On Wed, 27 Sep 2017 12:14:10 +0800 Yafang Shao [off-list ref] wrote:

2017-09-27 10:54 GMT+08:00 Andrew Morton [off-list ref]:

quoted

On Wed, 27 Sep 2017 09:38:21 +0800 Yafang Shao [off-list ref] wrote:

quoted

quoted

And existing scripts which do not do this will cease to work correctly,
no?

The existing scritpts won't work correctly. That's also what I have
worried before.

But under this condition, there's a error message generated by "sysctl
-w" to tell them the first setting was failure.
This error message may be a reminder to them that there are some
connections between background and direct limit, and should not set
arbitrary.
May that's better. I'm not sure.

Maybe we can leave the logic as-is and simply print a warning when an
illogical state exists.

You mean, just modified the  code as bellow ?
in function  domain_dirty_limits()
-        if (bg_thresh >= thresh)
+     if (bg_thresh >= thresh) {
+        pr_warn("vm direct limit should greater than background limit.\n");
          bg_thresh = thresh / 2;
+     }

Something like that.

will this generate lots of log ?

Well, it's one message per write to a procfs file, when that write
causes an errant state.  Sounds manageable?  It would be nice if we
could somehow help the operator to figure out that writing in a
different order will prevent the incorrect state (and hence the
warning).

--
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>