Linux IOMMU Development
 help / color / mirror / Atom feed
From: "yf.wang--- via iommu" <iommu@lists.linux-foundation.org>
To: <iommu@lists.linux-foundation.org>
Cc: wsd_upstream@mediatek.com, linux-kernel@vger.kernel.org,
	Libo.Kang@mediatek.com, yf.wang@mediatek.com,
	linux-mediatek@lists.infradead.org, Ning.Li@mediatek.com,
	matthias.bgg@gmail.com, stable@vger.kernel.org, will@kernel.org,
	linux-arm-kernel@lists.infradead.org
Subject: Re: [PATCH] iommu/iova: Free all CPU rcache for retry when iova alloc failure
Date: Fri, 4 Mar 2022 13:00:40 +0800	[thread overview]
Message-ID: <20220304050040.4599-1-yf.wang@mediatek.com> (raw)
In-Reply-To: <20220304044635.4273-1-yf.wang@mediatek.com>


[-- Attachment #1.1: Type: text/html, Size: 8202 bytes --]

[-- Attachment #1.2: Type: text/plain, Size: 6934 bytes --]

On Fri, 2022-03-04 at 12:46 +0800, yf.wang@mediatek.com wrote:
> From: Yunfei Wang <yf.wang@mediatek.com>
> 
> In alloc_iova_fast function, if an iova alloc request fail,
> it will free the iova ranges present in the percpu iova
> rcaches and free global iova rcache and then retry, but
> flushing CPU iova rcaches only for each online CPU, which
> will cause incomplete rcache cleaning, and iova rcaches of
> not online CPU cannot be flushed, because iova rcaches may
> also lead to fragmentation of iova space, so the next retry
> action may still be fail.
> 
> Based on the above, so need to flushing all iova rcaches
> for each possible CPU, use for_each_possible_cpu instead of
> for_each_online_cpu like in free_iova_rcaches function,
> so that all rcaches can be completely released to try
> replenishing IOVAs.
> 
> Signed-off-by: Yunfei Wang <yf.wang@mediatek.com>
> Cc: <stable@vger.kernel.org> # 5.4.*
> ---
>  drivers/iommu/iova.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/iommu/iova.c b/drivers/iommu/iova.c
> index b28c9435b898..5a0637cd7bc2 100644
> --- a/drivers/iommu/iova.c
> +++ b/drivers/iommu/iova.c
> @@ -460,7 +460,7 @@ alloc_iova_fast(struct iova_domain *iovad,
> unsigned long size,
>  
>  		/* Try replenishing IOVAs by flushing rcache. */
>  		flush_rcache = false;
> -		for_each_online_cpu(cpu)
> +		for_each_possible_cpu(cpu)
>  			free_cpu_cached_iovas(cpu, iovad);
>  		free_global_cached_iovas(iovad);
>  		goto retry;

The following is the actual stress test log dump: 
1. __alloc_and_insert_iova_range_fail\x1a
[name:iova&]__alloc_and_insert_iova_range_fail:  iovad:0xffffff80dc9e6008 {granule:0x1000, start_pfn:0x40000, dma_32bit_pfn:0x100000, max32_alloc_size:0x731f},size:0x731f

2. dump all iova nodes of rbtree, contain cached iova:
[name:iova&]dump_iova_rbtree start:
[name:iova&]index    iova            pfn_lo      size      pfn_hi
[name:iova&]   1 0xffffff80680f8f80  0x47000000  0x4c9000  0x474c8000
[name:iova&]   2 0xffffff813e69c140  0x47500000  0x100000  0x475ff000
  ...317 lines of log are omitted here...
[name:iova&] 319 0xffffff80c8cea540  0xffefc000  0x4000    0xffeff000
[name:iova&] 320 0xffffff80de3a14c0  0xfff00000  0x100000  0xfffff000
[name:iova&] 321 0xffffff80dc9e6060  0xfffffffffffff000 0x1000 0xfffffffffffff000
[name:iova&]dump_iova_rbtree done, iovad:0xffffff80dc9e6008, count:321, total_size:606964KB

