From ad2b321d56c3ca0a1d3c48d71b80c5dfdb02fb97 Mon Sep 17 00:00:00 2001
From: Nexoniarz <gamesfrank90@gmail.com>
Date: Tue, 11 Aug 2026 12:15:38 +0200
Subject: [PATCH] dev: raw disks are 64-bit, so a real drive stops wrapping at
 4 GB

struct file already carried a uint64 offset and bread already took a
uint64 block number. The narrowing was one boundary in the middle:
devsw's read and write took a 32-bit uint, so fileread passed a 64-bit
f->off into a 32-bit parameter and the top half was discarded.

The effect was not a refusal at 4 GB but silent wraparound past it. A
read at 5 GB returned the bytes at 1 GB, reporting success. disk.c had a
DISK_MAX_BYTES constant documenting the limit, and nothing anywhere
referenced it -- the cap was described but never enforced, which is the
worst of the three options.

The signature is now uint64 across all nine files that carry it, and
diskread/diskwrite no longer narrow the block number on the way to
bread. Widening this is a prerequisite for NVMe and for any real disk:
namespaces start well above 4 GB and the failure mode reads valid data
from the wrong place.

Verified end to end rather than by inspection. A 6 GB sparse image with
"PAST4GB-MARKER-OK" written at byte 5368709120, attached as /dev/hdb to a
ramdisk-root kernel so device 1 was free:

  # dd if=/dev/hdb bs=4096 skip=1310720 count=1 | od -c | head -2
  0000000   P   A   S   T   4   G   B   -   M   A   R   K   E   R   -   O

Under the old code that offset truncates to 1 GB, where the image is
zeroes. The whole path is exercised: f->off, devsw, diskread, bread and
ide.c's LBA48.

Worth recording for next time: `make ramroot` stages the ESP itself, and
running `make esp` afterwards rebuilds the kernel without RAMROOT and
silently undoes it. The tell is a 855 KB kernel where a 25 MB one is
expected.

110 [ok], no failures, on the default build.
---
 include/defs.h        | 16 ++++++++--------
 include/file.h        |  4 ++--
 kernel/core/console.c |  4 ++--
 kernel/dev/cdev.c     | 22 +++++++++++-----------
 kernel/dev/disk.c     | 23 ++++++++++++-----------
 kernel/dev/fb.c       |  4 ++--
 kernel/dev/input.c    |  4 ++--
 kernel/dev/mouse.c    |  2 +-
 kernel/dev/procfs.c   |  4 ++--
 9 files changed, 42 insertions(+), 41 deletions(-)

diff --git a/include/defs.h b/include/defs.h
index 204e57e..b234a2a 100644
--- a/include/defs.h
+++ b/include/defs.h
@@ -198,8 +198,8 @@ int             schfs_kreaddir_dev(uint dev, uint64 dir,
 uint            uptime(void);
 
 // procfs.c
-int             procread(struct inode*, uint, char*, int);
-int             procwrite(struct inode*, uint, char*, int);
+int             procread(struct inode*, uint64, char*, int);
+int             procwrite(struct inode*, uint64, char*, int);
 // Shared with the /proc filesystem (kernel/fs/procfs.c): one owns the
 // namespace, the other the content.
 int             proc_gen_global(int id, char *buf, int cap);
@@ -208,8 +208,8 @@ void            procfs_mount(void);
 
 // input.c -- raw evdev event devices
 void            inputinit(void);
-int             inputread(struct inode*, uint, char*, int);
-int             inputwrite(struct inode*, uint, char*, int);
+int             inputread(struct inode*, uint64, char*, int);
+int             inputwrite(struct inode*, uint64, char*, int);
 int             inputready(struct inode*);
 void            inputclose(struct inode*);
 void            inputopen(struct inode*);
@@ -431,13 +431,13 @@ void            fbclose(struct inode*);
 // Called from sys_mmap: a mapping of /dev/fb0 is a licence to draw with no
 // further syscalls, so it is the last the kernel hears about it.
 void            fb_mapped(void);
-int             fbread(struct inode*, uint, char*, int);
-int             fbwrite(struct inode*, uint, char*, int);
+int             fbread(struct inode*, uint64, char*, int);
+int             fbwrite(struct inode*, uint64, 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);
+int             diskread(struct inode*, uint64, char*, int);
+int             diskwrite(struct inode*, uint64, 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);
diff --git a/include/file.h b/include/file.h
index ac669c6..7cf76da 100644
--- a/include/file.h
+++ b/include/file.h
@@ -123,8 +123,8 @@ struct inode {
 // table mapping major device number to
 // device functions
 struct devsw {
-  int (*read)(struct inode*, uint, char*, int);
-  int (*write)(struct inode*, uint, char*, int);
+  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
diff --git a/kernel/core/console.c b/kernel/core/console.c
index 6062e24..481eca5 100644
--- a/kernel/core/console.c
+++ b/kernel/core/console.c
@@ -655,7 +655,7 @@ consoleintr(int (*getc)(void))
 }
 
