include "fs.h"
#include "file.h"
#include "proc.h"
#include "uaccess.h"
#include "errno.h"
#include "fsstat.h"
#include "schfs.h"
#include "sbuf.h"
#include "dbg.h"

#define PROC_BUFSZ 2048

// The formatting helpers these files are built out of now live in
// include/sbuf.h, because the debugger's introspection registry needs exactly
// the same thing and two copies would have drifted. See docs/architecture/debugging.md.

// --- the files -------------------------------------------------------------

// PHYSTOP is the top of usable physical memory; kalloc hands out everything
// above the kernel image. Reporting the total honestly and leaving "free" as a
// share of it is better than the invented constants sysinfo() used to return.
#define TOTAL_KB   (PHYSTOP / 1024)

static void
gen_meminfo(struct sbuf *sb)
{
  uint64 total = TOTAL_KB;
  uint64 freek = kalloc_free_bytes() / 1024;
  sb_kv(sb, "MemTotal:       ", total, "kB");
  sb_kv(sb, "MemFree:        ", freek, "kB");
  sb_kv(sb, "MemAvailable:   ", freek, "kB");
  sb_kv(sb, "Buffers:        ", 0, "kB");
  sb_kv(sb, "Cached:         ", 0, "kB");
  sb_kv(sb, "SwapTotal:      ", 0, "kB");
  sb_kv(sb, "SwapFree:       ", 0, "kB");
}


void gen_schinit(struct sbuf *sb);   /* defined at the end of this file */

static void
gen_uptime(struct sbuf *sb)
{
  // From the TSC, not from the tick count -- see monotonic() in
  // kernel/core/time.c. A tick lost while interrupts were disabled is lost
  // forever, so anything derived from ticks runs slow under load and never
  // catches up. This is the number people check when they suspect that.
  uint64 sec, usec;
  monotonic(&sec, &usec);
  sb_putu(sb, (uint)sec);
  sb_putc(sb, '.');
  sb_putu(sb, (uint)(usec / 100000));
  sb_putu(sb, (uint)((usec / 10000) % 10));
  sb_puts(sb, " 0.00\n");       // idle time: not tracked
}

static void
gen_cpuinfo(struct sbuf *sb)
{
  // Online CPUs, not present ones -- the same rule Linux follows, and the
  // reason matters: something reading this to size a thread pool would
  // otherwise be told a VM with four processors has four to schedule on, when
  // three of them are enumerated and idle. "cpu cores" stays the present count,
  // which is what that field means.
  // Real values, from CPUID. These used to read "SchOS" and "SchOS virtual
  // CPU" -- a thing to print when you have not asked the processor, and
  // actively unhelpful, because identifying the machine is the one question
  // /proc/cpuinfo exists to answer.
  //
  // Leaf 0 returns the vendor string in EBX:EDX:ECX, in that order, which is
  // not the order anyone expects. Leaves 0x80000002..4 hold the 48-character
  // brand string. Leaf 0x80000000 reports the highest extended leaf, and asking
  // for one the CPU does not implement returns the highest supported leaf's
  // value rather than an error -- so checking first is not optional.
  char vendor[13], brand[49];
  uint32 a, b, c, d;

  cpuid_leaf(0, &a, &b, &c, &d);
  *(uint32 *)&vendor[0] = b;
  *(uint32 *)&vendor[4] = d;
  *(uint32 *)&vendor[8] = c;
  vendor[12] = '\0';

  brand[0] = '\0';
  cpuid_leaf(0x80000000, &a, 0, 0, 0);
  if (a >= 0x80000004) {
    for (int leaf = 0; leaf < 3; leaf++) {
      cpuid_leaf(0x80000002 + leaf, &a, &b, &c, &d);
      *(uint32 *)&brand[leaf * 16 + 0]  = a;
      *(uint32 *)&brand[leaf * 16 + 4]  = b;
      *(uint32 *)&brand[leaf * 16 + 8]  = c;
      *(uint32 *)&brand[leaf * 16 + 12] = d;
    }
    brand[48] = '\0';
  }

