Re: [Xen-devel] Xen Security Advisory 206 - xenstore denial of service via repeated update

[Ian Jackson <ian.jackson@xxxxxxxxxxxxx>]

(dropping some of the lists)

Michael Young writes ("Re: [Xen-devel] Xen Security Advisory 206 - xenstore 
denial of service via repeated update"):
> On Wed, 29 Mar 2017, Xen.org security team wrote:
> >                    Xen Security Advisory XSA-206
> >                              version 9
> >            xenstore denial of service via repeated update
> 
> I am seeing a build failure from these patches when using gcc 7. The 
> problem is with
> xsa206-4.80002-xenstored-Log-when-the-write-transaction-rate-limit-.patch 
> because in tools/xenstore/xenstored_domain.c the patch adds the boolean 
> wrl_delay_logged to the structure "domain" but later it tries to increment 
> it, resulting in the error 
> xenstored_domain.c: In function 'wrl_apply_debit_actual':
> xenstored_domain.c:949:32: error: increment of a boolean expression 
> [-Werror=bool-operation]
>     if (!domain->wrl_delay_logged++) {

I think this warning is wrong.


1. Increment of a boolean expression is perfectly well-defined:

With `_Bool b;', `b++' is equivalent to `b += 1' (C99 6.5.3.1(2))

That in turn is equivalent to `b = b + 1' (except that the lvalue b
is evaluated only once) (C99 6.5.16.2(3))

The expression b + 1 is of type int (because 1 is of type int and b
gets promoted to int by the usual arithmetic conversions (6.3.1.8,
6.3.1.1) since _Bool is the lowest ranked type.  So the expression has
value 1 or 2.

The constraints for the assignment (6.5.16.1) are satisfied because
both operands are arithmetic.  The right operand is converted to the
type of the left (6.5.16.1(2)).  Ie we convert (int)1 or (int)2 to
_Bool.

This is defined to be a zero-test (6.3.1.2) so the assigned value is
1.


2. Increment of a boolean expression feels more idiomatic to me,
certainly in this case, than plain assignment of 1.  (It is also more
flexible in case the code should be changed to count rather than
simply flag.)


It's a shame that we can't disable the warning about incrementing or
decrementing booleans, from other possible useful warnings such as
attempts to bitwise-invert them.

Ian.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
https://lists.xen.org/xen-devel