]> git.karo-electronics.de Git - karo-tx-linux.git/commitdiff
x86,mem-hotplug: pass sync_global_pgds() a correct argument in remove_pagetable()
authorYasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Thu, 26 Jun 2014 00:42:18 +0000 (10:42 +1000)
committerStephen Rothwell <sfr@canb.auug.org.au>
Thu, 26 Jun 2014 00:42:18 +0000 (10:42 +1000)
When hot-adding memory after hot-removing memory, following call traces
are shown:

kernel BUG at arch/x86/mm/init_64.c:206!
...
 [<ffffffff815e0c80>] kernel_physical_mapping_init+0x1b2/0x1d2
 [<ffffffff815ced94>] init_memory_mapping+0x1d4/0x380
 [<ffffffff8104aebd>] arch_add_memory+0x3d/0xd0
 [<ffffffff815d03d9>] add_memory+0xb9/0x1b0
 [<ffffffff81352415>] acpi_memory_device_add+0x1af/0x28e
 [<ffffffff81325dc4>] acpi_bus_device_attach+0x8c/0xf0
 [<ffffffff813413b9>] acpi_ns_walk_namespace+0xc8/0x17f
 [<ffffffff81325d38>] ? acpi_bus_type_and_status+0xb7/0xb7
 [<ffffffff81325d38>] ? acpi_bus_type_and_status+0xb7/0xb7
 [<ffffffff813418ed>] acpi_walk_namespace+0x95/0xc5
 [<ffffffff81326b4c>] acpi_bus_scan+0x9a/0xc2
 [<ffffffff81326bff>] acpi_scan_bus_device_check+0x8b/0x12e
 [<ffffffff81326cb5>] acpi_scan_device_check+0x13/0x15
 [<ffffffff81320122>] acpi_os_execute_deferred+0x25/0x32
 [<ffffffff8107e02b>] process_one_work+0x17b/0x460
 [<ffffffff8107edfb>] worker_thread+0x11b/0x400
 [<ffffffff8107ece0>] ? rescuer_thread+0x400/0x400
 [<ffffffff81085aef>] kthread+0xcf/0xe0
 [<ffffffff81085a20>] ? kthread_create_on_node+0x140/0x140
 [<ffffffff815fc76c>] ret_from_fork+0x7c/0xb0
 [<ffffffff81085a20>] ? kthread_create_on_node+0x140/0x140

The patch-sets fix the issue.

This patch (of 2):

remove_pagetable() gets start argument and passes the argument to
sync_global_pgds().  In this case, the argument must not be modified.  If
the argument is modified and passed to sync_global_pgds(),
sync_global_pgds() does not correctly synchronize PGD to PGD entries of
all processes MM since synchronized range of memory [start, end] is wrong.

Unfortunately the start argument is modified in remove_pagetable().  So
this patch fixes the issue.

Signed-off-by: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Tang Chen <tangchen@cn.fujitsu.com>
Acked-by: Toshi Kani <toshi.kani@hp.com>
Cc: Gu Zheng <guz.fnst@cn.fujitsu.com>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
arch/x86/mm/init_64.c

index df1a9927ad29ef9aa727851e775badb0f43a5416..a5b245d3f0cd44fe04a146a69d9a0b663286d573 100644 (file)
@@ -975,19 +975,20 @@ static void __meminit
 remove_pagetable(unsigned long start, unsigned long end, bool direct)
 {
        unsigned long next;
+       unsigned long addr;
        pgd_t *pgd;
        pud_t *pud;
        bool pgd_changed = false;
 
-       for (; start < end; start = next) {
-               next = pgd_addr_end(start, end);
+       for (addr = start; addr < end; addr = next) {
+               next = pgd_addr_end(addr, end);
 
-               pgd = pgd_offset_k(start);
+               pgd = pgd_offset_k(addr);
                if (!pgd_present(*pgd))
                        continue;
 
                pud = (pud_t *)pgd_page_vaddr(*pgd);
-               remove_pud_table(pud, start, next, direct);
+               remove_pud_table(pud, addr, next, direct);
                if (free_pud_table(pud, pgd))
                        pgd_changed = true;
        }