On 16.10.20 11:38, Wei Yang wrote:
> On Mon, Oct 12, 2020 at 02:53:19PM +0200, David Hildenbrand wrote:
>> Currently, we do not support device block sizes that exceed the Linux
>> memory block size. For example, having a device block size of 1 GiB (e.g.,
>> gigantic pages in the hypervisor) won't work with 128 MiB Linux memory
>> blocks.
>>
>> Let's implement Big Block Mode (BBM), whereby we add/remove at least
>> one Linux memory block at a time. With a 1 GiB device block size, a Big
>> Block (BB) will cover 8 Linux memory blocks.
>>
>> We'll keep registering the online_page_callback machinery, it will be used
>> for safe memory hotunplug in BBM next.
>>
>> Note: BBM is properly prepared for variable-sized Linux memory
>> blocks that we might see in the future. So we won't care how many Linux
>> memory blocks a big block actually spans, and how the memory notifier is
>> called.
>>
>> Cc: "Michael S. Tsirkin" <m...@redhat.com>
>> Cc: Jason Wang <jasow...@redhat.com>
>> Cc: Pankaj Gupta <pankaj.gupta.li...@gmail.com>
>> Cc: Michal Hocko <mho...@kernel.org>
>> Cc: Oscar Salvador <osalva...@suse.de>
>> Cc: Wei Yang <richard.weiy...@linux.alibaba.com>
>> Cc: Andrew Morton <a...@linux-foundation.org>
>> Signed-off-by: David Hildenbrand <da...@redhat.com>
>> ---
>> drivers/virtio/virtio_mem.c | 484 ++++++++++++++++++++++++++++++------
>> 1 file changed, 402 insertions(+), 82 deletions(-)
>>
>> diff --git a/drivers/virtio/virtio_mem.c b/drivers/virtio/virtio_mem.c
>> index e68d0d99590c..4d396ef98a92 100644
>> --- a/drivers/virtio/virtio_mem.c
>> +++ b/drivers/virtio/virtio_mem.c
>> @@ -30,12 +30,18 @@ MODULE_PARM_DESC(unplug_online, "Try to unplug online
>> memory");
>> /*
>> * virtio-mem currently supports the following modes of operation:
>> *
>> - * * Sub Block Mode (SBM): A Linux memory block spans 1..X subblocks (SB).
>> The
>> + * * Sub Block Mode (SBM): A Linux memory block spans 2..X subblocks (SB).
>> The
>> * size of a Sub Block (SB) is determined based on the device block size,
>> the
>> * pageblock size, and the maximum allocation granularity of the buddy.
>> * Subblocks within a Linux memory block might either be plugged or
>> unplugged.
>> * Memory is added/removed to Linux MM in Linux memory block granularity.
>> *
>> + * * Big Block Mode (BBM): A Big Block (BB) spans 1..X Linux memory blocks.
>> + * Memory is added/removed to Linux MM in Big Block granularity.
>> + *
>> + * The mode is determined automatically based on the Linux memory block size
>> + * and the device block size.
>> + *
>> * User space / core MM (auto onlining) is responsible for onlining added
>> * Linux memory blocks - and for selecting a zone. Linux Memory Blocks are
>> * always onlined separately, and all memory within a Linux memory block is
>> @@ -61,6 +67,19 @@ enum virtio_mem_sbm_mb_state {
>> VIRTIO_MEM_SBM_MB_COUNT
>> };
>>
>> +/*
>> + * State of a Big Block (BB) in BBM, covering 1..X Linux memory blocks.
>> + */
>> +enum virtio_mem_bbm_bb_state {
>> + /* Unplugged, not added to Linux. Can be reused later. */
>> + VIRTIO_MEM_BBM_BB_UNUSED = 0,
>> + /* Plugged, not added to Linux. Error on add_memory(). */
>> + VIRTIO_MEM_BBM_BB_PLUGGED,
>> + /* Plugged and added to Linux. */
>> + VIRTIO_MEM_BBM_BB_ADDED,
>> + VIRTIO_MEM_BBM_BB_COUNT
>> +};
>> +
>> struct virtio_mem {
>> struct virtio_device *vdev;
>>
>> @@ -113,6 +132,9 @@ struct virtio_mem {
>> atomic64_t offline_size;
>> uint64_t offline_threshold;
>>
>> + /* If set, the driver is in SBM, otherwise in BBM. */
>> + bool in_sbm;
>> +
>> struct {
>> /* Id of the first memory block of this device. */
>> unsigned long first_mb_id;
>> @@ -151,9 +173,27 @@ struct virtio_mem {
>> unsigned long *sb_states;
>> } sbm;
>>
>> + struct {
>> + /* Id of the first big block of this device. */
>> + unsigned long first_bb_id;
>> + /* Id of the last usable big block of this device. */
>> + unsigned long last_usable_bb_id;
>> + /* Id of the next device bock to prepare when needed. */
>> + unsigned long next_bb_id;
>> +
>> + /* Summary of all big block states. */
>> + unsigned long bb_count[VIRTIO_MEM_BBM_BB_COUNT];
>> +
>> + /* One byte state per big block. See sbm.mb_states. */
>> + uint8_t *bb_states;
>> +
>> + /* The block size used for (un)plugged, adding/removing. */
>> + uint64_t bb_size;
>> + } bbm;
>
> Can we use a union here?
As I had the same thought initially, it most probably makes sense :)
Thanks!
--
Thanks,
David / dhildenb
