Always drop mmu_lock to allocate shadow pages in the TDP MMU when doing eager page splitting. Dropping mmu_lock during eager page splitting is cheap since KVM does not have to flush remote TLBs, and avoids stalling vCPU threads that are taking page faults in parallel. This change reduces 20%+ dips in MySQL throughput during live migration in a 160 vCPU VM while userspace is issuing CLEAR_DIRTY_LOG ioctls (tested with 1GiB and 8GiB CLEARs). Userspace could issue finer-grained CLEARs, which would also reduce contention on mmu_lock, but doing so will increase the rate of remote TLB flushing (KVM must flush TLBs before returning from CLEAR_DITY_LOG). When there isn't contention on mmu_lock[1], this change does not regress the time it takes to perform eager page splitting. [1] Tested with dirty_log_perf_test, which does not run vCPUs during eager page splitting, and with a 16 vCPU VM Live Migration with manual-protect disabled (where mmu_lock is held in read-mode). Cc: Bibo Mao <maobibo@xxxxxxxxxxx> Cc: Sean Christopherson <seanjc@xxxxxxxxxx> Signed-off-by: David Matlack <dmatlack@xxxxxxxxxx> --- v3: - Only drop mmu_lock during TDP MMU eager page splitting. This fixes the perfomance regression without regressing CLEAR_DIRTY_LOG performance on other architectures from frequent lock dropping. v2: https://lore.kernel.org/kvm/20240402213656.3068504-1-dmatlack@xxxxxxxxxx/ - Rebase onto kvm/queue [Marc] v1: https://lore.kernel.org/kvm/20231205181645.482037-1-dmatlack@xxxxxxxxxx/ arch/x86/kvm/mmu/tdp_mmu.c | 22 ++++------------------ 1 file changed, 4 insertions(+), 18 deletions(-) diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c index aaa2369a9479..2089d696e3c6 100644 --- a/arch/x86/kvm/mmu/tdp_mmu.c +++ b/arch/x86/kvm/mmu/tdp_mmu.c @@ -1385,11 +1385,11 @@ bool kvm_tdp_mmu_wrprot_slot(struct kvm *kvm, return spte_set; } -static struct kvm_mmu_page *__tdp_mmu_alloc_sp_for_split(gfp_t gfp) +static struct kvm_mmu_page *__tdp_mmu_alloc_sp_for_split(void) { + gfp_t gfp = GFP_KERNEL_ACCOUNT | __GFP_ZERO; struct kvm_mmu_page *sp; - gfp |= __GFP_ZERO; sp = kmem_cache_alloc(mmu_page_header_cache, gfp); if (!sp) @@ -1412,19 +1412,6 @@ static struct kvm_mmu_page *tdp_mmu_alloc_sp_for_split(struct kvm *kvm, kvm_lockdep_assert_mmu_lock_held(kvm, shared); - /* - * Since we are allocating while under the MMU lock we have to be - * careful about GFP flags. Use GFP_NOWAIT to avoid blocking on direct - * reclaim and to avoid making any filesystem callbacks (which can end - * up invoking KVM MMU notifiers, resulting in a deadlock). - * - * If this allocation fails we drop the lock and retry with reclaim - * allowed. - */ - sp = __tdp_mmu_alloc_sp_for_split(GFP_NOWAIT | __GFP_ACCOUNT); - if (sp) - return sp; - rcu_read_unlock(); if (shared) @@ -1433,7 +1420,7 @@ static struct kvm_mmu_page *tdp_mmu_alloc_sp_for_split(struct kvm *kvm, write_unlock(&kvm->mmu_lock); iter->yielded = true; - sp = __tdp_mmu_alloc_sp_for_split(GFP_KERNEL_ACCOUNT); + sp = __tdp_mmu_alloc_sp_for_split(); if (shared) read_lock(&kvm->mmu_lock); @@ -1524,8 +1511,7 @@ static int tdp_mmu_split_huge_pages_root(struct kvm *kvm, break; } - if (iter.yielded) - continue; + continue; } tdp_mmu_init_child_sp(sp, &iter); base-commit: 2d181d84af38146748042a6974c577fc46c3f1c3 -- 2.45.0.118.g7fe29c98d7-goog
