Re: [PATCH] Avoid fragment virtio-blk transfers by copying

[Avi Kivity <avi@qumranet.com>]

Marcelo Tosatti wrote:
> On Wed, Jun 25, 2008 at 01:44:35PM +0300, Avi Kivity wrote:
>   
>> Anthony Liguori wrote:
>>     
>>> A major source of performance loss for virtio-blk has been the fact that we
>>> split transfers into multiple requests.  This is particularly harmful if you
>>> have striped storage beneath your virtual machine.
>>>
>>> This patch copies the request data into a single contiguous buffer to ensure
>>> that we don't split requests.  This improves performance from about 80 MB/sec
>>> to about 155 MB/sec with my fibre channel link.  185 MB/sec is what we get on
>>> native so this gets us pretty darn close.
>>>
>>>   
>>>       
>> If the guest issues a request for a terabyte of memory, the host will  
>> try to allocate it and drop to swap/oom.  So we need to either fragment  
>> beyond some size, or to avoid copying and thus the need for allocation.
>>     
>
> The maximum request size for Linux guests is 512K (after tuning
> virtio-blk guest driver, current max is 124K). I'm not sure what the max
> number of requests is, but I guess is between 128 and 1024, Anthony?
>
>   

This is assuming 1 page per descriptor.  But the len field is 32 bits, 
so each descriptor can address up to 4GB.

> Point is that the guest is responsible for limiting the amount of data
> in-flight. A malicious guest can only hurt itself by attempting to DoS
> the host, with proper memory limits in place. IMO this issue should not
> be handled in the virtio-blk backend

I'd like kvm to be safe for use even without the rss controller.

-- 
error compiling committee.c: too many arguments to function