Re: [patch 12/12] mm: correctly synchronize rss-counters at exit/exec

[Oleg Nesterov <oleg@redhat.com>]

On 06/08, Konstantin Khlebnikov wrote:
>
> As result you can see "BUG: Bad rss-counter state mm:ffff88040783a680 idx:1 val:-1" in dmesg
>
> There left only one problem: nobody calls sync_mm_rss() after put_user() in mm_release().

Both callers call sync_mm_rss() to make check_mm() happy. But please
see the changelog, I think we should move it into mm_release(). See
the patch below (on top of v2 I sent). I need to recheck.

As for xacct_add_tsk(), yes it can "miss" that put_user(). But this
is what we have now, I think we do not care.

Oleg.

--- x/fs/exec.c
+++ x/fs/exec.c
@@ -822,7 +822,6 @@ static int exec_mmap(struct mm_struct *m
 	mm_release(tsk, old_mm);
 
 	if (old_mm) {
-		sync_mm_rss(old_mm);
 		/*
 		 * Make sure that if there is a core dump in progress
 		 * for the old mm, we get out and die instead of going
--- x/kernel/exit.c
+++ x/kernel/exit.c
@@ -656,7 +656,6 @@ static void exit_mm(struct task_struct *
 	if (!mm)
 		return;
 
-	sync_mm_rss(mm);
 	/*
 	 * Serialize with any possible pending coredump.
 	 * We must hold mmap_sem around checking core_state
--- x/kernel/taskstats.c
+++ x/kernel/taskstats.c
@@ -630,8 +630,7 @@ void taskstats_exit(struct task_struct *
 	if (!stats)
 		goto err;
 
-	if (tsk->mm)
-		sync_mm_rss(tsk->mm);
+	sync_mm_rss(tsk->mm);
 	fill_stats(tsk, stats);
 
 	/*
--- x/kernel/exit.c
+++ x/kernel/exit.c
@@ -656,7 +656,6 @@ static void exit_mm(struct task_struct *
 	if (!mm)
 		return;
 
-	sync_mm_rss(mm);
 	/*
 	 * Serialize with any possible pending coredump.
 	 * We must hold mmap_sem around checking core_state

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>