block: null_blk: fix 'Invalid parameters' when loading module
authorMing Lei <ming.lei@redhat.com>
Tue, 6 Mar 2018 04:07:13 +0000 (12:07 +0800)
committerJens Axboe <axboe@kernel.dk>
Tue, 6 Mar 2018 17:35:09 +0000 (10:35 -0700)
On ARM64, the default page size has been 64K on some distributions, and
we should allow ARM64 people to play null_blk.

This patch fixes the issue by extend page bitmap size for supporting
other non-4KB PAGE_SIZE.

Cc: Bart Van Assche <Bart.VanAssche@wdc.com>
Cc: Shaohua Li <shli@kernel.org>
Cc: Kyungchan Koh <kkc6196@fb.com>,
Cc: weiping zhang <zhangweiping@didichuxing.com>
Cc: Yi Zhang <yi.zhang@redhat.com>
Reported-by: Yi Zhang <yi.zhang@redhat.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
drivers/block/null_blk.c

index 6dc7e7cfca4a6e708d31cb4c42a7875ec2fe7d5b..d6be7a6d8ca623e5faf1a280f5dd2e99bb1311b5 100644 (file)
@@ -74,6 +74,7 @@ enum nullb_device_flags {
        NULLB_DEV_FL_CACHE      = 3,
 };
 
+#define MAP_SZ         ((PAGE_SIZE >> SECTOR_SHIFT) + 2)
 /*
  * nullb_page is a page in memory for nullb devices.
  *
@@ -88,10 +89,10 @@ enum nullb_device_flags {
  */
 struct nullb_page {
        struct page *page;
-       unsigned long bitmap;
+       DECLARE_BITMAP(bitmap, MAP_SZ);
 };
-#define NULLB_PAGE_LOCK (sizeof(unsigned long) * 8 - 1)
-#define NULLB_PAGE_FREE (sizeof(unsigned long) * 8 - 2)
+#define NULLB_PAGE_LOCK (MAP_SZ - 1)
+#define NULLB_PAGE_FREE (MAP_SZ - 2)
 
 struct nullb_device {
        struct nullb *nullb;
@@ -733,7 +734,7 @@ static struct nullb_page *null_alloc_page(gfp_t gfp_flags)
        if (!t_page->page)
                goto out_freepage;
 
-       t_page->bitmap = 0;
+       memset(t_page->bitmap, 0, sizeof(t_page->bitmap));
        return t_page;
 out_freepage:
        kfree(t_page);
@@ -743,13 +744,20 @@ out:
 
 static void null_free_page(struct nullb_page *t_page)
 {
-       __set_bit(NULLB_PAGE_FREE, &t_page->bitmap);
-       if (test_bit(NULLB_PAGE_LOCK, &t_page->bitmap))
+       __set_bit(NULLB_PAGE_FREE, t_page->bitmap);
+       if (test_bit(NULLB_PAGE_LOCK, t_page->bitmap))
                return;
        __free_page(t_page->page);
        kfree(t_page);
 }
 
+static bool null_page_empty(struct nullb_page *page)
+{
+       int size = MAP_SZ - 2;
+
+       return find_first_bit(page->bitmap, size) == size;
+}
+
 static void null_free_sector(struct nullb *nullb, sector_t sector,
        bool is_cache)
 {
@@ -764,9 +772,9 @@ static void null_free_sector(struct nullb *nullb, sector_t sector,
 
        t_page = radix_tree_lookup(root, idx);
        if (t_page) {
-               __clear_bit(sector_bit, &t_page->bitmap);
+               __clear_bit(sector_bit, t_page->bitmap);
 
-               if (!t_page->bitmap) {
+               if (null_page_empty(t_page)) {
                        ret = radix_tree_delete_item(root, idx, t_page);
                        WARN_ON(ret != t_page);
                        null_free_page(ret);
@@ -837,7 +845,7 @@ static struct nullb_page *__null_lookup_page(struct nullb *nullb,
        t_page = radix_tree_lookup(root, idx);
        WARN_ON(t_page && t_page->page->index != idx);
 
-       if (t_page && (for_write || test_bit(sector_bit, &t_page->bitmap)))
+       if (t_page && (for_write || test_bit(sector_bit, t_page->bitmap)))
                return t_page;
 
        return NULL;
@@ -900,10 +908,10 @@ static int null_flush_cache_page(struct nullb *nullb, struct nullb_page *c_page)
 
        t_page = null_insert_page(nullb, idx << PAGE_SECTORS_SHIFT, true);
 
-       __clear_bit(NULLB_PAGE_LOCK, &c_page->bitmap);
-       if (test_bit(NULLB_PAGE_FREE, &c_page->bitmap)) {
+       __clear_bit(NULLB_PAGE_LOCK, c_page->bitmap);
+       if (test_bit(NULLB_PAGE_FREE, c_page->bitmap)) {
                null_free_page(c_page);
-               if (t_page && t_page->bitmap == 0) {
+               if (t_page && null_page_empty(t_page)) {
                        ret = radix_tree_delete_item(&nullb->dev->data,
                                idx, t_page);
                        null_free_page(t_page);
@@ -919,11 +927,11 @@ static int null_flush_cache_page(struct nullb *nullb, struct nullb_page *c_page)
 
        for (i = 0; i < PAGE_SECTORS;
                        i += (nullb->dev->blocksize >> SECTOR_SHIFT)) {
-               if (test_bit(i, &c_page->bitmap)) {
+               if (test_bit(i, c_page->bitmap)) {
                        offset = (i << SECTOR_SHIFT);
                        memcpy(dst + offset, src + offset,
                                nullb->dev->blocksize);
-                       __set_bit(i, &t_page->bitmap);
+                       __set_bit(i, t_page->bitmap);
                }
        }
 
@@ -960,10 +968,10 @@ again:
                 * We found the page which is being flushed to disk by other
                 * threads
                 */
-               if (test_bit(NULLB_PAGE_LOCK, &c_pages[i]->bitmap))
+               if (test_bit(NULLB_PAGE_LOCK, c_pages[i]->bitmap))
                        c_pages[i] = NULL;
                else
-                       __set_bit(NULLB_PAGE_LOCK, &c_pages[i]->bitmap);
+                       __set_bit(NULLB_PAGE_LOCK, c_pages[i]->bitmap);
        }
 
        one_round = 0;
@@ -1016,7 +1024,7 @@ static int copy_to_nullb(struct nullb *nullb, struct page *source,
                kunmap_atomic(dst);
                kunmap_atomic(src);
 
-               __set_bit(sector & SECTOR_MASK, &t_page->bitmap);
+               __set_bit(sector & SECTOR_MASK, t_page->bitmap);
 
                if (is_fua)
                        null_free_sector(nullb, sector, true);
@@ -1846,10 +1854,6 @@ static int __init null_init(void)
        struct nullb *nullb;
        struct nullb_device *dev;
 
-       /* check for nullb_page.bitmap */
-       if (sizeof(unsigned long) * 8 - 2 < (PAGE_SIZE >> SECTOR_SHIFT))
-               return -EINVAL;
-
        if (g_bs > PAGE_SIZE) {
                pr_warn("null_blk: invalid block size\n");
                pr_warn("null_blk: defaults block size to %lu\n", PAGE_SIZE);