Re: [Qemu-devel] [RFC Patch 0/7]Qemu: Dynamic host pagecache change

[Kevin Wolf <kwolf@redhat.com>]

Am 01.02.2012 23:41, schrieb Eric Blake:
> On 01/31/2012 08:05 PM, Supriya Kannery wrote:
>>     For changing host pagecache setting of a running VM, it is
>> important to have a safe way of reopening its image file.
>>
>> Following patchset introduces:
>>  * a generic way to reopen image files safely. 
>>         In this approach, before reopening an image, for each
>>     block driver, its state will be stashed. Incase preparation
>>     (bdrv_reopen_prepare) for reopening returns success, the stashed 
>>     state will be cleared (bdrv_reopen_commit) and reopened state will 
>>     be used further. Incase preparation of reopening returns failure, 
>>     the state of the driver will be rolled back (bdrv_reopen_abort) 
>>     to the stashed state.
>>         This approach is extended to raw-posix, raw-win32 and vmdk block 
>>     drivers in this patchset. Once this is reviewed and finalised, I will 
>>     extend the implementation to other drivers like qcow2, qed etc..
> 
> How will this interplay with the goal of passing images in by fd rather
> than by name?  I'd really like to start thinking about how we plan on
> coordinating situations where an fd has to be reopened in order to
> switch flags (such as O_RDONLY becoming O_RDWR, or adding or subtracting
> O_DIRECT), but where SELinux or other isolation means that the
> management app (such as libvirt) has to do the open and pass the fd via
> 'getfd' monitor command.

That's easy: Either your host OS allows to change the respective flag
using fcntl(), which I believe is true for O_DIRECT/O_DSYNC and recent
Linux kernels, or you're out of luck.

Is there any reason for switching between rw/ro other than modifying the
backing file chain (i.e. creating or deleting a snapshot)?

Kevin