+static void dio_unlock_page_range(struct dio *dio)
+{
+ if (dio->lock_type != DIO_NO_LOCKING) {
+ remove_placeholder_pages(dio->inode->i_mapping,
+ dio->fspages_start_off,
+ dio->fspages_end_off);
+ dio->fspages_end_off = dio->fspages_start_off;
I wonder if it should call remove_placeholder_pages when _end_off !=
_start_off. That feels like a more direct test.
+ max_size = (dio->fspages_end_off - index) << PAGE_CACHE_SHIFT;
+ if (map_bh->b_size > max_size)
+ map_bh->b_size = max_size;
b_size = min(b_size, math) ?
+ /* if the mapping is mapped, they may be using a mmap'd portion
+ * of the file as the buffer for this io. That will deadlock
+ * with placeholders because the placeholder code forces the
+ * page fault handler to block. The (ugly) solution is to
+ * limit the span of inserted placeholders to the same
+ * increment we use for get_user_pages.
+ */
+ if (inode->i_mapping->nrpages ||
+ mapping_mapped(inode->i_mapping))
+ dio->fspages_span = DIO_PAGES;
+ else
+ dio->fspages_span = ULONG_MAX;
Couldn't the decision to use DIO_PAGES in the presence of a mapping
race with establishing a mapping? I fear we have to use DIO_PAGES
all the time, which at least has the benefit of getting rid of the
_span member of dio :).
- z
-
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
