From: John Garry <john.garry@huawei.com>
To: Robin Murphy <robin.murphy@arm.com>, <joro@8bytes.org>
Cc: <will.deacon@arm.com>, <linux-kernel@vger.kernel.org>,
<iommu@lists.linux-foundation.org>,
<ganapatrao.kulkarni@cavium.com>, <hch@lst.de>,
<m.szyprowski@samsung.com>
Subject: Re: [PATCH v2] iommu/dma: Use NUMA aware memory allocations in __iommu_dma_alloc_pages()
Date: Tue, 20 Nov 2018 15:42:16 +0000 [thread overview]
Message-ID: <c8b9ae20-215c-4778-c3ca-cb27b6b27dff@huawei.com> (raw)
In-Reply-To: <f9a91b43-ae77-ced5-3afc-57366e5126cb@arm.com>
On 20/11/2018 14:20, Robin Murphy wrote:
> On 20/11/2018 13:42, John Garry wrote:
>> From: Ganapatrao Kulkarni <ganapatrao.kulkarni@cavium.com>
>>
>> Change function __iommu_dma_alloc_pages() to allocate memory/pages
>> for DMA from respective device NUMA node.
>>
>> Signed-off-by: Ganapatrao Kulkarni <ganapatrao.kulkarni@cavium.com>
>> [JPG: Modifed to use kvzalloc() and fixed indentation]
>> Signed-off-by: John Garry <john.garry@huawei.com>
>> ---
>> Difference v1->v2:
>> - Add Ganapatrao's tag and change author
>>
>> This patch was originally posted by Ganapatrao in [1].
>>
>> However, after initial review, it was never reposted (due to lack of
>> cycles, I think). In addition, the functionality in its sibling patches
>> were merged through patches, as mentioned in [2]; this also refers to a
>> discussion on device local allocations vs CPU local allocations for DMA
>> pool, and which is better [3].
>>
>> However, as mentioned in [3], dma_alloc_coherent() uses the locality
>> information from the device - as in direct DMA - so this patch is just
>> applying this same policy.
>>
>> [1] https://lore.kernel.org/patchwork/patch/833004/
>> [2] https://lkml.org/lkml/2018/8/22/391
>> [3]
>> https://www.mail-archive.com/linux-kernel@vger.kernel.org/msg1692998.html
>>
>>
>> diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
>> index d1b0475..ada00bc 100644
>> --- a/drivers/iommu/dma-iommu.c
>> +++ b/drivers/iommu/dma-iommu.c
>> @@ -449,20 +449,17 @@ static void __iommu_dma_free_pages(struct page
>> **pages, int count)
>> kvfree(pages);
>> }
>> -static struct page **__iommu_dma_alloc_pages(unsigned int count,
>> - unsigned long order_mask, gfp_t gfp)
>> +static struct page **__iommu_dma_alloc_pages(struct device *dev,
>> + unsigned int count, unsigned long order_mask, gfp_t gfp)
>> {
>> struct page **pages;
>> - unsigned int i = 0, array_size = count * sizeof(*pages);
>> + unsigned int i = 0, nid = dev_to_node(dev);
>> order_mask &= (2U << MAX_ORDER) - 1;
>> if (!order_mask)
>> return NULL;
>> - if (array_size <= PAGE_SIZE)
>> - pages = kzalloc(array_size, GFP_KERNEL);
>> - else
>> - pages = vzalloc(array_size);
>> + pages = kvzalloc_node(count * sizeof(*pages), GFP_KERNEL, nid);
>
> The pages array is only accessed by the CPU servicing the
> iommu_dma_alloc() call, and is usually freed again before that call even
> returns. It's certainly never touched by the device, so forcing it to a
> potentially non-local node doesn't make a great deal of sense.
Right, it seems sensible to not make this allocation include the
device-locality requirement, so can leave as is. However modifying to
use kvzalloc() would seem ok.
>
>> if (!pages)
>> return NULL;
>> @@ -483,8 +480,10 @@ static struct page
>> **__iommu_dma_alloc_pages(unsigned int count,
>> unsigned int order = __fls(order_mask);
>> order_size = 1U << order;
>> - page = alloc_pages((order_mask - order_size) ?
>> - gfp | __GFP_NORETRY : gfp, order);
>> + page = alloc_pages_node(nid,
>> + (order_mask - order_size) ?
>> + gfp | __GFP_NORETRY : gfp,
>> + order);
>
> If we're touching this, can we sort out that horrendous ternary? FWIW I
> found I have a local version of the original patch which I tweaked at
> the time, and apparently I reworked this hunk as below, which does seem
> somewhat nicer for the same diffstat.
>
> Robin.
>
>
> @@ -446,10 +443,12 @@ static struct page
> **__iommu_dma_alloc_pages(unsigned int count,
> for (order_mask &= (2U << __fls(count)) - 1;
> order_mask; order_mask &= ~order_size) {
> unsigned int order = __fls(order_mask);
> + gfp_t alloc_flags = gfp;
>
> order_size = 1U << order;
> - page = alloc_pages((order_mask - order_size) ?
> - gfp | __GFP_NORETRY : gfp,
> order);
> + if (order_size < order_mask)
> + alloc_flags |= __GFP_NORETRY;
Sure, this can be included
> + page = alloc_pages_node(nid, alloc_flags, order);
> if (!page)
> continue;
> if (!order)
>
> .
>
Cheers,
John
prev parent reply other threads:[~2018-11-20 15:42 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-11-20 13:42 [PATCH v2] iommu/dma: Use NUMA aware memory allocations in __iommu_dma_alloc_pages() John Garry
2018-11-20 14:20 ` Robin Murphy
2018-11-20 15:42 ` John Garry [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=c8b9ae20-215c-4778-c3ca-cb27b6b27dff@huawei.com \
--to=john.garry@huawei.com \
--cc=ganapatrao.kulkarni@cavium.com \
--cc=hch@lst.de \
--cc=iommu@lists.linux-foundation.org \
--cc=joro@8bytes.org \
--cc=linux-kernel@vger.kernel.org \
--cc=m.szyprowski@samsung.com \
--cc=robin.murphy@arm.com \
--cc=will.deacon@arm.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).