mm/hotplug: fix hot remove failure in SPARSEMEM|!VMEMMAP case
authorBaoquan He <bhe@redhat.com>
Sun, 22 Mar 2020 01:22:13 +0000 (18:22 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sun, 22 Mar 2020 01:56:06 +0000 (18:56 -0700)
In section_deactivate(), pfn_to_page() doesn't work any more after
ms->section_mem_map is resetting to NULL in SPARSEMEM|!VMEMMAP case.  It
causes a hot remove failure:

  kernel BUG at mm/page_alloc.c:4806!
  invalid opcode: 0000 [#1] SMP PTI
  CPU: 3 PID: 8 Comm: kworker/u16:0 Tainted: G        W         5.5.0-next-20200205+ #340
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 0.0.0 02/06/2015
  Workqueue: kacpi_hotplug acpi_hotplug_work_fn
  RIP: 0010:free_pages+0x85/0xa0
  Call Trace:
   __remove_pages+0x99/0xc0
   arch_remove_memory+0x23/0x4d
   try_remove_memory+0xc8/0x130
   __remove_memory+0xa/0x11
   acpi_memory_device_remove+0x72/0x100
   acpi_bus_trim+0x55/0x90
   acpi_device_hotplug+0x2eb/0x3d0
   acpi_hotplug_work_fn+0x1a/0x30
   process_one_work+0x1a7/0x370
   worker_thread+0x30/0x380
   kthread+0x112/0x130
   ret_from_fork+0x35/0x40

Let's move the ->section_mem_map resetting after
depopulate_section_memmap() to fix it.

[akpm@linux-foundation.org: remove unneeded initialization, per David]
Fixes: ba72b4c8cf60 ("mm/sparsemem: support sub-section hotplug")
Signed-off-by: Baoquan He <bhe@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Wei Yang <richardw.yang@linux.intel.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: <stable@vger.kernel.org>
Link: http://lkml.kernel.org/r/20200307084229.28251-2-bhe@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/sparse.c

index 596b2a4..aadb729 100644 (file)
@@ -734,6 +734,7 @@ static void section_deactivate(unsigned long pfn, unsigned long nr_pages,
        struct mem_section *ms = __pfn_to_section(pfn);
        bool section_is_early = early_section(ms);
        struct page *memmap = NULL;
+       bool empty;
        unsigned long *subsection_map = ms->usage
                ? &ms->usage->subsection_map[0] : NULL;
 
@@ -764,7 +765,8 @@ static void section_deactivate(unsigned long pfn, unsigned long nr_pages,
         * For 2/ and 3/ the SPARSEMEM_VMEMMAP={y,n} cases are unified
         */
        bitmap_xor(subsection_map, map, subsection_map, SUBSECTIONS_PER_SECTION);
-       if (bitmap_empty(subsection_map, SUBSECTIONS_PER_SECTION)) {
+       empty = bitmap_empty(subsection_map, SUBSECTIONS_PER_SECTION);
+       if (empty) {
                unsigned long section_nr = pfn_to_section_nr(pfn);
 
                /*
@@ -779,13 +781,15 @@ 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 = (unsigned long)NULL;
        }
 
        if (section_is_early && memmap)
                free_map_bootmem(memmap);
        else
                depopulate_section_memmap(pfn, nr_pages, altmap);
+
+       if (empty)
+               ms->section_mem_map = (unsigned long)NULL;
 }
 
 static struct page * __meminit section_activate(int nid, unsigned long pfn,