[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <YwOQ5Q+aZRBB0xk+@dhcp22.suse.cz>
Date: Mon, 22 Aug 2022 16:21:25 +0200
From: Michal Hocko <mhocko@...e.com>
To: Christoph Hellwig <hch@....de>
Cc: iommu@...ts.linux.dev, Marek Szyprowski <m.szyprowski@...sung.com>,
Robin Murphy <robin.murphy@....com>,
David Rientjes <rientjes@...gle.com>,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH 3/3] dma-pool: limit DMA and DMA32 zone size pools
On Mon 22-08-22 14:27:59, Christoph Hellwig wrote:
> On Mon, Aug 22, 2022 at 02:09:13PM +0200, Michal Hocko wrote:
> > OK, so you are worried about configurations that lack ZONE_DMA32 because
> > that memory range is covered by ZONE_DMA? I thought though simply do not
> > have CONFIG_ZONE_DMA32. Or am I wrong here?
>
> Zones can disappear when they are empty. Normally that affects
> ZONE_NORMAL and ZONE_HIGHMEM, but it can also happen to ZONE_DMA32
> when there is not enough memory.
Correct but that is essentially the same problem as ZONE_DMA becoming
empty. Or maybe I am just completely missing what you are trying to
say/ask?
My whole point was that if those zones are configured and you cannot
pre-allocate the pool then this can be either a result of OOM on that
zone or zone completely empty. In either case the pool would be empty
and if that is something worth bothering^Wnotifying an admin about then
better provide something useful rather than over messaging for one case
and be silent about the other. See my point?
--
Michal Hocko
SUSE Labs
Powered by blists - more mailing lists