On 11/22/2017 04:00 AM, Johannes Weiner wrote:
On Mon, Nov 20, 2017 at 11:39:30AM -0800, Mike Kravetz wrote:
If the call __alloc_contig_migrate_range() in alloc_contig_range returns -EBUSY, processing continues so that test_pages_isolated() is called where there is a tracepoint to identify the busy pages. However, it is possible for busy pages to become available between the calls to these two routines. In this case, the range of pages may be allocated. Unfortunately, the original return code (ret == -EBUSY) is still set and returned to the caller. Therefore, the caller believes the pages were not allocated and they are leaked.
Update the return code with the value from test_pages_isolated().
Fixes: 8ef5849fa8a2 ("mm/cma: always check which page caused allocation failure") Cc: stable@vger.kernel.org Signed-off-by: Mike Kravetz mike.kravetz@oracle.com
Wow, good catch.
mm/page_alloc.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/mm/page_alloc.c b/mm/page_alloc.c index 77e4d3c5c57b..3605ca82fd29 100644 --- a/mm/page_alloc.c +++ b/mm/page_alloc.c @@ -7632,10 +7632,10 @@ int alloc_contig_range(unsigned long start, unsigned long end, } /* Make sure the range is really isolated. */
- if (test_pages_isolated(outer_start, end, false)) {
- ret = test_pages_isolated(outer_start, end, false);
- if (ret) { pr_info_ratelimited("%s: [%lx, %lx) PFNs busy\n", __func__, outer_start, end);
ret = -EBUSY;Essentially, an -EBUSY from __alloc_contig_migrate_range() doesn't mean anything, and we return 0 if the rest of the operations succeed.
Since we never plan on returning that particular -EBUSY, would it be more robust to reset it right then and there, rather than letting it run on in ret for more than a screenful?
It would also be good to note in that fall-through comment that the pages becoming free on their own is a distinct possibility.
As Michal points out, this is really subtle. It makes sense to make it as explicit as possible.
Ok, I thought about zero'ing ret right after the call to __alloc_contig_migrate_range and return of -EBUSY. It just didn't look right to me. But, you are correct. We should make this as explicit as possible. I will respin the patch as suggested and be sure to include an explicit comment when setting ret = 0.