  // Intel pads the brand string with leading spaces, which read as a broken
  // field rather than a name.
  const char *model = brand;
  while (*model == ' ')
    model++;
  if (*model == '\0')
    model = "unknown";

  for (int i = 0; i < ncpu_online; i++) {
    sb_kv(sb, "processor\t: ", i, 0);
    sb_puts(sb, "vendor_id\t: ");
    sb_puts(sb, vendor);
    sb_putc(sb, '\n');
    sb_puts(sb, "model name\t: ");
    sb_puts(sb, model);
    sb_putc(sb, '\n');
    sb_kv(sb, "cpu cores\t: ", ncpu, 0);
    sb_putc(sb, '\n');
  }
  if (ncpu_online < ncpu) {
    sb_kv(sb, "cpus present\t: ", ncpu, 0);
    sb_puts(sb, "note\t\t: additional processors are enumerated but not "
                "started; see docs/plan/roadmap.md\n");
  }
}

static void
gen_version(struct sbuf *sb)
{
  // The kernel's name, not the system's -- same as "Linux version 6.x" here.
  sb_puts(sb, "SchIFT version 0.1.0 (schift) #1 SMP x86_64\n");
}

static void
gen_stat(struct sbuf *sb)
{
  // USER_HZ, like times() -- everything that parses /proc/stat divides by 100.
  uint t = (uint)((uint64)uptime() * USER_HZ / HZ);
  sb_puts(sb, "cpu  ");
  sb_putu(sb, t); sb_puts(sb, " 0 0 0 0 0 0 0 0 0\n");
  // Online, for the same reason as cpuinfo: top and friends count the cpuN
  // lines here to decide how many cores to draw.
  for (int i = 0; i < ncpu_online; i++) {
    sb_puts(sb, "cpu");
    sb_putu(sb, i);
    sb_putc(sb, ' ');
    sb_putu(sb, t); sb_puts(sb, " 0 0 0 0 0 0 0 0 0\n");
  }
  sb_kv(sb, "processes ", (uint64)proc_num(), 0);
  sb_kv(sb, "procs_running ", 1, 0);
  sb_kv(sb, "btime ", 0, 0);
}

// Everything attached, whatever bus found it. The one place that answers "what
// is in this machine" across PCI, USB and platform devices at once -- which is
// the question a device manager is, so this is where one would read from.
//
// Columns rather than key/value: this is a table of like things, and a reader
// scanning for an unclaimed device wants the driver column to line up.
static void
gen_devices(struct sbuf *sb)
{
  sb_puts(sb, "name          bus       class     driver        id        device\n");
  for (int i = 0; i < device_count(); i++) {
    struct device *d = device_at(i);
    if (!d)
      break;
    // 14 for the name and driver columns: "virtio-gpu0" is eleven characters
    // and overran a ten-wide column, which sb_pad allows on purpose but which
    // made the whole table unreadable.
    sb_pad(sb, d->name, 14);
    sb_pad(sb, dev_bus_name(d->bus), 10);
    sb_pad(sb, dev_class_name(d->class), 10);
    // An unclaimed device is the interesting row, so it gets a visible marker
    // rather than an empty column.
    sb_pad(sb, d->driver ? d->driver : "(none)", 14);

    // vendor:product, the way lspci and lsusb print it. The kernel has no
    // business resolving these to names -- that is what /usr/share/usb.ids is
    // for, and it belongs to a userspace tool. What the kernel owes is the
    // number, in the form a lookup expects.
    if (d->vendor || d->product) {
      sb_puthex4(sb, d->vendor);
      sb_putc(sb, ':');
      sb_puthex4(sb, d->product);
      sb_putc(sb, ' ');
    } else {
      sb_puts(sb, "          ");
    }
    sb_puts(sb, d->desc[0] ? d->desc : "-");
    sb_putc(sb, '\n');
  }
}

static void
gen_loadavg(struct sbuf *sb)
{
  // No load tracking; report zeros but a truthful process count.
  sb_puts(sb, "0.00 0.00 0.00 1/");
  sb_putu(sb, (uint64)proc_num());
  sb_puts(sb, " 1\n");
}

// df and mountpoint parse this.
//
// It used to be three fixed lines, which was true while there was one
// filesystem and no way to mount a second. Now `mount -t schfs /dev/hdb /mnt`
// works, and a mount missing from here is invisible to df -- it printed a
// header and no row, because df looks the path up in this list before it calls
// statfs at all.
static void
gen_mounts(struct sbuf *sb)
{
  char path[MAXPATH], source[MAXPATH], type[16];
  uint fsdev;

  sb_puts(sb, "devtmpfs /dev devtmpfs rw 0 0\n");

  for (int i = 0; i < 8; i++) {
    if (mount_get(i, path, source, type, &fsdev) != 0)
      continue;
    // "/" is emitted last, below, so skip it if it somehow appears here.
    if (path[0] == '/' && path[1] == '\0')
      continue;
    sb_puts(sb, source[0] ? source : "none");
    sb_puts(sb, " ");
    sb_puts(sb, path);
    sb_puts(sb, " ");
    sb_puts(sb, type[0] ? type : "schfs");
    // A mounted snapshot is genuinely read-only, and saying so is what stops
    // anything trying to write to one and reporting a confusing EROFS later.
    sb_puts(sb, SCHFS_IS_SNAP_DEV(fsdev) ? " ro 0 0\n" : " rw 0 0\n");
  }

