Re: [PATCH] mm/page_alloc: Avoid KERN_CONT uses in warn_alloc

[Michal Hocko]

On Tue 07-11-17 08:03:27, Joe Perches wrote:
> On Tue, 2017-11-07 at 16:43 +0100, Michal Hocko wrote:
> > On Tue 07-11-17 07:34:25, Joe Perches wrote:
> []
> > > I believe, but have not tested, that using a specific width
> > > as an argument to %*pb[l] will constrain the number of
> > > spaces before the '(null)' output in any NULL pointer use.
> > > 
> > > So how about a #define like
> > > 
> > > /*
> > >  * nodemask_pr_args is only used with a "%*pb[l]" format for a nodemask.
> > >  * A NULL nodemask uses 6 to emit "(null)" without leading spaces.
> > >  */
> > > #define nodemask_pr_args(maskp)			\
> > > 	(maskp) ? MAX_NUMNODES : 6,		\
> > > 	(maskp) ? (maskp)->bits : NULL
> > 
> > Why not -1 then?
> 
> I believe it's the field width and not the precision that
> needs to be set.

But the first of the two arguments is the field with specifier, not the
precision. /me confused...

Anyway, the following works as expected when printing the OOM report:
[   47.005321] mem_eater invoked oom-killer: gfp_mask=0x14280ca(GFP_HIGHUSER_MOVABLE|__GFP_ZERO), nodemask=(null), order=0, oom_score_adj=0
[   47.007183] mem_eater cpuset=/ mems_allowed=0-1
[   47.007829] CPU: 3 PID: 3223 Comm: mem_eater Tainted: G        W       4.13.0-pr1-dirty #11

I hope I haven't overlooked anything
---