mm/page_alloc.c: fix freeing non-compound pages
authorMatthew Wilcox (Oracle) <willy@infradead.org>
Tue, 13 Oct 2020 23:56:04 +0000 (16:56 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Wed, 14 Oct 2020 01:38:33 +0000 (18:38 -0700)
commite320d3012d25b1fb5f3df4edb7bd44a1c362ec10
tree045231cbb1f31c38cbb22189a08d284398fc09ae
parenta9b576f7253e22528584f5aeb46edf0b6b007992
mm/page_alloc.c: fix freeing non-compound pages

Here is a very rare race which leaks memory:

Page P0 is allocated to the page cache.  Page P1 is free.

Thread A                Thread B                Thread C
find_get_entry():
xas_load() returns P0
Removes P0 from page cache
P0 finds its buddy P1
alloc_pages(GFP_KERNEL, 1) returns P0
P0 has refcount 1
page_cache_get_speculative(P0)
P0 has refcount 2
__free_pages(P0)
P0 has refcount 1
put_page(P0)
P1 is not freed

Fix this by freeing all the pages in __free_pages() that won't be freed
by the call to put_page().  It's usually not a good idea to split a page,
but this is a very unlikely scenario.

Fixes: e286781d5f2e ("mm: speculative page references")
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nick Piggin <npiggin@gmail.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20200926213919.26642-1-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
lib/Kconfig.debug
lib/Makefile
lib/test_free_pages.c [new file with mode: 0644]
mm/page_alloc.c