  // Root must be last: df keeps the last mount point that prefixes the path
  // it was asked about, so listing / first made `df /` report /proc instead.
  sb_puts(sb, "/dev/root / schfs rw 0 0\n");
}

// Where the time goes in the filesystem and block paths.
//
// Cycles, not seconds: the numbers are only ever compared against each other,
// and converting would mean knowing the TSC frequency, which this kernel does
// not. What matters is the ratios and the miss counts.
static void
gen_fsstat(struct sbuf *sb)
{
  sb_puts(sb, "bread_calls      "); sb_putu(sb, fsstat.bread_calls);
  sb_puts(sb, "\nbread_miss       "); sb_putu(sb, fsstat.bread_miss);
  sb_puts(sb, "\nbget_evict_dirty "); sb_putu(sb, fsstat.bget_evict_dirty);
  sb_puts(sb, "\ndisk_reads       "); sb_putu(sb, fsstat.disk_reads);
  sb_puts(sb, "\ndisk_writes      "); sb_putu(sb, fsstat.disk_writes);
  sb_puts(sb, "\ndisk_kcycles     "); sb_putu(sb, fsstat.disk_cycles / 1000);
  sb_puts(sb, "\nread_calls       "); sb_putu(sb, fsstat.read_calls);
  sb_puts(sb, "\nread_bytes       "); sb_putu(sb, fsstat.read_bytes);
  sb_puts(sb, "\nread_kcycles     "); sb_putu(sb, fsstat.read_cycles / 1000);
  sb_puts(sb, "\ngetdents_calls   "); sb_putu(sb, fsstat.getdents_calls);
  sb_puts(sb, "\ngetdents_walked  "); sb_putu(sb, fsstat.getdents_walked);
  sb_puts(sb, "\ngetdents_kcycles "); sb_putu(sb, fsstat.getdents_cycles / 1000);
  sb_puts(sb, "\ndescents         "); sb_putu(sb, fsstat.descents);
  sb_puts(sb, "\nexec_calls       "); sb_putu(sb, fsstat.exec_calls);
  sb_puts(sb, "\nexec_kcycles     "); sb_putu(sb, fsstat.exec_cycles / 1000);

  // Occupancy. The peaks are the diagnostic: a peak that climbs across a
  // session and never falls back after the commands finish is a reference
  // leak, and which line it is on says where to look.
  sb_puts(sb, "\ninode_used       "); sb_putu(sb, fsstat.inode_used);
  sb_puts(sb, "\ninode_peak       "); sb_putu(sb, fsstat.inode_peak);
  sb_puts(sb, " / "); sb_putu(sb, NINODE);
  sb_puts(sb, "\nfile_used        "); sb_putu(sb, fsstat.file_used);
  sb_puts(sb, "\nfile_peak        "); sb_putu(sb, fsstat.file_peak);
  sb_puts(sb, " / "); sb_putu(sb, NFILE);
  sb_puts(sb, "\nbuf_pinned       "); sb_putu(sb, fsstat.buf_pinned);
  sb_puts(sb, "\nbuf_pinned_peak  "); sb_putu(sb, fsstat.buf_pinned_peak);
  sb_puts(sb, " / "); sb_putu(sb, NBUF);

  // The SchFS bracket. `locked 1` with `owner 0` is the wedge: the volume lock
  // is held by something that is not in a transaction, so nothing will ever
  // release it and every later filesystem call sleeps in acquiresleep.
  for (int i = 0; i < SCHFS_VOL_SLOTS; i++) {
    int locked, owner, depth;
    schfs_lockinfo(i, &locked, &owner, &depth);
    if (locked < 0)
      continue;
    sb_puts(sb, "\nschfs"); sb_putu(sb, (uint64)i);
    sb_puts(sb, "  locked "); sb_putu(sb, (uint64)locked);
    sb_puts(sb, "  tx_owner_pid "); sb_putu(sb, (uint64)owner);
    sb_puts(sb, "  tx_depth "); sb_putu(sb, (uint64)depth);
  }
  sb_puts(sb, "\n");
}

static void
gen_filesystems(struct sbuf *sb)
{
  sb_puts(sb, "\tschfs\n");
  sb_puts(sb, "nodev\tproc\n");
  sb_puts(sb, "nodev\tdevtmpfs\n");
}

