Add Linux kernel module support for 32-bit architectures (eg. ARM)

This change replaces all 64-bit division and modulo operations with
specific Linux kernel macros and functions to support 32-bit hardware
architectures. Thus, the xloop kernel modules can also run on 32-bit
ARM architectures, such as the Raspberry Pi 1 running Raspberry Pi OS.

3 files changed, 32 insertions, 8 deletions

diff --git a/README.md b/README.md
index 8b06805..bd3da5b 100644
--- a/README.md
+++ b/README.md
@@ -11,6 +11,7 @@ This repository contains the source code for the xloop Linux kernel modules
 and the user space utility **xlosetup** to configure xloop devices. The xloop Linux kernel modules can be built for the following Linux kernel versions and Linux distributions:
 
   - Archlinux with **Linux kernel 5.9.x** or **5.4.x**
+  - Raspberry Pi OS with **Linux kernel 5.4.x**
   - Ubuntu 20.04 with **Linux kernel 5.4.x**
   - Ubuntu 18.04 with **Linux kernel 4.19.x**
   - CentOS 8 with **Linux kernel 4.18.x**
@@ -34,6 +35,18 @@ pacman -S git \
           rpm-tools
 ```
 
+#### Raspberry Pi OS with Linux kernel 5.4.x
+```shell
+apt-get install git \
+                make \
+                cmake \
+                gcc \
+                raspberrypi-kernel-headers \
+                libcap-dev \
+                libncurses-dev \
+                rpm
+```
+
 #### Ubuntu 20.04 with Linux kernel 5.4.x
 ```shell
 apt-get install git \
diff --git a/src/kernel/xloop_file_fmt_qcow_cache.c b/src/kernel/xloop_file_fmt_qcow_cache.c
index 236add5..14f0b75 100644
--- a/src/kernel/xloop_file_fmt_qcow_cache.c
+++ b/src/kernel/xloop_file_fmt_qcow_cache.c
@@ -21,6 +21,7 @@
 #include <linux/limits.h>
 #include <linux/fs.h>
 #include <linux/vmalloc.h>
+#include <linux/math64.h>
 
 #include "xloop_file_fmt_qcow_main.h"
 #include "xloop_file_fmt_qcow_cache.h"
@@ -35,8 +36,14 @@ static inline int __xloop_file_fmt_qcow_cache_get_table_idx(
 	struct xloop_file_fmt_qcow_cache *c, void *table)
 {
 	ptrdiff_t table_offset = (u8 *) table - (u8 *) c->table_array;
-	int idx = table_offset / c->table_size;
-	ASSERT(idx >= 0 && idx < c->size && table_offset % c->table_size == 0);
+	int idx = div_s64(table_offset, c->table_size);
+
+#ifdef ASSERT
+	s32 rem_table_offset_mod_table_size;
+	div_s64_rem(table_offset, c->table_size, &rem_table_offset_mod_table_size);
+	ASSERT(idx >= 0 && idx < c->size && rem_table_offset_mod_table_size == 0);
+#endif
+
 	return idx;
 }
 
@@ -127,6 +134,7 @@ static int __xloop_file_fmt_qcow_cache_do_get(struct xloop_file_fmt *xlo_fmt,
 	u64 min_lru_counter = U64_MAX;
 	int min_lru_index = -1;
 	u64 read_offset;
+	u64 offset_div_table_size;
 	size_t len;
 
 	ASSERT(offset != 0);
@@ -139,7 +147,9 @@ static int __xloop_file_fmt_qcow_cache_do_get(struct xloop_file_fmt *xlo_fmt,
 	}
 
 	/* Check if the table is already cached */
-	i = lookup_index = (offset / c->table_size * 4) % c->size;
+	offset_div_table_size = div_u64(offset, c->table_size) * 4;
+	div_u64_rem(offset_div_table_size, c->size, &lookup_index);
+	i = lookup_index;
 	do {
 		const struct xloop_file_fmt_qcow_cache_table *t =
 			&c->entries[i];
diff --git a/src/kernel/xloop_file_fmt_qcow_main.c b/src/kernel/xloop_file_fmt_qcow_main.c
index dfde76d..376273d 100644
--- a/src/kernel/xloop_file_fmt_qcow_main.c
+++ b/src/kernel/xloop_file_fmt_qcow_main.c
@@ -26,6 +26,7 @@
 #ifdef CONFIG_ZSTD_DECOMPRESS
 #include <linux/zstd.h>
 #endif
+#include <linux/math64.h>
 
 #include <xloop/version.h>
 
@@ -119,7 +120,7 @@ static int __qcow_file_fmt_validate_table(struct xloop_file_fmt *xlo_fmt,
 {
 	struct xloop_file_fmt_qcow_data *qcow_data = xlo_fmt->private_data;
 
-	if (entries > max_size_bytes / entry_len) {
+	if (entries > div_s64(max_size_bytes, entry_len)) {
 		dev_err(xloop_file_fmt_to_dev(xlo_fmt), "%s too large\n", table_name);
 		return -EFBIG;
 	}
@@ -776,7 +777,7 @@ static int qcow_file_fmt_init(struct xloop_file_fmt *xlo_fmt)
 
 	/* create cache for L2 */
 	virtual_disk_size = qcow_data->size;
-	max_l2_entries = DIV_ROUND_UP(virtual_disk_size, qcow_data->cluster_size);
+	max_l2_entries = DIV64_U64_ROUND_UP(virtual_disk_size, qcow_data->cluster_size);
 	max_l2_cache = round_up(
 		max_l2_entries * xloop_file_fmt_qcow_l2_entry_size(qcow_data), 
 		qcow_data->cluster_size);
@@ -791,7 +792,7 @@ static int qcow_file_fmt_init(struct xloop_file_fmt *xlo_fmt)
 	l2_cache_entry_size = min(qcow_data->cluster_size, (int)PAGE_SIZE);
 
 	/* calculate the number of cache tables */
-	l2_cache_size /= l2_cache_entry_size;
+	l2_cache_size = div_u64(l2_cache_size, l2_cache_entry_size);
 	if (l2_cache_size < QCOW_MIN_L2_CACHE_SIZE) {
 		l2_cache_size = QCOW_MIN_L2_CACHE_SIZE;
 	}
@@ -802,8 +803,8 @@ static int qcow_file_fmt_init(struct xloop_file_fmt *xlo_fmt)
 		goto free_l1_table;
 	}
 
-	qcow_data->l2_slice_size = 
-		l2_cache_entry_size / xloop_file_fmt_qcow_l2_entry_size(qcow_data);
+	qcow_data->l2_slice_size = div_u64(l2_cache_entry_size,
+		xloop_file_fmt_qcow_l2_entry_size(qcow_data));
 
 	qcow_data->l2_table_cache = xloop_file_fmt_qcow_cache_create(xlo_fmt,
 		l2_cache_size, l2_cache_entry_size);