mm/sparse.c: reset section's mem_map when fully deactivated
authorPingfan Liu <kernelfans@gmail.com>
Fri, 31 Jan 2020 06:11:10 +0000 (22:11 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 31 Jan 2020 18:30:36 +0000 (10:30 -0800)
After commit ba72b4c8cf60 ("mm/sparsemem: support sub-section hotplug"),
when a mem section is fully deactivated, section_mem_map still records
the section's start pfn, which is not used any more and will be
reassigned during re-addition.

In analogy with alloc/free pattern, it is better to clear all fields of
section_mem_map.

Beside this, it breaks the user space tool "makedumpfile" [1], which
makes assumption that a hot-removed section has mem_map as NULL, instead
of checking directly against SECTION_MARKED_PRESENT bit.  (makedumpfile
will be better to change the assumption, and need a patch)

The bug can be reproduced on IBM POWERVM by "drmgr -c mem -r -q 5" ,
trigger a crash, and save vmcore by makedumpfile

[1]: makedumpfile, commit e73016540293 ("[v1.6.7] Update version")

Link: http://lkml.kernel.org/r/1579487594-28889-1-git-send-email-kernelfans@gmail.com
Signed-off-by: Pingfan Liu <kernelfans@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Baoquan He <bhe@redhat.com>
Cc: Qian Cai <cai@lca.pw>
Cc: Kazuhito Hagio <k-hagio@ab.jp.nec.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/sparse.c

index 3822ecb..3918fc3 100644 (file)
@@ -789,7 +789,7 @@ static void section_deactivate(unsigned long pfn, unsigned long nr_pages,
                        ms->usage = NULL;
                }
                memmap = sparse_decode_mem_map(ms->section_mem_map, section_nr);
-               ms->section_mem_map = sparse_encode_mem_map(NULL, section_nr);
+               ms->section_mem_map = (unsigned long)NULL;
        }
 
        if (section_is_early && memmap)