vnet_ioctl.c
来自「xen虚拟机源代码安装包」· C语言 代码 · 共 510 行
C
510 行
/* * Copyright (C) 2004, 2005 Mike Wray <mike.wray@hp.com> * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by the * Free Software Foundation; either version 2 of the License, or (at your * option) any later version. * * This program is distributed in the hope that it will be useful, but * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY * or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License * for more details. * * You should have received a copy of the GNU General Public License along * with this program; if not, write to the Free software Foundation, Inc., * 59 Temple Place, suite 330, Boston, MA 02111-1307 USA * */#include <linux/config.h>#include <linux/module.h>#include <linux/types.h>#include <linux/kernel.h>#include <linux/errno.h>#include <asm/uaccess.h>#include <linux/slab.h>#include <linux/proc_fs.h>#include <linux/string.h>#include <linux/net.h>#include <linux/in.h>#include <linux/inet.h>#include <linux/netdevice.h>#include <sa.h>#include "vif.h"#include "vnet.h"#include "varp.h"#include "vnet_dev.h"#include "vnet_eval.h"#include "vnet_forward.h"#include "iostream.h"#include "kernel_stream.h"#include "mem_stream.h"#include "sys_string.h"#include "sys_net.h"#include "sxpr_parser.h"#define MODULE_NAME "VNET"#define DEBUG 1#undef DEBUG#include "debug.h"/** @file * * Kernel interface to files in /proc. * todo: Add a sysfs interface using kobject. */#define PROC_ROOT "/proc/"#define PROC_ROOT_LEN 6#define MODULE_ROOT PROC_ROOT "vnet"enum { VNET_POLICY = 1, VNET_VNETS, VNET_VIFS, VNET_VARP, VNET_PEERS,};typedef struct proc_dir_entry ProcEntry;typedef struct inode Inode;typedef struct file File;static int proc_open_fn(struct inode *inode, File *file);//static ssize_t proc_read_fn(File *file, char *buffer, size_t count, loff_t *offset);//static ssize_t proc_write_fn(File *file, const char *buffer, size_t count, loff_t *offset) ;//static int proc_flush_fn(File *file);static loff_t proc_lseek_fn(File * file, loff_t offset, int orig);static int proc_ioctl_fn(struct inode *inode, File *file, unsigned opcode, unsigned long arg);//static int proc_release_fn(struct inode *inode, File *file);static int ProcEntry_has_name(ProcEntry *entry, const char *name, int namelen){ dprintf("> name=%.*s entry=%.*s\n", namelen, name, entry->namelen, entry->name); if(!entry || !entry->low_ino) return FALSE; if(entry->namelen != namelen) return FALSE; return memcmp(name, entry->name, namelen) == 0;}// Set f->f_error on error?// Does interface stop r/w on first error?// Is release called after an error?//static int proc_get_parser(File *file, Parser **val){ int err = 0; Parser *parser = NULL; parser = file->private_data; if(!parser){ parser = Parser_new(); if(!parser){ err = -ENOMEM; goto exit; } file->private_data = parser; } exit: *val = parser; return err;}static int proc_open_fn(Inode *inode, File *file){ // User open. // Return errcode or 0 on success. // Can stuff data in file->private_data (void*). // Get entry from //ProcEntry *entry = (ProcEntry *)inode->u.generic_ip; //file->private_data = NULL; //file->f_dentry->d_ino is inode. // Check for user privilege - deny otherwise. // -EACCESS int err = 0; dprintf(">\n"); file->private_data = NULL; return err;}static ssize_t proc_read_fn(File *file, char *buffer, size_t count, loff_t *offset){ // User read. // Copy data to user buffer, increment offset by count, return count. dprintf(">\n"); count = 0; //if(copy_to_user(buffer, data, count)){ // return -EFAULT; //} //*offset += count; return count;}#if 0static ssize_t proc_write_fn(File *file, const char *buffer, size_t count, loff_t *offset) { return -EINVAL;}#endif#if 0static int