Re: [Qemu-discuss] QEMU caching modes

[Jan Schermer]

It should be.
But even if it is, the data _won't_ be coherent because write ordering will be 
non-existent. This will mess up most filesystems.

Jan

> On 20 Mar 2017, at 22:01, Narcis Garcia <address@hidden> wrote:
> 
> When cache=unsafe , is the data flushed as configured in host's Linux
> kernel?
> dirty_expire_centisecs
> dirty_writeback_centisecs
> 
> El 20/03/17 a les 17:21, Alberto Garcia ha escrit:
>> On Fri, Mar 17, 2017 at 09:52:27PM +0100, Jan Schermer wrote:
>> 
>>> I'm trying to understand when/what IO is synchronous and why, and so
>>> far what I'm seeing I can only describe as "weird".
>>> 
>>> With cache=none, all IO should pass from the QEMU process to the device 
>>> as-is. Right?
>> 
>> Here's the summary of what the 'cache' option does in practice:
>> 
>> cache=none         ->   direct I/O (the host page cache is bypassed).
>>                        See O_DIRECT in open(2)
>> 
>> cache=directsync   ->   direct I/O + flush after each write
>> 
>> cache=writethrough ->   flush after each write
>> 
>> cache=unsafe       ->   guests flushes are ignored
>> 
>> cache=writeback    ->   no special options (this is the default)
>> 
>>> If so, that's not what I see - I see (with btrace) write IO as
>>> asynchronous (W) in the guest, but synchronous (WS) on the host.
>> 
>> Looks like what you're seeing is the direct I/O.
>> 
>> Berto
>> 
>