Give userspace a raw disk: /dev/hda and /dev/hdb

Nothing outside the kernel could touch a disk. sys_mount does not even
look at the device it is handed -- it mounts the one SchFS volume found
at boot and ignores `source` -- so a program had no way to read a
partition table, format a volume or write a boot sector. An installer is
a program that needs all three, so this is the first piece of it.

A character device in devsw rather than a separate block layer, because
that is where the offset already lives: fileread/filewrite pass f->off
and advance it by what comes back, which is everything a seekable device
needs. The minor is the IDE device number, the same numbering iderw()
uses; two schemes for the same disks would be a bug waiting to happen.

Writes bypass every filesystem above them, which is inherent to a raw
device and why an installer must target a disk it has not mounted.

The nodes are created unconditionally, like every other entry in /dev.
That matters more than it looks: with no node, `echo x > /dev/hdb`
silently creates an ordinary file of that name and appears to work. It
did, and it agreed with itself on read-back; only the host's copy of the
disk showed nothing had been written.

st_rdev was never filled, so every node in /dev reported "0, 0" under
ls -l. Harmless while nothing read it, wrong as soon as something has to
tell two disks apart. Filled with the full Linux dev_t encoding rather
than (major << 8 | minor), which only works while both stay small.

Verified against the host as an independent oracle, both directions: the
guest wrote to /dev/hdb and the host saw those bytes at offset 0; the
host wrote at offset 4096 and the guest read them back with dd skip=.
ls -l now shows 1,1 for console, 2,0 for fb0 and 8,1 for hdb.

Known limit, commented in disk.c: devsw passes the offset as a 32-bit
uint, so a raw disk is addressable to 4 GB. That must be widened before
anyone installs onto a real disk.

bootcheck 108 [ok] with -Fb -Tty -Usb -Sock -Evdev -Ps.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Makefile

@@ -24,7 +24,7 @@ KERN_CORE = $(addprefix $(BUILD)/kernel/core/, \
               panic.o uaccess.o qr.o time.o)
 
 KERN_DEV  = $(addprefix $(BUILD)/kernel/dev/, \
-              bio.o cdev.o device.o fb.o ide.o input.o irq.o kbd.o mouse.o \
+              bio.o cdev.o device.o disk.o fb.o ide.o input.o irq.o kbd.o mouse.o \
               pci.o text.o ttf.o uart.o vga.o ramdisk.o procfs.o)
 
 # Drivers for hardware that has to be *found* rather than assumed. Anything

cmd/init/main.c

@@ -56,6 +56,16 @@ main(void)
     mknod("/dev/input/event0", 7, 0);   // keyboard
     mknod("/dev/input/event1", 7, 1);   // mouse
 
+    // Raw disks (kernel/dev/disk.c). The minor is the IDE device number, so
+    // hda is the disk the kernel booted from and hdb the second channel.
+    //
+    // Created unconditionally, the same way every other node here is: opening
+    // one when no drive is there fails with ENXIO, which is a better answer
+    // than a missing path, and it keeps this list a description of the device
+    // model rather than of one machine's hardware.
+    mknod("/dev/hda",      8,  0);
+    mknod("/dev/hdb",      8,  1);
+
     // --- /proc -----------------------------------------------------------
     // Synthesised status files, major 6 (kernel/dev/procfs.c). These are the
     // global ones; per-process /proc/<pid>/ directories need synthetic

cmd/schinit/main.c

@@ -444,6 +444,18 @@ bootstrap(void)
     mknod("/dev/input/event0", 7, 0);   // keyboard
     mknod("/dev/input/event1", 7, 1);   // mouse
 
+    // Raw disks (kernel/dev/disk.c). The minor is the IDE device number, so
+    // hda is device 0 and hdb device 1 -- the same numbering iderw() uses.
+    //
+    // Created unconditionally, like everything else here: opening one when no
+    // drive is present fails with ENXIO, which is a better answer than a
+    // missing path. It also matters that these exist even when absent -- with
+    // no node, `echo x > /dev/hdb` silently creates an ordinary file of that
+    // name and appears to succeed, which is a confusing way to discover the
+    // node was never made.
+    mknod("/dev/hda",      8,  0);
+    mknod("/dev/hdb",      8,  1);
+
     // /proc is a real filesystem (kernel/fs/procfs.c), not a handful of nodes:
     // it has to be, because one directory per process cannot be created ahead
     // of time. That is what makes ps and top work.

include/defs.h

@@ -404,6 +404,10 @@ void            fb_mapped(void);
 int             fbread(struct inode*, uint, char*, int);
 int             fbwrite(struct inode*, uint, char*, int);
 int             fb_ioctl(int request, addr_t arg);
+// dev/disk.c -- /dev/hdN, raw seekable access to a whole disk. The minor is the
+// IDE device number. Writes bypass every filesystem above them.
+int             diskread(struct inode*, uint, char*, int);
+int             diskwrite(struct inode*, uint, char*, int);
 // drivers/net/e1000 -- raw Ethernet. No protocol stack above it yet, so these
 // are the whole interface: queue a frame, take a frame, and the MAC.
 int             e1000_transmit(const void *frame, int len);

kernel/dev/cdev.c

@@ -104,4 +104,8 @@ struct devsw devsw[] = {
   // /dev/input/eventN (major 7) — raw evdev events, kernel/dev/input.c. The
   // minor is the device index, so event0 is the keyboard and event1 the mouse.
   [7] = { inputread, inputwrite, inputready, inputclose, inputopen },
+  // /dev/hdN (major 8) — raw disks, kernel/dev/disk.c. The minor is the IDE
+  // device number. No `ready` hook: a disk is always ready, which is what
+  // fileread assumes when one is absent.
+  [8] = { diskread,  diskwrite  },
 };

