s390/mm: fix refcount usage for 4K pgste
authorEric Farman <farman@linux.ibm.com>
Fri, 29 Jun 2018 17:54:01 +0000 (19:54 +0200)
committerMartin Schwidefsky <schwidefsky@de.ibm.com>
Mon, 2 Jul 2018 09:25:39 +0000 (11:25 +0200)
s390 no longer uses the _mapcount field in struct page to identify
the page table format being used. While the code was diligent in handling
the different mappings, it neglected to turn "off" the map bits when
alloc_pgste was being used. This resulted in bits remaining "on" in the
_refcount field, and thus an artifically huge "in use" count that prevents
the pages from actually being released by __free_page.

There's opportunity for improvement in the "1 vs 3" vs "1U vs 3U" vs
"0x1 vs 0x11" etc. variations for all these calls, I am just keeping
things simple compared to neighboring code.

Fixes: 620b4e903179 ("s390: use _refcount for pgtables")
Reported-by: Halil Pasic <pasic@linux.ibm.com>
Bisected-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Eric Farman <farman@linux.ibm.com>
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
arch/s390/mm/pgalloc.c

index 84bd632..e3bd562 100644 (file)
@@ -252,6 +252,8 @@ void page_table_free(struct mm_struct *mm, unsigned long *table)
                spin_unlock_bh(&mm->context.lock);
                if (mask != 0)
                        return;
+       } else {
+               atomic_xor_bits(&page->_refcount, 3U << 24);
        }
 
        pgtable_page_dtor(page);
@@ -304,6 +306,8 @@ static void __tlb_remove_table(void *_table)
                        break;
                /* fallthrough */
        case 3:         /* 4K page table with pgstes */
+               if (mask & 3)
+                       atomic_xor_bits(&page->_refcount, 3 << 24);
                pgtable_page_dtor(page);
                __free_page(page);
                break;