[LogFS] Fix bdev erases
Erases for block devices were always just emulated by writing 0xff.
Some time back the write was removed and only the page cache was
changed to 0xff. Superficialy a good idea with two problems:
1. Touching the page cache isn't necessary either.
2. However, writing out 0xff _is_ necessary for the journal. As the
journal is scanned linearly, an old non-overwritten commit entry
can be used on next mount and cause havoc.
This should fix both aspects.
diff --git a/fs/logfs/dev_bdev.c b/fs/logfs/dev_bdev.c
index 58a057b..9718c22 100644
--- a/fs/logfs/dev_bdev.c
+++ b/fs/logfs/dev_bdev.c
@@ -167,27 +167,91 @@
generic_unplug_device(bdev_get_queue(logfs_super(sb)->s_bdev));
}
-static int bdev_erase(struct super_block *sb, loff_t to, size_t len)
+
+static void erase_end_io(struct bio *bio, int err)
+{
+ const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
+ struct super_block *sb = bio->bi_private;
+ struct logfs_super *super = logfs_super(sb);
+
+ BUG_ON(!uptodate); /* FIXME: Retry io or write elsewhere */
+ BUG_ON(err);
+ BUG_ON(bio->bi_vcnt == 0);
+ bio_put(bio);
+ if (atomic_dec_and_test(&super->s_pending_writes))
+ wake_up(&wq);
+}
+
+static int do_erase(struct super_block *sb, u64 ofs, pgoff_t index,
+ size_t nr_pages)
{
struct logfs_super *super = logfs_super(sb);
- struct address_space *mapping = super->s_mapping_inode->i_mapping;
- struct page *page;
- pgoff_t index = to >> PAGE_SHIFT;
- int i, nr_pages = len >> PAGE_SHIFT;
+ struct bio *bio;
+ struct request_queue *q = bdev_get_queue(sb->s_bdev);
+ unsigned int max_pages = queue_max_hw_sectors(q) >> (PAGE_SHIFT - 9);
+ int i;
+
+ bio = bio_alloc(GFP_NOFS, max_pages);
+ BUG_ON(!bio); /* FIXME: handle this */
+
+ for (i = 0; i < nr_pages; i++) {
+ if (i >= max_pages) {
+ /* Block layer cannot split bios :( */
+ bio->bi_vcnt = i;
+ bio->bi_idx = 0;
+ bio->bi_size = i * PAGE_SIZE;
+ bio->bi_bdev = super->s_bdev;
+ bio->bi_sector = ofs >> 9;
+ bio->bi_private = sb;
+ bio->bi_end_io = erase_end_io;
+ atomic_inc(&super->s_pending_writes);
+ submit_bio(WRITE, bio);
+
+ ofs += i * PAGE_SIZE;
+ index += i;
+ nr_pages -= i;
+ i = 0;
+
+ bio = bio_alloc(GFP_NOFS, max_pages);
+ BUG_ON(!bio);
+ }
+ bio->bi_io_vec[i].bv_page = super->s_erase_page;
+ bio->bi_io_vec[i].bv_len = PAGE_SIZE;
+ bio->bi_io_vec[i].bv_offset = 0;
+ }
+ bio->bi_vcnt = nr_pages;
+ bio->bi_idx = 0;
+ bio->bi_size = nr_pages * PAGE_SIZE;
+ bio->bi_bdev = super->s_bdev;
+ bio->bi_sector = ofs >> 9;
+ bio->bi_private = sb;
+ bio->bi_end_io = erase_end_io;
+ atomic_inc(&super->s_pending_writes);
+ submit_bio(WRITE, bio);
+ return 0;
+}
+
+static int bdev_erase(struct super_block *sb, loff_t to, size_t len,
+ int ensure_write)
+{
+ struct logfs_super *super = logfs_super(sb);
BUG_ON(to & (PAGE_SIZE - 1));
BUG_ON(len & (PAGE_SIZE - 1));
- if (logfs_super(sb)->s_flags & LOGFS_SB_FLAG_RO)
+ if (super->s_flags & LOGFS_SB_FLAG_RO)
return -EROFS;
- for (i = 0; i < nr_pages; i++) {
- page = find_get_page(mapping, index + i);
- if (page) {
- memset(page_address(page), 0xFF, PAGE_SIZE);
- page_cache_release(page);
- }
+ if (ensure_write) {
+ /*
+ * Object store doesn't care whether erases happen or not.
+ * But for the journal they are required. Otherwise a scan
+ * can find an old commit entry and assume it is the current
+ * one, travelling back in time.
+ */
+ do_erase(sb, to, to >> PAGE_SHIFT, len >> PAGE_SHIFT);
}
+
return 0;
}