linux-kernel - suspicious capability check in ovl_ioctl_set

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>] [thread-next>] [day] [month] [year] [list]

Message-ID: <CACT4Y+bUfavwMVv2SEMve5pabE_AwsDO0YsRBGZtYqX59a77vA@mail.gmail.com>
Date:   Tue, 17 Nov 2020 08:56:15 +0100
From:   Dmitry Vyukov <dvyukov@...gle.com>
To:     Miklos Szeredi <miklos@...redi.hu>,
        overlayfs <linux-unionfs@...r.kernel.org>,
        LKML <linux-kernel@...r.kernel.org>
Cc:     Alexander Potapenko <glider@...gle.com>,
        Merna Zakaria <mernazakaria@...gle.com>,
        kasan-dev <kasan-dev@...glegroups.com>
Subject: suspicious capability check in ovl_ioctl_set_flags

Hi Miklos,

We've detected a suspicious double-fetch of user-space data in
ovl_ioctl_set_flags using a prototype tool (see report below [1]).

It points to ovl_ioctl_set_flags that does a capability check using
flags, but then the real ioctl double-fetches flags and uses
potentially different value:

static long ovl_ioctl_set_flags(struct file *file, unsigned int cmd,
                unsigned long arg, unsigned int flags)
{
...
    /* Check the capability before cred override */
    oldflags = ovl_iflags_to_fsflags(READ_ONCE(inode->i_flags));
    ret = vfs_ioc_setflags_prepare(inode, oldflags, flags);
    if (ret)
        goto unlock;
...
    ret = ovl_real_ioctl(file, cmd, arg);

All fs impls call vfs_ioc_setflags_prepare again, so the capability is
checked again.

But I think this makes the vfs_ioc_setflags_prepare check in overlayfs
pointless (?) and the "Check the capability before cred override"
comment misleading, user can skip this check by presenting benign
flags first and then overwriting them to non-benign flags. Or, if this
check is still needed... it is wrong (?). The code would need to
arrange for both ioctl's to operate on the same data then.
Does it make any sense?
Thanks

[1] BUG: multi-read in __x64_sys_ioctl  between ovl_ioctl and ext4_ioctl
======= First Address Range Stack =======
 df_save_stack+0x33/0x70 lib/df-detection.c:208
 add_address+0x2ac/0x352 lib/df-detection.c:47
 ovl_ioctl_set_fsflags fs/overlayfs/file.c:607 [inline]
 ovl_ioctl+0x7d/0x290 fs/overlayfs/file.c:654
 vfs_ioctl fs/ioctl.c:48 [inline]
 __do_sys_ioctl fs/ioctl.c:753 [inline]
 __se_sys_ioctl fs/ioctl.c:739 [inline]
 __x64_sys_ioctl+0xfc/0x140 fs/ioctl.c:739
 do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46
 entry_SYSCALL_64_after_hwframe+0x44/0xa9
======= Second Address Range Stack =======
 df_save_stack+0x33/0x70 lib/df-detection.c:208
 add_address+0x2ac/0x352 lib/df-detection.c:47
 ext4_ioctl+0x13b1/0x27f0 fs/ext4/ioctl.c:833
 vfs_ioctl+0x30/0x80 fs/ioctl.c:48
 ovl_real_ioctl+0xed/0x100 fs/overlayfs/file.c:539
 ovl_ioctl_set_flags+0x11d/0x180 fs/overlayfs/file.c:574
 ovl_ioctl_set_fsflags fs/overlayfs/file.c:610 [inline]
 ovl_ioctl+0x11e/0x290 fs/overlayfs/file.c:654
 vfs_ioctl fs/ioctl.c:48 [inline]
 __do_sys_ioctl fs/ioctl.c:753 [inline]
 __se_sys_ioctl fs/ioctl.c:739 [inline]
 __x64_sys_ioctl+0xfc/0x140 fs/ioctl.c:739
 do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46
 entry_SYSCALL_64_after_hwframe+0x44/0xa9
syscall number 16  System Call: __x64_sys_ioctl+0x0/0x140 fs/ioctl.c:800
First 0000000020000000 len 4 Caller vfs_ioctl fs/ioctl.c:48 [inline]
First 0000000020000000 len 4 Caller __do_sys_ioctl fs/ioctl.c:753 [inline]
First 0000000020000000 len 4 Caller __se_sys_ioctl fs/ioctl.c:739 [inline]
First 0000000020000000 len 4 Caller __x64_sys_ioctl+0xfc/0x140 fs/ioctl.c:739
Second 0000000020000000 len 4 Caller vfs_ioctl+0x30/0x80 fs/ioctl.c:48
==================================================================