On Tue, May 25, 2021 at 3:52 PM David Hildenbrand [off-list ref] wrote:

On 17.05.21 13:20, Dong Aisheng wrote:

quoted

Free section usage memory in case populate_section_memmap failed.
We use map_count to track the remain unused memory to be freed.

Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Dong Aisheng <aisheng.dong@nxp.com>
---
  mm/sparse.c | 2 ++
  1 file changed, 2 insertions(+)

diff --git a/mm/sparse.c b/mm/sparse.c
index 7ac481353b6b..98bfacc763da 100644
--- a/mm/sparse.c
+++ b/mm/sparse.c

@@ -549,12 +549,14 @@ static void __init sparse_init_nid(int nid, unsigned long pnum_begin,
                             __func__, nid);
                      pnum_begin = pnum;
                      sparse_buffer_fini();
+                     memblock_free_early(__pa(usage), map_count * mem_section_usage_size());
                      goto failed;
              }
              check_usemap_section_nr(nid, usage);
              sparse_init_one_section(__nr_to_section(pnum), pnum, map, usage,
                              SECTION_IS_EARLY);
              usage = (void *) usage + mem_section_usage_size();
+             map_count--;
      }
      sparse_buffer_fini();
      return;

Why care about optimizing something that literally never fails, and if
it would fail, leave the system in a sub-optimal, mostly unusable state?

It just fixes the code logic.
I'm going to listen to more comments, if everyone thinks it's not needed.
I may drop in V2.

Regards
Aisheng

--
Thanks,

David / dhildenb