kernel/dev/disk.c

@@ -0,0 +1,100 @@
+// Raw block devices: /dev/hda, /dev/hdb.
+//
+// Until this existed, nothing outside the kernel could touch a disk. sys_mount
+// does not even look at the device it is given -- it mounts the one SchFS
+// volume the kernel found at boot and ignores `source` -- so a program had no
+// way to read a partition table, format a volume, or write a boot sector. An
+// installer is exactly a program that needs all three.
+//
+// This is a character device in the devsw table rather than a separate block
+// layer, because that is where the offset already is: fileread/filewrite pass
+// f->off and advance it by whatever comes back, which is all a seekable device
+// needs. `dd if=/dev/hdb bs=4096 count=1` works, and so does lseek.
+//
+// The minor number is the IDE device number: 0 is the disk the kernel booted
+// from, 1 is the second channel. Same numbering iderw() uses, deliberately --
+// two numbering schemes for the same disks is a bug waiting to be written.
+//
+// WARNING, and the reason this file is short: writes here bypass the
+// filesystem entirely. Writing to a device with a mounted filesystem on it will
+// corrupt that filesystem, because the buffer cache above still believes its
+// own copy. That is inherent to a raw device and not a defect -- but it is why
+// an installer must target a disk it has not mounted.
+
+#include "types.h"
+#include "defs.h"
+#include "param.h"
+#include "spinlock.h"
+#include "sleeplock.h"
+#include "fs.h"
+#include "buf.h"
+#include "file.h"
+#include "errno.h"
+
+// A byte offset arrives as a 32-bit `uint` from the devsw signature, so a raw
+// device is addressable to 4 GB. Larger disks need the signature widened --
+// worth doing before anyone installs onto a real one, and called out here
+// rather than discovered as silent wraparound at the 4 GB mark.
+#define DISK_MAX_BYTES 0xFFFFFFFFULL
+
+int
+diskread(struct inode *ip, uint off, char *dst, int n)
+{
+  int dev = ip->minor;
+
+  if (!ide_disk_present(dev))
+    return -ENXIO;
+  if (n <= 0)
+    return 0;
+
+  int done = 0;
+  while (done < n) {
+    uint64 pos   = (uint64)off + (uint64)done;
+    uint   blk   = (uint)(pos / BSIZE);
+    uint   inblk = (uint)(pos % BSIZE);
+    int    want  = BSIZE - (int)inblk;
+    if (want > n - done)
+      want = n - done;
+
+    struct buf *b = bread((uint)dev, blk);
+    memmove(dst + done, b->data + inblk, want);
+    brelse(b);
+
+    done += want;
+  }
+  return done;
+}
+
+int
+diskwrite(struct inode *ip, uint off, char *src, int n)
+{
+  int dev = ip->minor;
+
+  if (!ide_disk_present(dev))
+    return -ENXIO;
+  if (n <= 0)
+    return 0;
+
+  int done = 0;
+  while (done < n) {
+    uint64 pos   = (uint64)off + (uint64)done;
+    uint   blk   = (uint)(pos / BSIZE);
+    uint   inblk = (uint)(pos % BSIZE);
+    int    want  = BSIZE - (int)inblk;
+    if (want > n - done)
+      want = n - done;
+
+    // Read-modify-write for anything that is not a whole block. bread() first
+    // even when the whole block is being replaced would be a wasted revolution,
+    // but distinguishing the two cases means trusting the caller's alignment;
+    // the disk is not the bottleneck in an installer and correctness here is
+    // worth more than the read.
+    struct buf *b = bread((uint)dev, blk);
+    memmove(b->data + inblk, src + done, want);
+    bwrite(b);
+    brelse(b);
+
+    done += want;
+  }
+  return done;
+}

kernel/sys/sys_fs.c

@@ -47,6 +47,17 @@ struct linux_stat {
 #define LINUX_S_IFREG  0100000
 #define LINUX_S_IFLNK  0120000
 
+// Linux's dev_t: 12 bits of major and 8 of minor packed into the low 32, with
+// the overflow of each above that. Spelled out in full rather than as
+// (major << 8 | minor), because that is how the other side decodes it and a
+// shortcut here would only work while both numbers stayed small.
+static uint64
+linux_makedev(uint32 major, uint32 minor)
+{
+    return ((uint64)(major & 0xfffu) << 8) | (uint64)(minor & 0xffu)
+         | ((uint64)(major & ~0xfffu) << 32) | ((uint64)(minor & ~0xffu) << 12);
+}
+
 static void
 stati_linux(struct inode *ip, struct linux_stat *st)
 {
@@ -74,6 +85,12 @@ stati_linux(struct inode *ip, struct linux_stat *st)
     }
     st->st_mode = type_bits | perm_bits;
 
+    // Which device this node actually is. Left at zero until now, so every
+    // entry in /dev printed as "0, 0" under ls -l -- harmless while nothing
+    // read it, and wrong the moment something has to tell two disks apart.
+    if (ip->type == T_DEV)
+        st->st_rdev = linux_makedev(ip->major, ip->minor);
+
     st->st_atime = ip->atime; st->st_mtime = ip->mtime; st->st_ctime = ip->ctime;
 }