This is the type of thing such that if we get to the code, it is guaranteed that NULL will
not be returned by virtue of the fact that it would have been caught much earlier.
I don't want to start littering the code with assert, etc.
It is not as if we have not thought about these things and made certain decisions.
From: Zheng, Lv
Sent: Sunday, September 20, 2015 5:59 PM
To: Sudip Mukherjee; Moore, Robert
Cc: Wysocki, Rafael J; Len Brown; devel(a)acpica.org
Subject: RE: [PATCH] ACPICA: fix possible NULL dereference
If it is intentional, can we put something like assert(), BUG_ON() there
so that the checker can always be coded but not compiled when the binary
> From: Sudip Mukherjee [mailto:[email protected]
> Sent: Friday, September 18, 2015 2:06 PM
> On Thu, Sep 10, 2015 at 07:06:02PM +0000, Moore, Robert wrote:
> > One of the things we always have to think about and look at is that
> > often, things like this are *guaranteed" to be valid simply by
> virtue of the fact that the code has gotten as far as it has.
> > So, rather than litter the code with NULL checks for every single
access, we know up front that the object is valid.
> > This isn't always true and there are some valid situations, but we
would appreciate it if you help track this down.
> I tried but looks like acpi is very tough to understand. The only
> argument I can give in favor of my patch is that in all the places
> this function acpi_ns_get_secondary_object() has been used the return
> has been checked, only the three calls in this file
> acpi/acpica/dsargs.c has not checked the return. Like you said it
> might be intentional also to not check the return only in this file.