lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Tue, 12 Feb 2019 12:39:59 +0000
From:   Tariq Toukan <tariqt@...lanox.com>
To:     Eric Dumazet <eric.dumazet@...il.com>,
        Ilias Apalodimas <ilias.apalodimas@...aro.org>,
        Matthew Wilcox <willy@...radead.org>,
        "brouer@...hat.com" <brouer@...hat.com>
CC:     David Miller <davem@...emloft.net>,
        "toke@...hat.com" <toke@...hat.com>,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        "mgorman@...hsingularity.net" <mgorman@...hsingularity.net>,
        "linux-mm@...ck.org" <linux-mm@...ck.org>
Subject: Re: [RFC, PATCH] net: page_pool: Don't use page->private to store
 dma_addr_t



On 2/11/2019 7:14 PM, Eric Dumazet wrote:
> 
> 
> On 02/11/2019 12:53 AM, Tariq Toukan wrote:
>>
> 
>> Hi,
>>
>> It's great to use the struct page to store its dma mapping, but I am
>> worried about extensibility.
>> page_pool is evolving, and it would need several more per-page fields.
>> One of them would be pageref_bias, a planned optimization to reduce the
>> number of the costly atomic pageref operations (and replace existing
>> code in several drivers).
>>
> 
> But the point about pageref_bias is to place it in a different cache line than "struct page"
> 
> The major cost is having a cache line bouncing between producer and consumer.
> 

pageref_bias is meant to be dirtied only by the page requester, i.e. the 
NIC driver / page_pool.
All other components (basically, SKB release flow / put_page) should 
continue working with the atomic page_refcnt, and not dirty the 
pageref_bias.

However, what bothers me more is another issue.
The optimization doesn't cleanly combine with the new page_pool 
direction for maintaining a queue for "available" pages, as the put_page 
flow would need to read pageref_bias, asynchronously, and act accordingly.

The suggested hook in put_page (to catch the 2 -> 1 "biased refcnt" 
transition) causes a problem to the traditional pageref_bias idea, as it 
implies a new point in which the pageref_bias field is read 
*asynchronously*. This would risk missing the this critical 2 -> 1 
transition! Unless pageref_bias is atomic...


> pageref_bias means the producer only have to read the "struct page" and not dirty it
> in the case the page can be recycled.
> 
> 
> 
>> I would replace this dma field with a pointer to an extensible struct,
>> that would contain the dma mapping (and other stuff in the near future).
>> This pointer fits perfectly with the existing unsigned long private;
>> they can share the memory, for both 32- and 64-bits systems.
>>
>> The only downside is one more pointer de-reference. This should be perf
>> tested.
>> However, when introducing the page refcnt bias optimization into
>> page_pool, I believe the perf gain would be guaranteed.
> 
> Only in some cases perhaps (when the cache line can be dirtied without performance hit)
> 

Powered by blists - more mailing lists