xen: pin correct PGD on suspend

Impact: fix Xen guest boot failure

commit eefb47f6a1e855653d275cb90592a3587ea93a09 ("xen: use
spin_lock_nest_lock when pinning a pagetable") changed xen_pgd_walk to
walk over mm->pgd rather than taking pgd as an argument.

This breaks xen_mm_(un)pin_all() because it makes init_mm.pgd readonly
instead of the pgd we are interested in and therefore the pin subsequently
fails.

(XEN) mm.c:2280:d15 Bad type (saw 00000000e8000001 != exp 0000000060000000) for mfn bc464 (pfn 21ca7)
(XEN) mm.c:2665:d15 Error while pinning mfn bc464

[   14.586913] 1 multicall(s) failed: cpu 0
[   14.586926] Pid: 14, comm: kstop/0 Not tainted 2.6.28-rc5-x86_32p-xenU-00172-gee2f6cc #200
[   14.586940] Call Trace:
[   14.586955]  [<c030c17a>] ? printk+0x18/0x1e
[   14.586972]  [<c0103df3>] xen_mc_flush+0x163/0x1d0
[   14.586986]  [<c0104bc1>] __xen_pgd_pin+0xa1/0x110
[   14.587000]  [<c015a330>] ? stop_cpu+0x0/0xf0
[   14.587015]  [<c0104d7b>] xen_mm_pin_all+0x4b/0x70
[   14.587029]  [<c022bcb9>] xen_suspend+0x39/0xe0
[   14.587042]  [<c015a330>] ? stop_cpu+0x0/0xf0
[   14.587054]  [<c015a3cd>] stop_cpu+0x9d/0xf0
[   14.587067]  [<c01417cd>] run_workqueue+0x8d/0x150
[   14.587080]  [<c030e4b3>] ? _spin_unlock_irqrestore+0x23/0x40
[   14.587094]  [<c014558a>] ? prepare_to_wait+0x3a/0x70
[   14.587107]  [<c0141918>] worker_thread+0x88/0xf0
[   14.587120]  [<c01453c0>] ? autoremove_wake_function+0x0/0x50
[   14.587133]  [<c0141890>] ? worker_thread+0x0/0xf0
[   14.587146]  [<c014509c>] kthread+0x3c/0x70
[   14.587157]  [<c0145060>] ? kthread+0x0/0x70
[   14.587170]  [<c0109d1b>] kernel_thread_helper+0x7/0x10
[   14.587181]   call  1/3: op=14 arg=[c0415000] result=0
[   14.587192]   call  2/3: op=14 arg=[e1ca2000] result=0
[   14.587204]   call  3/3: op=26 arg=[c1808860] result=-22

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

1 files changed, 14 insertions, 7 deletions

diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c
index 688936044dc9..636ef4caa52d 100644
--- a/arch/x86/xen/mmu.c
+++ b/arch/x86/xen/mmu.c
@@ -661,12 +661,11 @@ void xen_set_pgd(pgd_t *ptr, pgd_t val)
  * For 64-bit, we must skip the Xen hole in the middle of the address
  * space, just after the big x86-64 virtual hole.
  */
-static int xen_pgd_walk(struct mm_struct *mm,
-                        int (*func)(struct mm_struct *mm, struct page *,
-                                    enum pt_level),
-                        unsigned long limit)
+static int __xen_pgd_walk(struct mm_struct *mm, pgd_t *pgd,
+                          int (*func)(struct mm_struct *mm, struct page *,
+                                      enum pt_level),
+                          unsigned long limit)
 {
-        pgd_t *pgd = mm->pgd;
         int flush = 0;
         unsigned hole_low, hole_high;
         unsigned pgdidx_limit, pudidx_limit, pmdidx_limit;
@@ -753,6 +752,14 @@ out:
         return flush;
 }
 
+static int xen_pgd_walk(struct mm_struct *mm,
+                        int (*func)(struct mm_struct *mm, struct page *,
+                                    enum pt_level),
+                        unsigned long limit)
+{
+        return __xen_pgd_walk(mm, mm->pgd, func, limit);
+}
+
 /* If we're using split pte locks, then take the page's lock and
    return a pointer to it.  Otherwise return NULL. */
 static spinlock_t *xen_pte_lock(struct page *page, struct mm_struct *mm)
@@ -854,7 +861,7 @@ static void __xen_pgd_pin(struct mm_struct *mm, pgd_t *pgd)
 
         xen_mc_batch();
 
-         if (xen_pgd_walk(mm, xen_pin_page, USER_LIMIT)) {
+        if (__xen_pgd_walk(mm, pgd, xen_pin_page, USER_LIMIT)) {
                 /* re-enable interrupts for flushing */
                 xen_mc_issue(0);
 
@@ -998,7 +1005,7 @@ static void __xen_pgd_unpin(struct mm_struct *mm, pgd_t *pgd)
                        PT_PMD);
 #endif
 
-        xen_pgd_walk(mm, xen_unpin_page, USER_LIMIT);
+        __xen_pgd_walk(mm, pgd, xen_unpin_page, USER_LIMIT);
 
         xen_mc_issue(0);
 }