// --- dispatch --------------------------------------------------------------

#define PROC_MEMINFO      1
#define PROC_UPTIME       2
#define PROC_CPUINFO      3
#define PROC_VERSION      4
#define PROC_STAT         5
#define PROC_LOADAVG      6
#define PROC_MOUNTS       7
#define PROC_FILESYSTEMS  8
#define PROC_SCHFS        9
#define PROC_SYSRQ        10
#define PROC_PCI          11
#define PROC_INTERRUPTS   12
#define PROC_DEVICES      13
#define PROC_FSSTAT       14
#define PROC_NETIF        15
#define PROC_KMSG         17
#define PROC_RESOLV       18

// Render global file `id` (a PROC_* constant) into `buf`. Returns its length.
//
// Split out of procread() so the /proc *filesystem* (kernel/fs/procfs.c) can
// use the same generators: that file owns the namespace, this one owns the
// content, and they have no other business with each other.
int
proc_gen_global(int id, char *buf, int cap)
{
  struct sbuf sb = { buf, 0, cap };

  switch (id) {
  case PROC_MEMINFO:     gen_meminfo(&sb);     break;
  case PROC_UPTIME:      gen_uptime(&sb);      break;
  case PROC_CPUINFO:     gen_cpuinfo(&sb);     break;
  case PROC_VERSION:     gen_version(&sb);     break;
  case PROC_STAT:        gen_stat(&sb);        break;
  case PROC_LOADAVG:     gen_loadavg(&sb);     break;
  case PROC_MOUNTS:      gen_mounts(&sb);      break;
  case PROC_FSSTAT:      gen_fsstat(&sb);      break;
  case PROC_FILESYSTEMS: gen_filesystems(&sb); break;
  // SchFS renders straight into the buffer rather than through sbuf, because
  // it lives in kernel/fs/schfs and must not depend on this file's helpers.
  case PROC_SCHFS:       sb.len = schfs_kinfo(buf, PROC_BUFSZ); break;
  // Same arrangement as SchFS above: pci.c renders its own lines so that it
  // does not depend on this file's helpers.
  case PROC_PCI:         sb.len = pci_info(buf, PROC_BUFSZ); break;
  // Likewise irq.c: it owns the table, so it renders it.
  case PROC_INTERRUPTS:  sb.len = irq_info(buf, PROC_BUFSZ); break;
  case 16:               gen_schinit(&sb);       break;
  // And likewise net.c, which is the only file here allowed to include lwIP's
  // headers -- reaching into the stack from this one would put them on every
  // kernel file's include path.
  case PROC_NETIF:       sb.len = net_info(buf, PROC_BUFSZ); break;
  case PROC_RESOLV:      sb.len = net_resolv(buf, PROC_BUFSZ); break;
  case PROC_DEVICES:     gen_devices(&sb);     break;
  default:               return 0;
  }
  return sb.len;
}

int
procread(struct inode *ip, uint64 off, char *dst, int n)
{
  // The log is served straight out of its ring rather than regenerated into a
  // scratch buffer like everything else here: it is 128 KB, and this function
  // renders the whole file on every read into an array on a 4 KB kernel stack.
  // See klog_read().
  if (ip->minor == PROC_KMSG)
    return klog_read(off, dst, n);

  char buf[PROC_BUFSZ];
  int len = proc_gen_global(ip->minor, buf, PROC_BUFSZ);

  if (len <= 0 || (int)off >= len)
    return 0;                       // EOF
  int avail = len - (int)off;
  if (n > avail)
    n = avail;
  memmove(dst, buf + off, n);
  return n;
}

// /proc is read-only, with one deliberate exception.
//
// /proc/sysrq is the same idea as Linux's sysrq-trigger: a way to make the
// kernel do something drastic on purpose. Right now the only command is 'c',
// crash — which exists because the crash screen is otherwise the one part of
// the system that can only be tested by having a bug. A screen you cannot look
// at until something goes wrong is a screen nobody has ever actually read.
// One sysrq command byte. Shared by the device path (procwrite, below) and
// the filesystem path (kernel/fs/procfs.c), so the two cannot drift.
static void
proc_sysrq_cmd(char cmd)
{
  switch (cmd) {
  case 'c':
    panic("sysrq: crash requested from /proc/sysrq");
  case 'u':
    xhci_dump();
    break;
  case 's':
    usb_storage_dump();
    break;
  case 'd':
    display_selftest();
    break;
  case 'n':
    netdev_selftest();
    break;
  case 't':
    text_selftest();
    break;
  case 'i':
    // Kill PID 1, on purpose.
    //
    // Signals cannot do this -- init_ignores() drops them, which is correct
    // and is why the two-init scheme was solving a problem that did not exist.
    // Setting `killed` directly is the kernel deciding, not userspace asking:
    // init exits at its next trap return and proc_reincarnate() takes over.
    //
    // docs/architecture/init.md asks for this to be "tested by doing it rather than
    // by reasoning", which is the only honest way to know the path works.
    proc_kill_init();
    cprintf("sysrq: pid 1 asked to exit -- it should come back\n");
    break;

