linux-kernel - compat_ioctl question/problem

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Message-ID: <CAJhxZY4H85f3H6T4Qwgy79tDosbgrHY=3=VV1XaTh1R2qFsJww@mail.gmail.com>
Date:	Wed, 10 Feb 2016 15:03:56 -0500
From:	Soohoon pho <soohoon.photo@...il.com>
To:	linux-kernel@...r.kernel.org
Subject: compat_ioctl question/problem

Hi,

following is the code in question.

+static int compat_hdio_ioctl(struct inode *inode, struct file *file,
+                      struct gendisk *disk, unsigned int cmd,
unsigned long arg)
+{
+          mm_segment_t old_fs = get_fs();
+          unsigned long kval;
+          unsigned int __user *uvp;
+          int error;
+
+          set_fs(KERNEL_DS);
+          error = blkdev_driver_ioctl(inode, file, disk,
+                                              cmd, (unsigned long)(&kval));
+          set_fs(old_fs);
+
+          if (error == 0) {
+                      uvp = compat_ptr(arg);
+                      if (put_user(kval, uvp))
+                                  error = -EFAULT;
+          }
+          return error;
+}


kval is local so it has random value.
But one of syscall like HDIO_GET_32BIT/ATA_IOC_GET_IO32 only updates
one byte so if kval has 0xaaaaaaaa and ioctl() updates one byte then
it becomes 0xaaaaaa01.
And put_user() writes 4bytes.

Because of that I’m having problem with 32bit hdparm.

void process_dev (char *devname)
{
        static long parm, multcount;   <-- parm is static so it’s zero
…..
        if (do_defaults || get_io32bit) {
                if (0 == ioctl(fd, HDIO_GET_32BIT, &parm)) {


so parm becomes 0xaaaaaa01 and reports wrong mode.

If hdparm is 64bit then 64bit syscall will update 1byte of parm so it
becomes 0x1.

What would be a good fix if I can't use 64bit hdparm?
-       Modify hdparm to look at only 1byte
-       Initialize kval to zero
-       Copy parm to kval before blkdev_driver_ioctl() to be more
accurately close to 64bit syscall.


Thanks,
Soohoon.