qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] [PATCH] block/nfs: add knob to set readahead


From: Eric Blake
Subject: Re: [Qemu-devel] [PATCH] block/nfs: add knob to set readahead
Date: Mon, 23 Jun 2014 15:05:07 -0600
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:24.0) Gecko/20100101 Thunderbird/24.6.0

On 06/23/2014 02:47 PM, Peter Lieven wrote:

>>> +#ifdef LIBNFS_FEATURE_READAHEAD
>>> +        } else if (!strcmp(qp->p[i].name, "readahead")) {
>>> +            nfs_set_readahead(client->context, atoi(qp->p[i].value));
>>> +#endif
>>
>> I'm not a fan of adding even more special-casing to the file-name URI
>> without also adding it to the QAPI schema for use in the blockdev-add
>> command.  Of course, we don't have structured nfs options for
>> blockdev-add yet, but maybe it's time to start thinking about that
>> addition before we do this addition.
> 
> I would like to leave this for a follow-up and I promise to take care of this.
> If you could give me a pointer of an existing driver that does this as a 
> reference
> I would be grateful.

A good example of a recent conversion to structured options would be the
blkdebug and blkverify backends (look around commit 1bf20b82), or
proposed addition of new backends (such as archipelego,
https://lists.gnu.org/archive/html/qemu-devel/2014-06/msg05309.html).
It may be a bigger task than you anticipate, and certainly won't make
the 2.1 timeframe, but it's worth doing so if you're up to the task,
more power to you :)

> 
> For the read ahead parameter its likely not needed as (at least in my use 
> case)
> the read ahead makes only sense when converting a compressed QCOW2 Template 
> File
> which lives on an NFS Share to a RAW Device. So I use it with qemu-img.
> For a Container File that is used as a VM hard drive I would likely not use 
> read ahead
> as the operating system itself will implement some sort of read ahead already.

Even if qemu-img is likely to be the only one ever specifying the
option, it still makes sense to expose it via QMP, as we are gradually
trying to move qemu-img over to more exclusive use of QMP (or at least
the underlying functions reached by QMP) rather than ad-hoc code.  For
example,
https://lists.gnu.org/archive/html/qemu-devel/2014-06/msg01822.html).

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org

Attachment: signature.asc
Description: OpenPGP digital signature


reply via email to

[Prev in Thread] Current Thread [Next in Thread]