lguest Hypercalls

一  guest

1.1  直接方式 hcall


1.2  lazy方式  async_hcall
最多可以batch LHCALL_RING_SIZE个hypercalls.  在合适的时机选择lazy_mode, 进入async_hcall处理方式.
具体应用见arch_enter_lazy_mmu_mode 等函数

见下面的注释
/*G:035 Notice the lazy_hcall() above, rather than hcall().  This is our first
 * real optimization trick!
 *
 * When lazy_mode is set, it means we're allowed to defer all hypercalls and do
 * them as a batch when lazy_mode is eventually turned off.  Because hypercalls
 * are reasonably expensive, batching them up makes sense.  For example, a
 * large mmap might update dozens of page table entries: that code calls
 * lguest_lazy_mode(PARAVIRT_LAZY_MMU), does the dozen updates, then calls
 * lguest_lazy_mode(PARAVIRT_LAZY_NONE).
 *
 * So, when we're in lazy mode, we call async_hypercall() to store the call for
 * future processing.  When lazy mode is turned off we issue a hypercall to
 * flush the stored calls.
 *
 * There's also a hack where "mode" is set to "PARAVIRT_LAZY_FLUSH" which
 * indicates we're to flush any outstanding calls immediately.  This is used
 * when an interrupt handler does a kmap_atomic(): the page table changes must
 * happen immediately even if we're in the middle of a batch.  Usually we're
 * not, though, so there's nothing to do. */
static enum paravirt_lazy_mode lazy_mode; /* Note: not SMP-safe! */

lguest_data 由guest 和host 共享, 所以guest累积async hcall时不需要切换到host
struct lguest_data
{   
    /* 0xFF == done (set by Host), 0 == pending (set by Guest). */
    u8 hcall_status[LHCALL_RING_SIZE];
    /* The actual registers for the hypercalls. */
    struct hcall_ring hcalls[LHCALL_RING_SIZE];
};
1.3 xen 很早就支持,称为multicall

二 HOST

见do_hypercalls