无法在 ARM64 上的 Linux 内核中使用 set_memory_rw

Question

我正在尝试开发一个挂钩read() 系统调用的内核模块。由于某种原因，set_memory_rw() 函数似乎不起作用。

我看到了另一个类似的问题，但我真的不明白该怎么做。

我正在使用 Raspberry-pi 4 开发 Kali 4.19.93

我的代码：

#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/moduleparam.h>
#include <linux/syscalls.h>
#include <linux/kallsyms.h>
#include <linux/slab.h>
#include <linux/kern_levels.h>
#include <asm/unistd.h>
#include <asm/cacheflush.h>
#include <linux/semaphore.h>
#include <asm/set_memory.h>

MODULE_LICENSE("GPL");
MODULE_AUTHOR("Omri Ben David");
MODULE_DESCRIPTION("Hooking Linux System calls");
MODULE_VERSION("1.0");

unsigned long** SYS_CALL_TABLE = (unsigned long**) 0xc02011c4;

asmlinkage ssize_t (*original_read) (int fd, char *buf, size_t count);
asmlinkage ssize_t HookRead(unsigned int fd, char * buf, size_t count)
{
    printk(KERN_INFO "Rootkit_Debug: Yay you entered my function!!\n Now you can read\n");
    return (*original_read)(fd,buf,count);
}   

void (*seek)(unsigned long, int);
void (*hide)(unsigned long, int);

static int __init SetHooks(void)
{
    printk(KERN_INFO "Hooks Will now be set, hold on tight\n");
    printk(KERN_INFO "System calls table is at address %p\n",SYS_CALL_TABLE);

    original_read = (void*) SYS_CALL_TABLE[__NR_read];
    
    seek = (void*) kallsyms_lookup_name("set_memory_rw");
    hide = (void*) kallsyms_lookup_name("set_memory_ro");

    (*seek)((unsigned long)SYS_CALL_TABLE, 1);
    SYS_CALL_TABLE[__NR_read] = (unsigned long*)HookRead;
    (*hide)((unsigned long)SYS_CALL_TABLE, 1);
    printk(KERN_INFO "System calls hooked successfully\n");

    return 0;
}

static void __exit HookCleanup(void)
{
    printk(KERN_INFO "System calls restore initiated\n");

    (*seek)((unsigned long)SYS_CALL_TABLE, 1);
    SYS_CALL_TABLE[__NR_read] = (unsigned long*) original_read;
    (*hide)((unsigned long)SYS_CALL_TABLE, 1);

    printk(KERN_INFO "System successfully restored. hope you had fun");
}

module_init(SetHooks);
module_exit(HookCleanup);

如何使set_memory_rw() 函数工作以覆盖系统调用表？还是我应该使用其他方法？

Answer 1

因此，正如我在上面的 cmets 中所说，似乎函数 change_memory_common()（由 set_memory_ro/rw() 使用）在应用请求的权限之前会进行检查。这记录在注释中：

/*
 * Kernel VA mappings are always live, and splitting live section
 * mappings into page mappings may cause TLB conflicts. This means
 * we have to ensure that changing the permission bits of the range
 * we are operating on does not result in such splitting.
 *
 * Let's restrict ourselves to mappings created by vmalloc (or vmap).
 * Those are guaranteed to consist entirely of page mappings, and
 * splitting is never needed.
 *
 * So check whether the [addr, addr + size) interval is entirely
 * covered by precisely one VM area that has the VM_ALLOC flag set.
 */
area = find_vm_area((void *)addr);
if (!area ||
    end > (unsigned long)area->addr + area->size ||
    !(area->flags & VM_ALLOC))
    return -EINVAL;

该函数似乎只适用于通过vmalloc() 或vmap() 创建的映射，而sys_call_table 不存在于此类映射中。

问题似乎与 TLB 冲突有关。我真的不确定为什么这会导致 TLB 冲突，因为函数 seems to properly call flush_tlb_kernel_range()，但我不是 ARM 专家，所以我可能会遗漏一些东西。我想有人可以打电话给flush_tlb_all()，但它似乎没有必要。欢迎任何其他见解！

在任何情况下，为了练习系统调用劫持，您可以重新编写自己的set_memory_common() 和set_memory_rw/ro() 版本以避免此检查。一种更简单的方法是只为所需地址获取适当的 PTE，然后更改权限，但我没有为此查看所有无数的宏。

最后但同样重要的是，由于sys_call_table 最终可能会跨越页面边界，因此在对页面应用更改时最好使用syscall_table + __NR_read 而不是仅使用sys_call_table。

