Re: [PATCH PROTOTYPE 4/6] memory: Extend ram_block_discard_(require|disable) by two discard types

[David Hildenbrand <david@redhat.com>]

On 20.10.20 21:17, Peter Xu wrote:
> On Thu, Sep 24, 2020 at 06:04:21PM +0200, David Hildenbrand wrote:
>> We want to separate the two cases whereby
>> - balloning drivers do random discards on random guest memory (e.g.,
>>   virtio-balloon) - uncoordinated discards
>> - paravirtualized memory devices do discards in well-known granularity,
>>   and always know which block is currently accessible or inaccessible by
>>   a guest. - coordinated discards
>>
>> This will be required to get virtio_mem + vfio running - vfio still
>> wants to block random memory ballooning.
>>
>> Cc: Paolo Bonzini <pbonzini@redhat.com>
>> Cc: "Michael S. Tsirkin" <mst@redhat.com>
>> Cc: Alex Williamson <alex.williamson@redhat.com>
>> Cc: Wei Yang <richardw.yang@linux.intel.com>
>> Cc: Dr. David Alan Gilbert <dgilbert@redhat.com>
>> Cc: Igor Mammedov <imammedo@redhat.com>
>> Cc: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
>> Cc: Peter Xu <peterx@redhat.com>
>> Signed-off-by: David Hildenbrand <david@redhat.com>
>> ---
>>  exec.c                | 109 ++++++++++++++++++++++++++++++++++--------
>>  include/exec/memory.h |  36 ++++++++++++--
>>  2 files changed, 121 insertions(+), 24 deletions(-)
>>
>> diff --git a/exec.c b/exec.c
>> index e34b602bdf..83098e9230 100644
>> --- a/exec.c
>> +++ b/exec.c
>> @@ -4098,52 +4098,121 @@ void mtree_print_dispatch(AddressSpaceDispatch *d, MemoryRegion *root)
>>   * If positive, discarding RAM is disabled. If negative, discarding RAM is
>>   * required to work and cannot be disabled.
>>   */
>> -static int ram_block_discard_disabled;
>> +static int uncoordinated_discard_disabled;
>> +static int coordinated_discard_disabled;
> 
> Instead of duplicating the codes, how about start to make it an array?
> 
> Btw, iiuc these flags do not need atomic operations at all, because all callers
> should be with BQL and called majorly during machine start/reset.  Even not, I
> think we can also use a mutex, maybe it could make things simpler.  No strong
> opinion, though.
> 

I remember there were some !BQL users (but I might be confusing it with
postcopy code that once used to inhibit the balloon without BQL). Will
double-check. Simplifying it is certainly a good idea.

(we want to be able to check from virtio-balloon code repeatedly without
taking a mutex over and over again :) )

Thanks!

-- 
Thanks,

David / dhildenb