Re: [PATCH] lib/test_kmod: Fix an integer overflow test
On Sat, Feb 24, 2018 at 10:06:01PM +, Luis R. Rodriguez wrote: > On Sat, Feb 24, 2018 at 11:45:16AM +0300, Dan Carpenter wrote: > > On Sat, Feb 24, 2018 at 02:59:41AM +, Luis R. Rodriguez wrote: > > > On Mon, Jan 22, 2018 at 01:27:54PM +0300, Dan Carpenter wrote: > > > > The main problem is that the parentheses are in the wrong place and the > > > > unlikely() call returns either 0 or 1 so it's never less than zero. > > > > > > Doh, thanks, yes. Seems worth considering a grammar rule for it. > > > > > > > The other problem is that signed integer overflows like "INT_MAX + 1" > > > > are > > > > undefined behavior. > > > > > > Likewise. > > > > > > This seems like another possible generic typo issue. But I would not > > > resolve it > > > the way you did, in this particular case below num_test_devs represents > > > the > > > number of already registered devs, before we increment. So the way to > > > resolve > > > this would be: > > > > > > if (num_test_devs + 1 == INT_MAX) > > > > > > I'll get this upstream, thanks! > > > > There is no issue if num_test_devs is INT_MAX. But capping it at > > INT_MAX - 1 is also fine. > > If num_test_devs is INT_MAX, then doing num_test_devs + 1 overflows > and as you noted that is undefined? If it's INT_MAX we never do "num_test_devs + 1", we return a NULL. regards, dan carpenter
Re: [PATCH] lib/test_kmod: Fix an integer overflow test
On Sat, Feb 24, 2018 at 11:45:16AM +0300, Dan Carpenter wrote: > On Sat, Feb 24, 2018 at 02:59:41AM +, Luis R. Rodriguez wrote: > > On Mon, Jan 22, 2018 at 01:27:54PM +0300, Dan Carpenter wrote: > > > The main problem is that the parentheses are in the wrong place and the > > > unlikely() call returns either 0 or 1 so it's never less than zero. > > > > Doh, thanks, yes. Seems worth considering a grammar rule for it. > > > > > The other problem is that signed integer overflows like "INT_MAX + 1" are > > > undefined behavior. > > > > Likewise. > > > > This seems like another possible generic typo issue. But I would not > > resolve it > > the way you did, in this particular case below num_test_devs represents the > > number of already registered devs, before we increment. So the way to > > resolve > > this would be: > > > > if (num_test_devs + 1 == INT_MAX) > > > > I'll get this upstream, thanks! > > There is no issue if num_test_devs is INT_MAX. But capping it at > INT_MAX - 1 is also fine. If num_test_devs is INT_MAX, then doing num_test_devs + 1 overflows and as you noted that is undefined? Luis
Re: [PATCH] lib/test_kmod: Fix an integer overflow test
On Sat, Feb 24, 2018 at 02:59:41AM +, Luis R. Rodriguez wrote: > On Mon, Jan 22, 2018 at 01:27:54PM +0300, Dan Carpenter wrote: > > The main problem is that the parentheses are in the wrong place and the > > unlikely() call returns either 0 or 1 so it's never less than zero. > > Doh, thanks, yes. Seems worth considering a grammar rule for it. > > > The other problem is that signed integer overflows like "INT_MAX + 1" are > > undefined behavior. > > Likewise. > > This seems like another possible generic typo issue. But I would not resolve > it > the way you did, in this particular case below num_test_devs represents the > number of already registered devs, before we increment. So the way to resolve > this would be: > > if (num_test_devs + 1 == INT_MAX) > > I'll get this upstream, thanks! There is no issue if num_test_devs is INT_MAX. But capping it at INT_MAX - 1 is also fine. regards, dan carpenter
Re: [PATCH] lib/test_kmod: Fix an integer overflow test
On Mon, Jan 22, 2018 at 01:27:54PM +0300, Dan Carpenter wrote: > The main problem is that the parentheses are in the wrong place and the > unlikely() call returns either 0 or 1 so it's never less than zero. Doh, thanks, yes. Seems worth considering a grammar rule for it. > The other problem is that signed integer overflows like "INT_MAX + 1" are > undefined behavior. Likewise. This seems like another possible generic typo issue. But I would not resolve it the way you did, in this particular case below num_test_devs represents the number of already registered devs, before we increment. So the way to resolve this would be: if (num_test_devs + 1 == INT_MAX) I'll get this upstream, thanks! Luis > Fixes: d9c6a72d6fa2 ("kmod: add test driver to stress test the module loader") > Signed-off-by: Dan Carpenter > > diff --git a/lib/test_kmod.c b/lib/test_kmod.c > index e372b97eee13..30fd6d9e5361 100644 > --- a/lib/test_kmod.c > +++ b/lib/test_kmod.c > @@ -1141,7 +1141,7 @@ static struct kmod_test_device > *register_test_dev_kmod(void) > mutex_lock(®_dev_mutex); > > /* int should suffice for number of devices, test for wrap */ > - if (unlikely(num_test_devs + 1) < 0) { > + if (num_test_devs == INT_MAX) { > pr_err("reached limit of number of test devices\n"); > goto out; > } > -- Luis Rodriguez, SUSE LINUX GmbH Maxfeldstrasse 5; D-90409 Nuernberg
[PATCH] lib/test_kmod: Fix an integer overflow test
The main problem is that the parentheses are in the wrong place and the unlikely() call returns either 0 or 1 so it's never less than zero. The other problem is that signed integer overflows like "INT_MAX + 1" are undefined behavior. Fixes: d9c6a72d6fa2 ("kmod: add test driver to stress test the module loader") Signed-off-by: Dan Carpenter diff --git a/lib/test_kmod.c b/lib/test_kmod.c index e372b97eee13..30fd6d9e5361 100644 --- a/lib/test_kmod.c +++ b/lib/test_kmod.c @@ -1141,7 +1141,7 @@ static struct kmod_test_device *register_test_dev_kmod(void) mutex_lock(®_dev_mutex); /* int should suffice for number of devices, test for wrap */ - if (unlikely(num_test_devs + 1) < 0) { + if (num_test_devs == INT_MAX) { pr_err("reached limit of number of test devices\n"); goto out; }