 int
-consoleread(struct inode *ip, uint off, char *dst, int n)
+consoleread(struct inode *ip, uint64 off, char *dst, int n)
 {
   uint target;
   int c;
@@ -729,7 +729,7 @@ console_input_ready(void)
 }
 
 int
-consolewrite(struct inode *ip, uint off, char *buf, int n)
+consolewrite(struct inode *ip, uint64 off, char *buf, int n)
 {
   int i;
 
diff --git a/kernel/dev/cdev.c b/kernel/dev/cdev.c
index b9c1361..2071a91 100644
--- a/kernel/dev/cdev.c
+++ b/kernel/dev/cdev.c
@@ -8,33 +8,33 @@
 #include "x86.h"
 #include "errno.h"
 
-extern int consoleread(struct inode*, uint, char*, int);
-extern int consolewrite(struct inode*, uint, char*, int);
-extern int mouseread(struct inode*, uint, char*, int);
+extern int consoleread(struct inode*, uint64, char*, int);
+extern int consolewrite(struct inode*, uint64, char*, int);
+extern int mouseread(struct inode*, uint64, char*, int);
 
 static uint64 rseed = 0x2A99554433221100ULL;
 
 int
-nullwrite(struct inode *ip, uint off, char *dst, int n)
+nullwrite(struct inode *ip, uint64 off, char *dst, int n)
 {
   return n;
 }
 
 int
-nullread(struct inode *ip, uint off, char *dst, int n)
+nullread(struct inode *ip, uint64 off, char *dst, int n)
 {
   return 0;
 }
 
 int
-zeroread(struct inode *ip, uint off, char *dst, int n)
+zeroread(struct inode *ip, uint64 off, char *dst, int n)
 {
   memset(dst, 0, n);
   return n;
 }
 
 int
-randomread(struct inode *ip, uint off, char *dst, int n)
+randomread(struct inode *ip, uint64 off, char *dst, int n)
 {
   for(int i = 0; i < n; i++){
     rseed = rseed * 6364136223846793005ULL + 1442695040888963407ULL;
@@ -46,19 +46,19 @@ randomread(struct inode *ip, uint off, char *dst, int n)
 // /dev/full — reads as an endless run of zero bytes, every write fails with
 // ENOSPC. Programs use it to exercise their out-of-disk-space paths.
 int
-fullwrite(struct inode *ip, uint off, char *dst, int n)
+fullwrite(struct inode *ip, uint64 off, char *dst, int n)
 {
   return -ENOSPC;
 }
 
 int
-mousewrite(struct inode *ip, uint off, char *dst, int n)
+mousewrite(struct inode *ip, uint64 off, char *dst, int n)
 {
   return n;
 }
 
 int
-memread(struct inode *ip, uint off, char *dst, int n)
+memread(struct inode *ip, uint64 off, char *dst, int n)
 {
   switch(ip->minor) {
     case 1: return consoleread(ip, off, dst, n); 
@@ -73,7 +73,7 @@ memread(struct inode *ip, uint off, char *dst, int n)
 }
 
 int
-memwrite(struct inode *ip, uint off, char *dst, int n)
+memwrite(struct inode *ip, uint64 off, char *dst, int n)
 {
   switch(ip->minor) {
     case 1: return consolewrite(ip, off, dst, n);
diff --git a/kernel/dev/disk.c b/kernel/dev/disk.c
index 256b6ba..dd6ead7 100644
--- a/kernel/dev/disk.c
+++ b/kernel/dev/disk.c
@@ -31,14 +31,15 @@
 #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
+// The byte offset is 64-bit, all the way from struct file's off through the
+// devsw signature to bread's block number. It used to narrow to a 32-bit uint
+// at the devsw boundary, which capped a raw device at 4 GB -- and the constant
+// that documented the cap was never referenced by anything, so the failure was
+// not a refusal at 4 GB but silent wraparound past it. Nothing here may narrow
+// it again: an NVMe namespace or any modern disk is far beyond that.
 
 int
-diskread(struct inode *ip, uint off, char *dst, int n)
+diskread(struct inode *ip, uint64 off, char *dst, int n)
 {
   int dev = ip->minor;
 
