Re: [PATCHv2] evdev: fix evdev_write return value on partial writes

[Dmitry Torokhov <dmitry.torokhov@gmail.com>]

On Thu, Jan 27, 2011 at 01:43:17PM +0100, Peter Korsgaard wrote:
> >>>>> "Henrik" == Henrik Rydberg <rydberg@euromail.se> writes:
> 
>  Henrik> I won't argue against this case (with < 0) being frequent, but one
>  Henrik> should really check "n < len" to be safe. Hopefully Dmitry has some
>  Henrik> more input.
>  >> 
>  >> No, the point is that write (and read) can consume less data than
>  >> requested, without it being an error. Robust userspace code should
>  >> adjust buffer address / size and redo the work until all data is
>  >> transferred or an error occurs.
> 
>  Henrik> Shouldn't the error be on (!len || len % smallest_acceptable_chunk),
>  Henrik> then? Which makes me wonder about regressions - perhaps accumulating
>  Henrik> partial writes in evdev is more safe from that perspective.
> 
> No, writing more than 1 complete struct should just consume the full
> structs and return the number of bytes consumed, similar to all other
> cases in the kernel where we return a length < count.
> 
> No sane userspace will write anything else than a multiple of
> sizeof(input_event) though.
> 
> I doubt this will introduce any regressions (but you never know). The
> only situation I can see is if userspace would fill out a proper struct
> input_dev but use a wrong (too small) length in the write call. We used
> to accept these, but with the patch here it will -EINVAL.

I think that returning -EINVAL is good idea, so I am going to apply it.
I also think we can change this loop to do-while kind since we already
checked there enough space for an event.

-- 
Dmitry


Input: evdev - fix evdev_write return value on partial writes

From: Peter Korsgaard <jacmet@sunsite.dk>

As was recently brought up on the busybox list
(http://lists.busybox.net/pipermail/busybox/2011-January/074565.html),
evdev_write doesn't properly check the count argument, which will
lead to a return value > count on partial writes if the remaining bytes
are accessible - causing userspace confusion.

Fix it by only handling each full input_event structure and return -EINVAL
if less than 1 struct was written, similar to how it is done in evdev_read.

Reported-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
---

 drivers/input/evdev.c |   10 ++++++----
 1 files changed, 6 insertions(+), 4 deletions(-)


diff --git a/drivers/input/evdev.c b/drivers/input/evdev.c
index c8471a2..7f42d3a 100644
--- a/drivers/input/evdev.c
+++ b/drivers/input/evdev.c
@@ -321,6 +321,9 @@ static ssize_t evdev_write(struct file *file, const char __user *buffer,
 	struct input_event event;
 	int retval;
 
+	if (count < input_event_size())
+		return -EINVAL;
+
 	retval = mutex_lock_interruptible(&evdev->mutex);
 	if (retval)
 		return retval;
@@ -330,17 +333,16 @@ static ssize_t evdev_write(struct file *file, const char __user *buffer,
 		goto out;
 	}
 
-	while (retval < count) {
-
+	do {
 		if (input_event_from_user(buffer + retval, &event)) {
 			retval = -EFAULT;
 			goto out;
 		}
+		retval += input_event_size();
 
 		input_inject_event(&evdev->handle,
 				   event.type, event.code, event.value);
-		retval += input_event_size();
-	}
+	} while (retval + input_event_size() <= count);
 
  out:
 	mutex_unlock(&evdev->mutex);