On 11/24/21 14:27, Andrew Morton wrote:

On Wed, 24 Nov 2021 13:23:42 +0000 Matthew Wilcox [off-list ref] wrote:

quoted

On Wed, Nov 24, 2021 at 03:08:49AM +0000, cgel.zte@gmail.com wrote:

quoted

From: chiminghao <redacted>

Fix the following coccinelle report:
./mm/memory_hotplug.c:2210:2-5:
WARNING  Use BUG_ON instead of if condition followed by BUG.

What coccinelle script is reporting this?

quoted

-	if (try_remove_memory(start, size))
-		BUG();
+	BUG_ON(try_remove_memory(start, size));

I really, really, really do not like this.  For functions with
side-effects, this is bad style.  If it's a pure predicate, then
sure, but this is bad.

I don't like it either.  Yes, BUG() is special but it's such dangerous
practice.  I'd vote to change coccinelle.

Definitely! Or at least use a safer pattern/habit, with just a passive
variable in the BUG_ON() call, approximately like this:

diff --git a/mm/memory_hotplug.c b/mm/memory_hotplug.c
index 852041f6be41..48bd5ff341e7 100644
--- a/mm/memory_hotplug.c
+++ b/mm/memory_hotplug.c

@@ -2201,13 +2201,12 @@ static int __ref try_remove_memory(u64 start, u64 size)

*/
void __remove_memory(u64 start, u64 size)
{
-
+       int ret = try_remove_memory(start, size);
/*
* trigger BUG() if some memory is not offlined prior to calling this
* function
*/
-       if (try_remove_memory(start, size))
-               BUG();
+       BUG_ON(ret);
}

/*

...and by the way, while going to type that, I immediately stumbled upon
another pre-existing case of this sort of thing, in try_remove_memory(),
which does this:

static int __ref try_remove_memory(u64 start, u64 size)
{
	struct vmem_altmap mhp_altmap = {};
	struct vmem_altmap *altmap = NULL;
	unsigned long nr_vmemmap_pages;
	int rc = 0, nid = NUMA_NO_NODE;

	BUG_ON(check_hotplug_memory_range(start, size));

...


thanks,
-- 
John Hubbard
NVIDIA