  case 'g':
    // Every registered view, straight to the console.
    //
    // This is docs/architecture/debugging.md's level 0: no process, no filesystem, no
    // client program. If /dev/dbg cannot be opened -- because the filesystem
    // is what is broken -- this still works, and on a machine with only a
    // serial line it is the whole debugger.
    dbg_dump_all();
    break;

  case 'w': {
    // Wedge the supervisor: make schinitd hang, on purpose, so PID 1's
    // heartbeat timeout is exercised instead of reasoned about. There is no
    // way back from this; reboot afterwards.
    int pid = proc_wedge("schinitd");
    if (pid)
      cprintf("sysrq: wedged schinitd (pid %d) -- pid 1 should notice "
              "within %d seconds\n", pid, 10);
    else
      cprintf("sysrq: no process named schinitd to wedge\n");
    break;
  }
  default:
    cprintf("sysrq: unknown command '%c' "
            "(known: c = crash, u = dump USB, s = read the USB disk, "
            "d = display self-test, t = text self-test, n = net self-test, "
            "g = dump every debug view, "
            "w = wedge the init supervisor, i = kill pid 1)\n",
            cmd);
    break;
  }
}

// `user_src` says whether `src` is a user address that still needs copying in;
// the filesystem hands one over, the device path has already copied.
int
proc_write_sysrq(int user_src, uint64 src, uint n)
{
  char c;

  if (n < 1)
    return -1;
  if (user_src) {
    if (copy_from_user(&c, src, 1) < 0)
      return -EFAULT;
  } else {
    c = *(char *)src;
  }
  proc_sysrq_cmd(c);
  return (int)n;
}

// The character-device path: /proc/sysrq reached through the old mknod'd node
// rather than through the /proc filesystem. `src` is already a kernel buffer
// here, which is why it passes user_src = 0.
int
procwrite(struct inode *ip, uint64 off, char *src, int n)
{
  (void)off;

  if (ip->minor != PROC_SYSRQ)
    return -1;
  return proc_write_sysrq(0, (uint64)src, (uint)n);
}

// ---------------------------------------------------------------------------
// /proc/schinit -- the job table
// ---------------------------------------------------------------------------
//
// docs/architecture/init.md, "State outside the process". PID 1 writes this
// whenever the table changes and
// anyone may read it; a supervisor that has just been restarted reads it back
// and knows what is running without having started any of it.
//
// The kernel holds the text and does not interpret it. That is deliberate: a
// procfs generator runs in the kernel, and walking PID 1's memory to answer a
// read would put this file's format into the kernel, where it would then have
// to be kept in step with a userspace program. Instead the kernel is a
// pinboard -- PID 1 knows the format and knows when it changed.
//
// procfs rather than a file on disk for four reasons: it survives a read-only
// root, it survives a full disk, it costs no I/O, and it is already the
// pattern this kernel uses for exactly this kind of question.

#define SCHINIT_TABLE_MAX 2048

static char schinit_table[SCHINIT_TABLE_MAX];
static int  schinit_table_len;

int
proc_write_schinit(int user_src, uint64 src, uint n)
{
  if (n > SCHINIT_TABLE_MAX)
    n = SCHINIT_TABLE_MAX;

  // A write replaces the table rather than appending to it. There is one
  // writer and it always sends the whole thing, so append would only be a way
  // to accumulate stale rows.
  if (user_src) {
    if (copy_from_user(schinit_table, src, n) < 0)
      return -EFAULT;
  } else {
    memmove(schinit_table, (char *)src, n);
  }
  schinit_table_len = (int)n;
  return (int)n;
}

void
gen_schinit(struct sbuf *sb)
{
  if (schinit_table_len == 0) {
    sb_puts(sb, "generation 0\nsupervisor 0\n");
    return;
  }
  for (int i = 0; i < schinit_table_len; i++)
    sb_putc(sb, schinit_table[i]);
}