Hey there, let's break down why you're seeing that vfs_fstatat undefined error and how to fix it for kernel 5.10.113.

What's Causing the Issue? #

Starting with Linux kernel 5.10, the vfs_stat() function you're using isn't a standalone exported function anymore—it's now a macro that expands to a call to vfs_fstatat(). The problem is vfs_fstatat() isn't exported to kernel modules (it's not in the module symbol table), so your module can't link against it.

Your existing code handles the user-space access switching correctly for 5.10+, but the vfs_stat() call itself is no longer compatible with module code in newer kernels.

The Fix #

We need to switch to using kernel APIs that are exported and intended for module use in 5.10+. Here's the updated version of your file_stat function:

#include <linux/fs.h>
#include <linux/sched.h>
#include <linux/uaccess.h>

int file_stat(char *path, struct kstat *st){
    mm_segment_t oldfs;
    int ret = 0;

    #if LINUX_VERSION_CODE < KERNEL_VERSION(5, 10, 0)
        // Keep your existing logic for kernels older than 5.10
        oldfs = get_fs();
        #if LINUX_VERSION_CODE < KERNEL_VERSION(5, 0, 0)
            set_fs(get_ds());
        #else
            set_fs(KERNEL_DS);
        #endif
        ret = vfs_stat(path, st);
        set_fs(oldfs);
    #else
        // New logic for 5.10+ kernels
        struct path file_path;
        // Resolve user-space path to kernel's internal path structure
        ret = user_path_at_empty(current->fs, path, AT_FDCWD, &file_path, 0);
        if (!ret) {
            // Get file attributes using the exported vfs_getattr
            ret = vfs_getattr(&file_path, st, STATX_ALL, AT_STATX_SYNC_AS_STAT);
            path_put(&file_path); // Don't forget to release the path reference
        }
    #endif

    return ret;
}

Key Changes Explained #

• Pre-5.10 Kernels: We leave your original code intact—vfs_stat() is still a valid, exported function here, and the fs switching logic works as expected.
• 5.10+ Kernels:
  • user_path_at_empty() safely converts the user-space path string into a kernel struct path, which is the preferred way to handle paths in modern kernel code.
  • vfs_getattr() is the exported API for retrieving file attributes in 5.10+. It handles all necessary permission checks and populates the struct kstat correctly, just like vfs_stat() used to.
  • Always call path_put() after working with a struct path to release its reference and avoid memory leaks.

Quick Checks #

• Make sure your module's Makefile is pointing to the correct 5.10 kernel headers—mismatched headers can cause unexpected errors.
• If you hit issues with STATX_ALL, verify that CONFIG_STATX is enabled in your kernel config (it's usually enabled by default in most distributions).

内容的提问来源于stack exchange，提问作者Jarvis Bao