Commit 05b7b7c7 authored Oct 12, 2018 by Zhenhua Huang

mm: give up CPU when swap cache in inconsistence state



The case is for __read_swap_cache_async():
If swap cache is deleted by __delete_from_swap_cache(), but
SWAP_HAS_CACHE flag not cleared yet by swapcache_free(). eg,
when sched out in delete_from_swap_cache(), or __remove_mapping().
In the mean time, current process unfortunately is RT,
then cond_sched() may not working. Introduce using usleep(500, 1000)
to give up CPU to another task.

CRs-Fixed: 2285534
Change-Id: Icf55fc2b01d95a552d3f3651c2d78d115fd17b1f
Signed-off-by: Zhenhua Huang <zhenhuah@codeaurora.org>

parent f5feb668

mm/swap_state.c

+5 −1

Original line number	Diff line number	Diff line
		@@ -17,6 +17,7 @@
		#include <linux/blkdev.h>
		#include <linux/pagevec.h>
		#include <linux/migrate.h>
		#include <linux/delay.h>

		#include <asm/pgtable.h>

		@@ -350,8 +351,11 @@ struct page *__read_swap_cache_async(swp_entry_t entry, gfp_t gfp_mask,
		* busy looping, we just conditionally invoke the
		* scheduler here, if there are some more important
		* tasks to run.
		*
		* cond_resched() may not work if the process is RT.
		* We need a usleep_range() give up CPU to another task.
		*/
		cond_resched();
		usleep_range(500, 1000);
		continue;
		}
		if (err) { /* swp entry is obsolete ? */