3. for_each_possible_cpu dump all cached iovas by iova_magazine pfns:
cpu:*-1: online cpu, cpu:*-0: not online cpu
[name:iova&]dump_iova_list, for_each_possible_cpu iovad:0xffffff80dc9e6008

3.1 online cpu cached iovas:
[name:iova&]dump_cpu_cached_iovas, iovad:0xffffff80dc9e6008, cpu:0-1, cache:0, iova_size:0x1000
[name:iova&]mag    iova              pfn_lo      size    pfn_h
[name:iova&]0/2  0xffffff8089c93c40  0xb8aff000  0x1000  0xb8aff000
[name:iova&]1/2  0xffffff8072046c40  0x716cf000  0x1000  0x716cf000
... omit other online cpu cached iovas ...

3.2 not online cpu cached iovas\x1a
... omit other not online cpu cached iovas ...
[name:iova&]dump_cpu_cached_iovas, iovad:0xffffff80dc9e6008, cpu:7-0, cache:5, iova_size:0x20000
[name:iova&]mag    iova               pfn_lo      size    pfn_hi
[name:iova&]0/10  0xffffff82b728ea40  0xdfbe0000  0x20000  0xdfbff000
[name:iova&]1/10  0xffffff82b728e480  0xee1e0000  0x20000  0xee1ff000
[name:iova&]2/10  0xffffff81a9fc5440  0xf27c0000  0x20000  0xf27df000
[name:iova&]3/10  0xffffff816f213800  0xd7fe0000  0x20000  0xd7fff000
[name:iova&]4/10  0xffffff80aa3fbf00  0xdb0e0000  0x20000  0xdb0ff000
[name:iova&]5/10  0xffffff8109217b00  0x4a7e0000  0x20000  0x4a7ff000
[name:iova&]6/10  0xffffff8109217540  0xb55a0000  0x20000  0xb55bf000
[name:iova&]7/10  0xffffff82b728e580  0xf33c0000  0x20000  0xf33df000
[name:iova&]8/10  0xffffff8109217500  0x7b5e0000  0x20000  0x7b5ff000
[name:iova&]9/10  0xffffff81a9fc5a80  0xf2da0000  0x20000  0xf2dbf000
[name:iova&]dump_iova_list, done iovad:0xffffff80dc9e6008, cpu_mask:0xff(11111111), online_cpu_mask:(00111111)

4. alloc fail, flushing rcache and retry:
[name:iova&]alloc_iova_fast, flushing rcache and retry, iovad:0xffffff80dc9e6008, size:0x731f, limit_pfn:0xfffff

5. retry and alloc fail again.
[name:iova&]__alloc_and_insert_iova_range_fail, iovad:0xffffff80dc9e6008 {granule:0x1000, start_pfn:0x40000, dma_32bit_pfn:0x100000, max32_alloc_size:0x731f},size:0x731f

6. dump all iova nodes of rbtree, contain cached iova:
[name:iova&]dump_iova_rbtree start:
[name:iova&]index    iova                pfn_lo      size    pfn_hi
[name:iova&]   1 0xffffff80680f8f80 0x47000000 0x4c9000  0x474c8000    
[name:iova&]   2 0xffffff813e69c140 0x47500000 0x100000  0x475ff000    
... 215 lines of log are omitted here ...
[name:iova&] 217 0xffffff80c8cea540 0xffefc000 0x4000    0xffeff000    
[name:iova&] 218 0xffffff80de3a14c0 0xfff00000 0x100000  0xfffff000    
[name:iova&] 219 0xffffff80dc9e6060 0xfffffffffffff000 0x1000 0xfffffffffffff000
[name:iova&]dump_iova_rbtree done, iovad:0xffffff80dc9e6008, count:219, total_size:603640KB

