[ David added to CC ]
On Wed, 14 Nov 2018, Eric Biggers wrote:
> From: Eric Biggers <ebiggers@xxxxxxxxxx>
>
> When a UHID_CREATE command is written to the uhid char device, a
> copy_from_user() is done from a user pointer embedded in the command.
> When the address limit is KERNEL_DS, e.g. as is the case during
> sys_sendfile(), this can read from kernel memory. Alternatively,
> information can be leaked from a setuid binary that is tricked to write
> to the file descriptor. Therefore, forbid UHID_CREATE in these cases.
>
> No other commands in uhid_char_write() are affected by this bug and
> UHID_CREATE is marked as "obsolete", so apply the restriction to
> UHID_CREATE only rather than to uhid_char_write() entirely.
>
> Thanks to Dmitry Vyukov for adding uhid definitions to syzkaller and to
> Jann Horn for commit 9da3f2b740544 ("x86/fault: BUG() when uaccess
> helpers fault on kernel addresses"), allowing this bug to be found.
>
> Reported-by: syzbot+72473edc9bf4eb1c6556@xxxxxxxxxxxxxxxxxxxxxxxxx
> Fixes: d365c6cfd337 ("HID: uhid: add UHID_CREATE and UHID_DESTROY events")
> Cc: <stable@xxxxxxxxxxxxxxx> # v3.6+
> Cc: Jann Horn <jannh@xxxxxxxxxx>
> Cc: Andy Lutomirski <luto@xxxxxxxxxx>
> Signed-off-by: Eric Biggers <ebiggers@xxxxxxxxxx>
Thanks for the patch. I however believe the fix below is more generic, and
would prefer taking that one in case noone sees any major flaw in that
I've overlooked. Thanks.
From: David Herrmann <dh.herrmann@xxxxxxxxx>
Subject: [PATCH] HID: uhid: prevent splice(2)
The kernel has a default implementation of splice(2) for writing from a
pipe into an arbitrary file. This behavior can be overriden by
providing an f_op.splice_write() callback.
Unfortunately, the default implementation of splice_write() takes page
by page from the source pipe, calls kmap() and passes the mapped page
as kernel-address to f_op.write(). Thus, it uses standard write(2) to
implement splice(2). However, since the page is kernel-mapped, they
have to `set_fs(get_ds())`. This is mostly fine, but UHID takes
command-streams through write(2), and thus it might interpret the data
taken as pointers. If called with KERNEL_DS, you can trick UHID to
allow kernel-space pointers as well.
As a simple fix, prevent splice(2) on UHID. It is unsecure, but it is
also non-functional. We need a linear mapping of the input in UHID, so
chunked input from splice(2) makes no sense, anyway.
Reported-by: syzbot+72473edc9bf4eb1c6556@xxxxxxxxxxxxxxxxxxxxxxxxx
Signed-off-by: David Herrmann <dh.herrmann@xxxxxxxxx>
---
drivers/hid/uhid.c | 10 ++++++++++
1 file changed, 10 insertions(+)
diff --git a/drivers/hid/uhid.c b/drivers/hid/uhid.c
index 3c5507313606..fefedc0b4dc6 100644
--- a/drivers/hid/uhid.c
+++ b/drivers/hid/uhid.c
@@ -753,6 +753,15 @@ static ssize_t uhid_char_write(struct file *file, const char __user *buffer,
return ret ? ret : count;
}
+static ssize_t uhid_char_splice_write(struct pipe_inode_info *pipe,
+ struct file *out,
+ loff_t *ppos,
+ size_t len,
+ unsigned int flags)
+{
+ return -EOPNOTSUPP;
+}
+
static __poll_t uhid_char_poll(struct file *file, poll_table *wait)
{
struct uhid_device *uhid = file->private_data;
@@ -771,6 +780,7 @@ static const struct file_operations uhid_fops = {
.release = uhid_char_release,
.read = uhid_char_read,
.write = uhid_char_write,
+ .splice_write = uhid_char_splice_write,
.poll = uhid_char_poll,
.llseek = no_llseek,
};
--
Jiri Kosina
SUSE Labs
