Skip to content

Commit c36c04c

Browse files
johnhubbardtorvalds
authored andcommitted
Revert "mm/gup: small refactoring: simplify try_grab_page()"
This reverts commit 54d516b That commit did a refactoring that effectively combined fast and slow gup paths (again). And that was again incorrect, for two reasons: a) Fast gup and slow gup get reference counts on pages in different ways and with different goals: see Linus' writeup in commit cd1adf1 ("Revert "mm/gup: remove try_get_page(), call try_get_compound_head() directly""), and b) try_grab_compound_head() also has a specific check for "FOLL_LONGTERM && !is_pinned(page)", that assumes that the caller can fall back to slow gup. This resulted in new failures, as recently report by Will McVicker [1]. But (a) has problems too, even though they may not have been reported yet. So just revert this. Link: https://lore.kernel.org/r/20220131203504.3458775-1-willmcvicker@google.com [1] Fixes: 54d516b ("mm/gup: small refactoring: simplify try_grab_page()") Reported-and-tested-by: Will McVicker <willmcvicker@google.com> Cc: Christoph Hellwig <hch@lst.de> Cc: Minchan Kim <minchan@google.com> Cc: Matthew Wilcox <willy@infradead.org> Cc: Christian Borntraeger <borntraeger@de.ibm.com> Cc: Heiko Carstens <hca@linux.ibm.com> Cc: Vasily Gorbik <gor@linux.ibm.com> Cc: stable@vger.kernel.org # 5.15 Signed-off-by: John Hubbard <jhubbard@nvidia.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1 parent d394bb7 commit c36c04c

1 file changed

Lines changed: 30 additions & 5 deletions

File tree

mm/gup.c

Lines changed: 30 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -124,8 +124,8 @@ static inline struct page *try_get_compound_head(struct page *page, int refs)
124124
* considered failure, and furthermore, a likely bug in the caller, so a warning
125125
* is also emitted.
126126
*/
127-
struct page *try_grab_compound_head(struct page *page,
128-
int refs, unsigned int flags)
127+
__maybe_unused struct page *try_grab_compound_head(struct page *page,
128+
int refs, unsigned int flags)
129129
{
130130
if (flags & FOLL_GET)
131131
return try_get_compound_head(page, refs);
@@ -208,10 +208,35 @@ static void put_compound_head(struct page *page, int refs, unsigned int flags)
208208
*/
209209
bool __must_check try_grab_page(struct page *page, unsigned int flags)
210210
{
211-
if (!(flags & (FOLL_GET | FOLL_PIN)))
212-
return true;
211+
WARN_ON_ONCE((flags & (FOLL_GET | FOLL_PIN)) == (FOLL_GET | FOLL_PIN));
213212

214-
return try_grab_compound_head(page, 1, flags);
213+
if (flags & FOLL_GET)
214+
return try_get_page(page);
215+
else if (flags & FOLL_PIN) {
216+
int refs = 1;
217+
218+
page = compound_head(page);
219+
220+
if (WARN_ON_ONCE(page_ref_count(page) <= 0))
221+
return false;
222+
223+
if (hpage_pincount_available(page))
224+
hpage_pincount_add(page, 1);
225+
else
226+
refs = GUP_PIN_COUNTING_BIAS;
227+
228+
/*
229+
* Similar to try_grab_compound_head(): even if using the
230+
* hpage_pincount_add/_sub() routines, be sure to
231+
* *also* increment the normal page refcount field at least
232+
* once, so that the page really is pinned.
233+
*/
234+
page_ref_add(page, refs);
235+
236+
mod_node_page_state(page_pgdat(page), NR_FOLL_PIN_ACQUIRED, 1);
237+
}
238+
239+
return true;
215240
}
216241

217242
/**

0 commit comments

Comments
 (0)