7. for_each_possible_cpu dump all cached iovas by iova_magazine pfns:
cpu:*-1: online cpu, cpu:*-0: not online cpu
[name:iova&]dump_iova_list, for_each_possible_cpu iovad:0xffffff80dc9e6008

7.1 online cpu cached iovas empty\x1a
[name:iova&]dump_cpu_cached_iovas, iovad:0xffffff80dc9e6008, cpu:0-1, cache:0, iova_size:0x1000 empty
[name:iova&]dump_cpu_cached_iovas, iovad:0xffffff80dc9e6008, cpu:0-1, cache:1, iova_size:0x2000 empty
... omit other online cpus ...
[name:iova&]dump_cpu_cached_iovas, iovad:0xffffff80dc9e6008, cpu:5-1, cache:4, iova_size:0x10000 empty
[name:iova&]dump_cpu_cached_iovas, iovad:0xffffff80dc9e6008, cpu:5-1, cache:5, iova_size:0x20000 empty

7.1 not online cpu cached iovas\x1a still have cached iovas:
... omit other not online cpu cached iovas ...
[name:iova&]dump_cpu_cached_iovas, iovad:0xffffff80dc9e6008, cpu:7-0, cache:5, iova_size:0x20000
[name:iova&]mag    iova               pfn_lo      size    pfn_hi
[name:iova&]0/10  0xffffff82b728ea40  0xdfbe0000  0x20000  0xdfbff000
[name:iova&]1/10  0xffffff82b728e480  0xee1e0000  0x20000  0xee1ff000
[name:iova&]2/10  0xffffff81a9fc5440  0xf27c0000  0x20000  0xf27df000
[name:iova&]3/10  0xffffff816f213800  0xd7fe0000  0x20000  0xd7fff000
[name:iova&]4/10  0xffffff80aa3fbf00  0xdb0e0000  0x20000  0xdb0ff000
[name:iova&]5/10  0xffffff8109217b00  0x4a7e0000  0x20000  0x4a7ff000
[name:iova&]6/10  0xffffff8109217540  0xb55a0000  0x20000  0xb55bf000
[name:iova&]7/10  0xffffff82b728e580  0xf33c0000  0x20000  0xf33df000
[name:iova&]8/10  0xffffff8109217500  0x7b5e0000  0x20000  0x7b5ff000
[name:iova&]9/10  0xffffff81a9fc5a80  0xf2da0000  0x20000  0xf2dbf000
[name:iova&]dump_iova_list, done iovad:0xffffff80dc9e6008, cpu_mask:0xff(11111111), online_cpu_mask:0x3f(00111111)

Thanks,
Yunfei.

[-- Attachment #2: Type: text/plain, Size: 156 bytes --]

_______________________________________________
iommu mailing list
iommu@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/iommu

  reply	other threads:[~2022-03-04  5:06 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-03-04  4:46 [PATCH] iommu/iova: Free all CPU rcache for retry when iova alloc failure yf.wang--- via iommu
2022-03-04  5:00 ` yf.wang--- via iommu [this message]
2022-03-04  9:22 ` John Garry via iommu
2022-03-07  3:18   ` yf.wang--- via iommu
2022-03-04 14:03 ` Robin Murphy
2022-03-07  3:32   ` yf.wang--- via iommu
2022-03-07  3:49   ` yf.wang--- via iommu

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=20220304050040.4599-1-yf.wang@mediatek.com \
    --to=iommu@lists.linux-foundation.org \
    --cc=Libo.Kang@mediatek.com \
    --cc=Ning.Li@mediatek.com \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mediatek@lists.infradead.org \
    --cc=matthias.bgg@gmail.com \
    --cc=stable@vger.kernel.org \
    --cc=will@kernel.org \
    --cc=wsd_upstream@mediatek.com \
    --cc=yf.wang@mediatek.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