Re: [PATCH] iommu/iova: silence warnings under memory pressure

[Qian Cai <cai@lca.pw>]

On Fri, 2019-11-15 at 13:13 -0800, Joe Perches wrote:
> On Fri, 2019-11-15 at 15:50 -0500, Qian Cai wrote:
> > When running heavy memory pressure workloads, this 5+ old system is
> > throwing endless warnings below because disk IO is too slow to recover
> > from swapping. Since the volume from alloc_iova_fast() could be large,
> > once it calls printk(), it will trigger disk IO (writing to the log
> > files) and pending softirqs which could cause a loop and no progress
> > from memory reclaim for days.
> 
> []
> > diff --git a/drivers/iommu/iova.c b/drivers/iommu/iova.c
> 
> []
> > @@ -233,7 +233,7 @@ static int __alloc_and_insert_iova_range(struct iova_domain *iovad,
> >  
> >  struct iova *alloc_iova_mem(void)
> >  {
> > -	return kmem_cache_alloc(iova_cache, GFP_ATOMIC);
> > +	return kmem_cache_alloc(iova_cache, GFP_ATOMIC | __GFP_NOWARN);
> >  }
> >  EXPORT_SYMBOL(alloc_iova_mem);
> 
> Is notification ever useful?
> 
> If so, maybe something like:
> 
> struct iova *alloc_iova_mem(void)
> {
> 	void *mem = kmem_cache_alloc(iova_cache, GFP_ATOMIC | __GFP_NOWARN)_
> 
> 	WARN_RATELIMIT(!mem, "%s: unable to alloc cache\n", __func__);
> 
> 	return mem;
> }
> 
> or maybe use printk_deferred or prink_deferred_once
> 
> ?
> 

Forgot to mentioned that errors are also reported in hpsa driver which is fine.

hpsa 0000:03:00.0: DMAR: Allocating 1-page iova failed

but warn_alloc() is way too expense for this old server.