10:05 a.m.
From: Justin M. Forbes <jforbes(a)fedoraproject.org>
Fix gcc 12 warning address of 'mem_section' will never be NULL
Signed-off-by: Justin M. Forbes <jforbes(a)fedoraproject.org>
diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h
index blahblah..blahblah 100644
--- a/include/linux/mmzone.h
+++ b/include/linux/mmzone.h
@@ -1390,11 +1390,9 @@ static inline unsigned long *section_to_usemap(struct mem_section
*ms)
static inline struct mem_section *__nr_to_section(unsigned long nr)
{
#ifdef CONFIG_SPARSEMEM_EXTREME
- if (!mem_section)
+ if (!mem_section || !mem_section[SECTION_NR_TO_ROOT(nr)])
return NULL;
#endif
- if (!mem_section[SECTION_NR_TO_ROOT(nr)])
- return NULL;
return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
}
extern size_t mem_section_usage_size(void);
--
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1598
12:59 p.m.
New subject: [OS-BUILD PATCH] Fix gcc 12 warning address of 'mem_section' will
never
be NULL
From: Waiman Long on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1598#note_8284...
Actually, the correct fix should be "if (!*mem_section")", it is a bug in
the
upstream code.
3:22 p.m.
New subject: [OS-BUILD PATCH] Fix gcc 12 warning address of 'mem_section' will
never
be NULL
From: Justin M. Forbes on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1598#note_8285...
Updated with your patch which Andrew just pulled into mm.
3:52 p.m.
New subject: [OS-BUILD PATCH] Fix gcc 12 warning address of 'mem_section' will
never
be NULL
From: Justin M. Forbes on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1598#note_8285...
Actually, that doesn't solve the whole issue, and creates a new warning for
infinite recursion:
https://kojipkgs.fedoraproject.org/work/tasks/2682/82252682/build.log
4:48 p.m.
New subject: [OS-BUILD PATCH] Fix gcc 12 warning address of 'mem_section' will
never
be NULL
From: Waiman Long on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1598#note_8286...
I see, you are right. The subsequently statement should be pull inside the
CONFIG_SPARSEMEM_EXTREME block. So Fedora is built with
CONFIG_SPARSEMEM_EXTREME off. Right?
6:45 p.m.
New subject: [OS-BUILD PATCH] Fix gcc 12 warning address of 'mem_section' will
never
be NULL
From: Justin M. Forbes on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1598#note_8286...
Thanks, yes, it is on for x86_64 and off for everything else. The only thing
that fails is armv7 though.
12:40 p.m.
New subject: [OS-BUILD PATCH] Fix gcc 12 warning address of 'mem_section' will
never
be NULL
From: Rafael Aquini on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1598#note_8297...
Well, any pointer to pointer can be NULL, either by being not-initialized or
explicitly assigned so,
so the compiler warning (-Waddress) might be considering some other contextual
knowledge, as in:
mm/sparse.c:
225 /* Record a memory area against a node. */
226 static void __init memory_present(int nid, unsigned long start, unsigned
long end)
227 {
228 unsigned long pfn;
229
230 #ifdef CONFIG_SPARSEMEM_EXTREME
231 if (unlikely(!mem_section)) {
232 unsigned long size, align;
233
234 size = sizeof(struct mem_section *) * NR_SECTION_ROOTS;
235 align = 1 << (INTERNODE_CACHE_SHIFT);
236 mem_section = memblock_alloc(size, align);
...
The warning seems off, in a first glance, specially because there's no throw
for memory_present l:231...
Can't we just hint that preprocessor wrapped branch within __nr_to_section()
as unlikely, instead?
12:48 p.m.
New subject: [OS-BUILD PATCH] Fix gcc 12 warning address of 'mem_section' will
never
be NULL
From: Rafael Aquini on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1598#note_8297...
or perhaps what the new compiler really wants here is an explicit comparison,
as in
```
if ((void *)mem_section != NULL)
...
```
4:14 p.m.
New subject: [OS-BUILD PATCH] Fix gcc 12 warning address of 'mem_section' will
never
be NULL
From: Herton R. Krzesinski on gitlab.com
https://gitlab.com/cki-project/kernel-ark/-/merge_requests/1598#note_8951...
I think this thread can be resolved, doing it now.
752
days inactive
809
days old
kernel@lists.fedoraproject.org
8 comments
4 participants
participants (4)
-
Herton R. Krzesinski (via Email Bridge) -
Justin M. Forbes (via Email Bridge)
-
Rafael Aquini (via Email Bridge)
-
Waiman Long (via Email Bridge)