Procprotect kernel module for protecting /proc in LXC containers
Sapan Bhatia [Tue, 24 Apr 2012 06:50:09 +0000 (02:50 -0400)]
procprotect.c [new file with mode: 0644]

diff --git a/procprotect.c b/procprotect.c
new file mode 100644 (file)
index 0000000..d49a41b
--- /dev/null
@@ -0,0 +1,228 @@
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/types.h>
+#include <linux/kernel.h>
+#include <linux/fs_struct.h>
+#include <linux/fs.h>
+#include <linux/mm.h>
+#include <linux/reboot.h>
+#include <linux/delay.h>
+#include <linux/proc_fs.h>
+#include <asm/uaccess.h>
+#include <linux/sysrq.h>
+#include <linux/timer.h>
+#include <linux/time.h>
+#include <linux/lglock.h>
+#include <linux/init.h>
+#include <linux/idr.h>
+#include <linux/namei.h>
+#include <linux/bitops.h>
+#include <linux/mount.h>
+#include <linux/dcache.h>
+#include <linux/spinlock.h>
+#include <linux/completion.h>
+#include <linux/sched.h>
+#include <linux/seq_file.h>
+#include <linux/kprobes.h>
+#include <linux/kallsyms.h>
+#include <linux/nsproxy.h>
+
+#define VERSION_STR "0.0.1"
+
+#ifndef CONFIG_X86_64
+#error "This code does not support your architecture"
+#endif
+
+static char *aclpath __devinitdata = "procprotect";
+static struct qstr aclqpath;
+
+module_param(aclpath, charp, 0);
+MODULE_PARM_DESC(aclpath, "Root of directory that stores acl tags for /proc files.");
+
+MODULE_AUTHOR("Sapan Bhatia <sapanb@cs.princeton.edu>");
+MODULE_DESCRIPTION("Lightweight ACLs for /proc.");
+MODULE_LICENSE("GPL");
+MODULE_VERSION(VERSION_STR);
+
+struct procprotect_ctx {
+       struct inode **inode;
+};
+
+struct acl_entry {
+       unsigned int ino;
+       struct hlist_node hlist;
+};
+
+#define HASH_SIZE (1<<16)
+
+struct hlist_head procprotect_hash[HASH_SIZE];
+
+struct proc_dir_entry *proc_entry;
+
+/*
+  Entry point of intercepted call. We need to do two things here:
+  - Decide if we need the heavier return hook to be called
+  - Save the first argument, which is in a register, for consideration in the return hook
+*/
+static int do_lookup_entry(struct kretprobe_instance *ri, struct pt_regs *regs) {
+       int ret = -1;
+       struct procprotect_ctx *ctx;
+       struct nameidata *nd = (struct nameidata *) regs->di;
+       struct qstr *q = (struct qstr *) regs->si;
+       struct dentry *parent = nd->path.dentry;
+       struct inode *pinode = parent->d_inode;
+       
+       if (pinode->i_sb->s_magic == PROC_SUPER_MAGIC) {        
+               ctx = (struct procprotect_ctx *) ri->data;
+               ctx->inode = (struct inode **) regs->cx;
+               ret = 0;
+       }
+
+       return ret;
+}
+
+static int run_acl(unsigned long ino) {
+       struct hlist_node *n;
+       struct acl_entry *entry;
+       hlist_for_each_entry_rcu(entry, 
+               n, &procprotect_hash[ino & (HASH_SIZE-1)],
+               hlist) {
+               if (entry->ino==ino) {
+                       return 0;
+               }
+       }
+       return 1;
+}
+
+/* The entry hook ensures that the return hook is only called for
+accesses to /proc */
+
+static int do_lookup_ret(struct kretprobe_instance *ri, struct pt_regs *regs)
+{
+       struct procprotect_ctx *ctx = (struct procprotect_ctx *) ri->data;
+       int ret = regs->ax;
+       
+       if (ret==0) {
+               /* The kernel is going to honor the request. Here's where we step in */
+               struct inode *inode = *(ctx->inode);
+               //printk(KERN_CRIT "Checking inode %x number %u",inode,inode->i_ino);
+               if (!run_acl(inode->i_ino)) {
+                       if (current->nsproxy->mnt_ns!=init_task.nsproxy->mnt_ns)
+                               regs->ax = -EPERM;