@@ -49,8 +50,8 @@ diskread(struct inode *ip, uint off, char *dst, int n)
 
   int done = 0;
   while (done < n) {
-    uint64 pos   = (uint64)off + (uint64)done;
-    uint   blk   = (uint)(pos / BSIZE);
+    uint64 pos   = off + (uint64)done;
+    uint64 blk   = pos / BSIZE;
     uint   inblk = (uint)(pos % BSIZE);
     int    want  = BSIZE - (int)inblk;
     if (want > n - done)
@@ -66,7 +67,7 @@ diskread(struct inode *ip, uint off, char *dst, int n)
 }
 
 int
-diskwrite(struct inode *ip, uint off, char *src, int n)
+diskwrite(struct inode *ip, uint64 off, char *src, int n)
 {
   int dev = ip->minor;
 
@@ -77,8 +78,8 @@ diskwrite(struct inode *ip, uint off, char *src, int n)
 
   int done = 0;
   while (done < n) {
-    uint64 pos   = (uint64)off + (uint64)done;
-    uint   blk   = (uint)(pos / BSIZE);
+    uint64 pos   = off + (uint64)done;
+    uint64 blk   = pos / BSIZE;
     uint   inblk = (uint)(pos % BSIZE);
     int    want  = BSIZE - (int)inblk;
     if (want > n - done)
diff --git a/kernel/dev/fb.c b/kernel/dev/fb.c
index ad8c344..9abe8de 100644
--- a/kernel/dev/fb.c
+++ b/kernel/dev/fb.c
@@ -64,7 +64,7 @@ pixel_format(struct fb_var_screeninfo *v, unsigned int bpp)
 }
 
 int
-fbread(struct inode *ip, uint off, char *dst, int n)
+fbread(struct inode *ip, uint64 off, char *dst, int n)
 {
   volatile uchar *fb = vga_fb_ptr();
   uint64 size = vga_fb_size();
@@ -98,7 +98,7 @@ fb_mapped(void)
 }
 
 int
-fbwrite(struct inode *ip, uint off, char *src, int n)
+fbwrite(struct inode *ip, uint64 off, char *src, int n)
 {
   volatile uchar *fb = vga_fb_ptr();
   uint64 size = vga_fb_size();
diff --git a/kernel/dev/input.c b/kernel/dev/input.c
index bc1b399..8c10a62 100644
--- a/kernel/dev/input.c
+++ b/kernel/dev/input.c
@@ -233,7 +233,7 @@ input_ioctl(struct inode *ip, int request, addr_t arg)
 // desynchronise the reader permanently -- every later event would be split
 // across two reads and reassembled wrong.
 int
-inputread(struct inode *ip, uint off, char *dst, int n)
+inputread(struct inode *ip, uint64 off, char *dst, int n)
 {
   struct input_dev *d;
   int got = 0;
@@ -266,7 +266,7 @@ inputread(struct inode *ip, uint off, char *dst, int n)
 // evdev accepts writes to set keyboard LEDs. There are no LEDs to set here,
 // and a program that writes one should not see an error for it.
 int
-inputwrite(struct inode *ip, uint off, char *src, int n)
+inputwrite(struct inode *ip, uint64 off, char *src, int n)
 {
   return n;
 }
diff --git a/kernel/dev/mouse.c b/kernel/dev/mouse.c
index ac21454..05f90b0 100644
--- a/kernel/dev/mouse.c
+++ b/kernel/dev/mouse.c
@@ -197,7 +197,7 @@ uint8 get_mouse_buttons(void) {
 }
 
 int
-mouseread(struct inode *ip, uint off, char *dst, int n)
+mouseread(struct inode *ip, uint64 off, char *dst, int n)
 {
   struct mouse_data m;
   int size = sizeof(struct mouse_data);
diff --git a/kernel/dev/procfs.c b/kernel/dev/procfs.c
index 72cb391..0879258 100644
--- a/kernel/dev/procfs.c
+++ b/kernel/dev/procfs.c
@@ -346,7 +346,7 @@ proc_gen_global(int id, char *buf, int cap)
 }
 
 int
-procread(struct inode *ip, uint off, char *dst, int n)
+procread(struct inode *ip, uint64 off, char *dst, int n)
 {
   char buf[PROC_BUFSZ];
   int len = proc_gen_global(ip->minor, buf, PROC_BUFSZ);
@@ -422,7 +422,7 @@ proc_write_sysrq(int user_src, uint64 src, uint n)
 // 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, uint off, char *src, int n)
+procwrite(struct inode *ip, uint64 off, char *src, int n)
 {
   (void)off;