@@ -1358,14 +1358,6 @@ static int __wait_on_page_locked_async(struct page *page,
return ret;
}
-static int wait_on_page_locked_async(struct page *page,
- struct wait_page_queue *wait)
-{
- if (!PageLocked(page))
- return 0;
- return __wait_on_page_locked_async(compound_head(page), wait, false);
-}
-
/**
* put_and_wait_on_page_locked - Drop a reference and wait for it to be unlocked
* @page: The page to wait for.
@@ -2259,6 +2251,7 @@ static struct page *gfbr_read_page(struct kiocb *iocb,
return error != AOP_TRUNCATED_PAGE ? ERR_PTR(error) : NULL;
}
+ page = thp_head(page);
if (!PageUptodate(page)) {
error = lock_page_for_iocb(iocb, page);
if (unlikely(error)) {
@@ -2292,64 +2285,42 @@ static struct page *gfbr_update_page(struct kiocb *iocb,
struct inode *inode = mapping->host;
int error;
- /*
- * See comment in do_read_cache_page on why
- * wait_on_page_locked is used to avoid unnecessarily
- * serialisations and why it's safe.
- */
if (iocb->ki_flags & IOCB_WAITQ) {
- error = wait_on_page_locked_async(page,
- iocb->ki_waitq);
- } else {
- error = wait_on_page_locked_killable(page);
- }
- if (unlikely(error)) {
- put_page(page);
- return ERR_PTR(error);
+ error = lock_page_async(page, iocb->ki_waitq);
+ if (error) {
+ put_page(page);
+ return ERR_PTR(error);
+ }
+ } else if (!trylock_page(page)) {
+ put_and_wait_on_page_locked(page, TASK_KILLABLE);
+ return NULL;
}
+
if (PageUptodate(page))
- return page;
+ goto uptodate;
if (inode->i_blkbits == PAGE_SHIFT ||
!mapping->a_ops->is_partially_uptodate)
- goto page_not_up_to_date;
+ goto readpage;
/* pipes can't handle partially uptodate pages */
if (unlikely(iov_iter_is_pipe(iter)))
- goto page_not_up_to_date;
- if (!trylock_page(page))
- goto page_not_up_to_date;
- /* Did it get truncated before we got the lock? */
+ goto readpage;
if (!page->mapping)
- goto page_not_up_to_date_locked;
+ goto truncated;
if (!mapping->a_ops->is_partially_uptodate(page,
- pos & ~PAGE_MASK, count))
- goto page_not_up_to_date_locked;
+ pos & (thp_size(page) - 1), count))
+ goto readpage;
+uptodate:
unlock_page(page);
return page;
-page_not_up_to_date:
- /* Get exclusive access to the page ... */
- error = lock_page_for_iocb(iocb, page);
- if (unlikely(error)) {
- put_page(page);
- return ERR_PTR(error);
- }
-
-page_not_up_to_date_locked:
- /* Did it get truncated before we got the lock? */
- if (!page->mapping) {
- unlock_page(page);
- put_page(page);
- return NULL;
- }
-
- /* Did somebody else fill it already? */
- if (PageUptodate(page)) {
- unlock_page(page);
- return page;
- }
-
+readpage:
+ page += (pos / PAGE_SIZE) - page->index;
return gfbr_read_page(iocb, mapping, page);
+truncated:
+ unlock_page(page);
+ put_page(page);
+ return NULL;
}
static struct page *gfbr_create_page(struct kiocb *iocb,
@@ -2443,6 +2414,7 @@ static int gfbr_get_pages(struct kiocb *iocb, struct iov_iter *iter,
err = PTR_ERR_OR_ZERO(page);
break;
}
+ pages[i] = page;
}
}
We need to tell readpage which subpage we're actually interested in (by passing the subpage to gfbr_read_page()), and if it does split the THP, we need to update the page in the page array to be the subpage. For page splitting to succeed, the thread asking to split the page has to be the only one with a reference to the page. Calling wait_on_page_locked() while holding a reference to the page will effectively prevent this from happening with sufficient threads waiting on the same page. Use put_and_wait_on_page_locked() to sleep without holding a reference to the page, then retry the page lookup after the page is unlocked. Since we now get the page lock a little earlier in gfbr_update_page(), we can eliminate a number of duplicate checks. The original intent (commit ebded02788b5 ("avoid unnecessary calls to lock_page when waiting for IO to complete during a read") behind getting the page lock later was to avoid re-locking the page after it has been brought uptodate by another thread. We will still avoid that because we go through the normal lookup path again after the winning thread has brought the page uptodate. Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org> --- mm/filemap.c | 76 +++++++++++++++++----------------------------------- 1 file changed, 24 insertions(+), 52 deletions(-)