+               }
+       }
+       
+       return 0;
+}
+
+static struct kretprobe proc_probe = {
+       .entry_handler = (kprobe_opcode_t *) do_lookup_entry,
+        .handler = (kprobe_opcode_t *) do_lookup_ret,
+       .maxactive = 20,
+       .data_size = sizeof(struct procprotect_ctx)
+};
+
+static void add_entry(char *pathname) {
+       struct path path;
+       if (kern_path(pathname, 0, &path)) {
+               printk(KERN_CRIT "Path lookup failed for %s",pathname);
+       }       
+       else {
+               unsigned int ino = path.dentry->d_inode->i_ino;
+               struct acl_entry *entry;
+               entry = kmalloc(GFP_KERNEL, sizeof(struct acl_entry));
+               entry->ino = ino;
+               
+               if (!entry) {
+                       printk(KERN_CRIT "Could not allocate memory for %s",pathname);
+               }
+               else {
+                       if (run_acl(ino)) {
+                               hlist_add_head_rcu(&entry->hlist,&procprotect_hash[ino&(HASH_SIZE-1)]);
+                               printk(KERN_CRIT "Added inode %u",ino);
+                       }
+                       else {
+                               printk(KERN_CRIT "Did not add inode %u, already in list", ino);
+                       }
+               }
+       }
+}
+
+
+static void __exit procprotect_exit(void)
+{
+       unregister_kretprobe(&proc_probe);
+       struct hlist_node *n;
+       struct acl_entry *entry;
+       int i;
+
+       for (i=0;i<HASH_SIZE;i++) {
+               hlist_for_each_entry_rcu(entry, 
+                       n, &procprotect_hash[i],
+                       hlist) {
+                               kfree(entry);
+               }
+       }
+       
+       remove_proc_entry("procprotect",NULL);
+       printk("Procprotect: Stopped procprotect.\n");
+}
+
+
+
+int procfile_write(struct file *file, const char *buffer, unsigned long count, void *data) {           
+       char pathname[PATH_MAX];
+
+       if (current->nsproxy->mnt_ns!=init_task.nsproxy->mnt_ns)
+               return -EPERM;
+
+       if (copy_from_user(pathname, buffer, count)) {
+               return -EFAULT;
+       }
+       if (count && (pathname[count-1]==10 || pathname[count-1]==13)) {
+               pathname[count-1]='\0';
+       }
+       else
+               pathname[count]='\0';
+       add_entry(pathname);    
+       printk(KERN_CRIT "Length of buffer=%d",strlen(pathname));
+       return count;
+}
+
+static int __init procprotect_init(void)
+{
+       printk("Procprotect: starting procprotect version %s with ACLs at path %s.\n",
+              VERSION_STR, aclpath);
+       int ret;
+       int i;
+
+       for(i=0;i<HASH_SIZE;i++) {
+               INIT_HLIST_HEAD(&procprotect_hash[i]);
+       }
+
+         aclqpath.name = aclpath;
+         aclqpath.len = strnlen(aclpath, PATH_MAX);
+
+          proc_probe.kp.addr = 
+                  (kprobe_opcode_t *) kallsyms_lookup_name("do_lookup");
+          if (!proc_probe.kp.addr) {
+                  printk("Couldn't find %s to plant kretprobe\n", "do_execve");
+                  return -1;
+          }
+  
+          if ((ret = register_kretprobe(&proc_probe)) <0) {
+                  printk("register_kretprobe failed, returned %d\n", ret);
+                  return -1;
+          }
+          printk("Planted kretprobe at %p, handler addr %p\n",
+                 proc_probe.kp.addr, proc_probe.handler);
+
+       proc_entry = create_proc_entry("procprotect", 0644, NULL);
+       proc_entry->write_proc = procfile_write;
+        return ret;
+}
+
+
+
+module_init(procprotect_init);
+module_exit(procprotect_exit);