这是一个工作示例：

// SPDX-License-Identifier: GPL-3.0
#include <linux/init.h>     // module_{init,exit}()
#include <linux/module.h>   // THIS_MODULE, MODULE_VERSION, ...
#include <linux/kernel.h>   // printk(), pr_*()
#include <linux/kallsyms.h> // kallsyms_lookup_name()
#include <asm/syscall.h>    // syscall_fn_t, __NR_*
#include <asm/ptrace.h>     // struct pt_regs
#include <asm/tlbflush.h>   // flush_tlb_kernel_range()
#include <asm/pgtable.h>    // {clear,set}_pte_bit(), set_pte()
#include <linux/vmalloc.h>  // vm_unmap_aliases()
#include <linux/mm.h>       // struct mm_struct, apply_to_page_range()
#include <linux/kconfig.h>  // IS_ENABLED()

#ifdef pr_fmt
#undef pr_fmt
#endif
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

static struct mm_struct *init_mm_ptr;
static syscall_fn_t *syscall_table;
static syscall_fn_t original_read;

/********** HELPERS **********/

// From arch/arm64/mm/pageattr.c.
struct page_change_data {
    pgprot_t set_mask;
    pgprot_t clear_mask;
};

// From arch/arm64/mm/pageattr.c.
static int change_page_range(pte_t *ptep, unsigned long addr, void *data)
{
    struct page_change_data *cdata = data;
    pte_t pte = READ_ONCE(*ptep);

    pte = clear_pte_bit(pte, cdata->clear_mask);
    pte = set_pte_bit(pte, cdata->set_mask);

    set_pte(ptep, pte);
    return 0;
}

// From arch/arm64/mm/pageattr.c.
static int __change_memory_common(unsigned long start, unsigned long size,
                  pgprot_t set_mask, pgprot_t clear_mask)
{
    struct page_change_data data;
    int ret;

    data.set_mask = set_mask;
    data.clear_mask = clear_mask;

    ret = apply_to_page_range(init_mm_ptr, start, size, change_page_range, &data);

    flush_tlb_kernel_range(start, start + size);
    return ret;
}

// Simplified set_memory_rw() from arch/arm64/mm/pageattr.c.
static int set_page_rw(unsigned long addr)
{
    vm_unmap_aliases();    
    return __change_memory_common(addr, PAGE_SIZE, __pgprot(PTE_WRITE), __pgprot(PTE_RDONLY));
}

// Simplified set_memory_ro() from arch/arm64/mm/pageattr.c.
static int set_page_ro(unsigned long addr)
{
    vm_unmap_aliases();
    return __change_memory_common(addr, PAGE_SIZE, __pgprot(PTE_RDONLY), __pgprot(PTE_WRITE));
}

/********** ACTUAL MODULE **********/

static long myread(const struct pt_regs *regs)
{
    pr_info("read() called\n");
    return original_read(regs);
}

static int __init modinit(void)
{
    int res;

    pr_info("init\n");

    // Shouldn't fail.
    init_mm_ptr = (struct mm_struct *)kallsyms_lookup_name("init_mm");
    syscall_table = (syscall_fn_t *)kallsyms_lookup_name("sys_call_table");

    original_read = syscall_table[__NR_read];

    res = set_page_rw((unsigned long)(syscall_table + __NR_read) & PAGE_MASK);
    if (res != 0) {
        pr_err("set_page_rw() failed: %d\n", res);
        return res;
    }

    syscall_table[__NR_read] = myread;

    res = set_page_ro((unsigned long)(syscall_table + __NR_read) & PAGE_MASK);
    if (res != 0) {
        pr_err("set_page_ro() failed: %d\n", res);
        return res;
    }

    pr_info("init done\n");

    return 0;
}

static void __exit modexit(void)
{
    int res;

    pr_info("exit\n");

    res = set_page_rw((unsigned long)(syscall_table + __NR_read) & PAGE_MASK);
    if (res != 0) {
        pr_err("set_page_rw() failed: %d\n", res);
        return;
    }

    syscall_table[__NR_read] = original_read;

    res = set_page_ro((unsigned long)(syscall_table + __NR_read) & PAGE_MASK);
    if (res != 0)
        pr_err("set_page_ro() failed: %d\n", res);

    pr_info("goodbye\n");
}

module_init(modinit);
module_exit(modexit);
MODULE_VERSION("0.1");
MODULE_DESCRIPTION("Syscall hijack on arm64.");
MODULE_AUTHOR("Marco Bonelli");
MODULE_LICENSE("GPL");