proc_flush_fn(File *file){ // User flush. int writing = (file->f_flags & O_ACCMODE) == O_WRONLY; int f_count = atomic_read(&file->f_count); if (writing && f_count == 1) { ProcEntry *pentry = (ProcEntry *)file->f_dentry->d_inode->u.generic_ip; // ... } return retval;}#endif#ifndef SEEK_SETenum { /** Offset from start. */ SEEK_SET = 0, /** Offset from current position. */ SEEK_CUR = 1, /** Offset from size of file. */ SEEK_END = 2};#endif /* !SEEK_SET */static loff_t proc_lseek_fn(File * file, loff_t offset, int from){ // User lseek. dprintf(">\n"); switch(from){ case SEEK_SET: break; case SEEK_CUR: offset += file->f_pos; break; case SEEK_END: return -EINVAL; default: return -EINVAL; } if(offset < 0) return -EINVAL; file->f_pos = offset; return offset;}static int proc_ioctl_fn(Inode *inode, File *file, unsigned opcode, unsigned long arg){ // User ioctl. dprintf(">\n"); return 0;}static ssize_t proc_policy_write_fn(File *file, const char *buffer, size_t count, loff_t *offset) { // User write. // Copy data into kernel space from buffer. // Increment offset by count, return count (or code). int err = 0; char *data = NULL; Parser *parser = NULL; err = proc_get_parser(file, &parser); if(err) goto exit; data = allocate(count); if(!data){ err = -ENOMEM; goto exit; } err = copy_from_user(data, buffer, count); if(err) goto exit; *offset += count; err = Parser_input(parser, data, count); exit: deallocate(data); err = (err < 0 ? err : count); return err;}static int proc_policy_release_fn(Inode *inode, File *file){ // User close. // Cleanup file->private_data, return errcode. int err = 0; Parser *parser = NULL; Sxpr obj, l; dprintf(">\n"); err = proc_get_parser(file, &parser); if(err) goto exit; err = Parser_input(parser, NULL, 0); if(err) goto exit; obj = parser->val; for(l = obj; CONSP(l); l = CDR(l)){ err = vnet_eval(CAR(l), iostdout, NULL); if(err) break; } exit: Parser_free(parser); file->private_data = NULL; dprintf("< err=%d\n", err); return err;}static int proc_io_open(Inode *inode, File *file, IOStream **val){ int err = 0; IOStream *io = mem_stream_new(); if(!io){ err = -ENOMEM; goto exit; } file->private_data = io; exit: *val = (err ? NULL: io); return err;}static ssize_t proc_io_read_fn(File *file, char *buffer, size_t count, loff_t *offset){ // User read. // Copy data to user buffer, increment offset by count, return count. int err = 0; char kbuf[1024] = {}; int kbuf_n = sizeof(kbuf); int k, n = 0; char *ubuf = buffer; IOStream *io = file->private_data; dprintf(">\n"); if(!io) goto exit; while(n < count){ k = count - n; if(k > kbuf_n){ k = kbuf_n; } k = IOStream_read(io, kbuf, k); if(k <= 0) break; if(copy_to_user(ubuf, kbuf, k)){ err = -EFAULT; goto exit; } n += k; ubuf += k; } *offset += n; exit: return (err ? err : n);}static int proc_io_release_fn(Inode *inode, File *file){ // User close. int err = 0; IOStream *io = file->private_data; if(io) IOStream_close(io); dprintf("< err=%d\n", err); return err;}static int proc_vnets_open_fn(Inode *inode, File *file){ int err = 0; IOStream *io; if(proc_io_open(inode, file, &io)) goto exit; vnet_print(io); exit: return err;}static int proc_vifs_open_fn(Inode *inode, File *file){ int err = 0; IOStream *io; if(proc_io_open(inode, file, &io)) goto exit; vif_print(io); exit: return err;}static int proc_peers_open_fn(Inode *inode, File *file){ int err = 0; IOStream *io; if(proc_io_open(inode, file, &io)) goto exit; vnet_peer_print(io); exit: return err;}static int proc_varp_open_fn(Inode *inode, File *file){ int err = 0; IOStream *io; if(proc_io_open(inode, file, &io)) goto exit; varp_print(io); exit: return err;}static struct file_operations proc_policy_ops = { open: proc_open_fn, read: proc_read_fn, write: proc_policy_write_fn, //flush: proc_flush_fn, llseek: proc_lseek_fn, ioctl: proc_ioctl_fn, release: proc_policy_release_fn,};static struct file_operations proc_vnets_ops = { open: proc_vnets_open_fn, read: proc_io_read_fn, release: proc_io_release_fn,};static struct file_operations proc_vifs_ops = { open: proc_vifs_open_fn, read: proc_io_read_fn, release: proc_io_release_fn,};static struct file_operations proc_peers_ops = { open: proc_peers_open_fn, read: proc_io_read_fn, release: proc_io_release_fn,};static struct file_operations proc_varp_ops = { open: proc_varp_open_fn, read: proc_io_read_fn, release: proc_io_release_fn,};static ProcEntry *proc_fs_root = &proc_root;static int proc_path_init(const char *path, const char **rest){ int err = 0; if(!path){ err = -EINVAL; goto exit; } if(*path == '/'){ if(strncmp(PROC_ROOT, path, PROC_ROOT_LEN)){ err = -EINVAL; } else { path += PROC_ROOT_LEN; } } exit: *rest = path; return err;}/** Parse a path relative to `dir'. If dir is null or the proc root * the path is relative to "/proc/", and the leading "/proc/" may be * supplied. * */static ProcEntry * ProcFS_lookup(const char *path, ProcEntry *dir){ const char *pathptr = path, *next = NULL; ProcEntry *entry, *result = NULL; int pathlen; if(dir && (dir != proc_fs_root)){ entry = dir; } else { if(proc_path_init(path, &pathptr)) goto exit; entry = proc_fs_root; } if(!pathptr || !*pathptr) goto exit; while(1){ next = strchr(pathptr, '/'); pathlen = (next ? next - pathptr : strlen(pathptr)); for(entry = entry->subdir; entry ; entry = entry->next) { if(ProcEntry_has_name(entry, pathptr, pathlen)) break; } if (!entry) break; if(!next){ result = entry; break; } pathptr = next + 1; } exit: return result;}static ProcEntry *ProcFS_register(const char *name, ProcEntry *dir, int val, struct file_operations *ops){ mode_t mode = 0; ProcEntry *entry; entry = create_proc_entry(name, mode, dir); if(entry){ entry->proc_fops = ops; entry->data = (void*)val; // Whatever data we need. } return entry;}static ProcEntry *ProcFS_mkdir(const char *name, ProcEntry *parent){ ProcEntry *entry = NULL; entry = ProcFS_lookup(name, parent); if(!entry){ const char *path; if(proc_path_init(name, &path)) goto exit; entry = proc_mkdir(path, parent); } exit: return entry;}static void ProcFS_remove(const char *name, ProcEntry *parent){ remove_proc_entry(name, parent);}static void ProcFS_rmrec_entry(ProcEntry *entry){ if(entry){ // Don't want to remove /proc itself! if(entry->parent == entry) return; while(entry->subdir){ ProcFS_rmrec_entry(entry->subdir); } dprintf("> remove %s\n", entry->name); ProcFS_remove(entry->name, entry->parent); }}static void ProcFS_rmrec(const char *name, ProcEntry *parent){ ProcEntry *entry; dprintf("> name=%s\n", name); entry = ProcFS_lookup(name, parent); if(entry){ ProcFS_rmrec_entry(entry); } dprintf("<\n");}void __init ProcFS_init(void){ ProcEntry *root_entry; ProcEntry *policy_entry; ProcEntry *vnets_entry; ProcEntry *vifs_entry; ProcEntry *peers_entry; ProcEntry *varp_entry; dprintf(">\n"); root_entry = ProcFS_mkdir(MODULE_ROOT, NULL); if(!root_entry) goto exit; policy_entry = ProcFS_register("policy", root_entry, VNET_POLICY, &proc_policy_ops); vnets_entry = ProcFS_register("vnets", root_entry, VNET_VNETS, &proc_vnets_ops); vifs_entry = ProcFS_register("vifs", root_entry, VNET_VIFS, &proc_vifs_ops); peers_entry = ProcFS_register("peers", root_entry, VNET_PEERS, &proc_peers_ops); varp_entry = ProcFS_register("varp", root_entry, VNET_VARP, &proc_varp_ops); exit: dprintf("<\n");}void __exit ProcFS_exit(void){ dprintf(">\n"); ProcFS_rmrec(MODULE_ROOT, NULL); dprintf("<\n");}
⌨️ 快捷键说明
复制代码Ctrl + C
搜索代码Ctrl + F
全屏模式F11
增大字号Ctrl + =
减小字号Ctrl + -
显示快捷键?