Hi! This emerged from the ongoing efforts to consolidate the entry code maze on x86 along with the tracing and instrumentation induced violations of RCU constraints. While working on this I stumbled over the KVM async page fault handler and kvm_async_pf_task_wait() in particular. It took me a while to realize that the randomly sprinkled around rcu_irq_enter()/exit() invocations are just cargo cult programming. Several patches "fixed" RCU splats by curing the symptoms without realizing that the code is fundametally flawed from a design perspective. Aside of that Andy noticed that the way the async page fault handler is implemented can be improved which in turn allows further simplification vs. CR2 consistency and the general exception entry handling on x86. I'm sending this out as a stand alone series against mainline, but I'd prefer to take this along with the rest if the entry code rework. If you look at the changelog then don't be afraid, most of the added lines are comments which were painfully absent in the original code. Thanks, tglx --- entry/entry_32.S | 8 -- entry/entry_64.S | 4 - include/asm/kvm_para.h | 21 ++++- include/asm/x86_init.h | 2 kernel/kvm.c | 193 +++++++++++++++++++++++++++++++++---------------- kernel/traps.c | 2 kernel/x86_init.c | 1 kvm/mmu/mmu.c | 2 mm/fault.c | 19 ++++ 9 files changed, 171 insertions(+), 81 deletions(-)
