[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [Xen-devel] [PATCH v2 1/6] x86/xpti: avoid copying L4 page table contents when possible
For mitigation of Meltdown the current L4 page table is copied to the cpu local root page table each time a 64 bit pv guest is entered. Copying can be avoided in cases where the guest L4 page table hasn't been modified while running the hypervisor, e.g. when handling interrupts or any hypercall not modifying the L4 page table or %cr3. So add a per-cpu flag whether the copying should be performed and set that flag only when loading a new %cr3 or modifying the L4 page table. This includes synchronization of the cpu local root page table with other cpus, so add a special synchronization flag for that case. A simple performance check (compiling the hypervisor via "make -j 4") in dom0 with 4 vcpus shows a significant improvement: - real time drops from 112 seconds to 103 seconds - system time drops from 142 seconds to 131 seconds Signed-off-by: Juergen Gross <jgross@xxxxxxxx> --- To be applied on top of Jan's "Meltdown band-aid overhead reduction" series --- xen/arch/x86/mm.c | 32 +++++++++++++++++++------------- xen/arch/x86/pv/domain.c | 1 + xen/arch/x86/smp.c | 2 ++ xen/arch/x86/x86_64/asm-offsets.c | 1 + xen/arch/x86/x86_64/entry.S | 8 ++++++-- xen/include/asm-x86/current.h | 8 ++++++++ xen/include/asm-x86/flushtlb.h | 2 ++ 7 files changed, 39 insertions(+), 15 deletions(-) diff --git a/xen/arch/x86/mm.c b/xen/arch/x86/mm.c index d8d3ee2ecd..fdc1636817 100644 --- a/xen/arch/x86/mm.c +++ b/xen/arch/x86/mm.c @@ -509,6 +509,8 @@ void make_cr3(struct vcpu *v, mfn_t mfn) void write_ptbase(struct vcpu *v) { + get_cpu_info()->root_pgt_changed = this_cpu(root_pgt) && is_pv_vcpu(v) && + !is_pv_32bit_vcpu(v); write_cr3(v->arch.cr3); } @@ -3704,18 +3706,22 @@ long do_mmu_update( break; rc = mod_l4_entry(va, l4e_from_intpte(req.val), mfn, cmd == MMU_PT_UPDATE_PRESERVE_AD, v); - /* - * No need to sync if all uses of the page can be accounted - * to the page lock we hold, its pinned status, and uses on - * this (v)CPU. - */ - if ( !rc && !cpu_has_no_xpti && - ((page->u.inuse.type_info & PGT_count_mask) > - (1 + !!(page->u.inuse.type_info & PGT_pinned) + - (pagetable_get_pfn(curr->arch.guest_table) == mfn) + - (pagetable_get_pfn(curr->arch.guest_table_user) == - mfn))) ) - sync_guest = true; + if ( !rc && !cpu_has_no_xpti ) + { + get_cpu_info()->root_pgt_changed = true; + /* + * No need to sync if all uses of the page can be + * accounted to the page lock we hold, its pinned + * status, and uses on this (v)CPU. + */ + if ( (page->u.inuse.type_info & PGT_count_mask) > + (1 + !!(page->u.inuse.type_info & PGT_pinned) + + (pagetable_get_pfn(curr->arch.guest_table) == + mfn) + + (pagetable_get_pfn(curr->arch.guest_table_user) == + mfn)) ) + sync_guest = true; + } break; case PGT_writable_page: @@ -3830,7 +3836,7 @@ long do_mmu_update( cpumask_andnot(mask, pt_owner->dirty_cpumask, cpumask_of(cpu)); if ( !cpumask_empty(mask) ) - flush_mask(mask, FLUSH_TLB_GLOBAL); + flush_mask(mask, FLUSH_TLB_GLOBAL | FLUSH_ROOT_PGTBL); } perfc_add(num_page_updates, i); diff --git a/xen/arch/x86/pv/domain.c b/xen/arch/x86/pv/domain.c index 868a23fd7e..7742d522f5 100644 --- a/xen/arch/x86/pv/domain.c +++ b/xen/arch/x86/pv/domain.c @@ -238,6 +238,7 @@ static void _toggle_guest_pt(struct vcpu *v, bool force_cr3) /* Don't flush user global mappings from the TLB. Don't tick TLB clock. */ asm volatile ( "mov %0, %%cr3" : : "r" (v->arch.cr3) : "memory" ); + get_cpu_info()->root_pgt_changed = true; if ( !(v->arch.flags & TF_kernel_mode) ) return; diff --git a/xen/arch/x86/smp.c b/xen/arch/x86/smp.c index 033dd05958..60b0657ab7 100644 --- a/xen/arch/x86/smp.c +++ b/xen/arch/x86/smp.c @@ -207,6 +207,8 @@ void invalidate_interrupt(struct cpu_user_regs *regs) unsigned int flags = flush_flags; ack_APIC_irq(); perfc_incr(ipis); + if ( flags & FLUSH_ROOT_PGTBL ) + get_cpu_info()->root_pgt_changed = true; if ( (flags & FLUSH_VCPU_STATE) && __sync_local_execstate() ) flags &= ~(FLUSH_TLB | FLUSH_TLB_GLOBAL); if ( flags & ~(FLUSH_VCPU_STATE | FLUSH_ORDER_MASK) ) diff --git a/xen/arch/x86/x86_64/asm-offsets.c b/xen/arch/x86/x86_64/asm-offsets.c index e925e6589c..c9225b06c1 100644 --- a/xen/arch/x86/x86_64/asm-offsets.c +++ b/xen/arch/x86/x86_64/asm-offsets.c @@ -144,6 +144,7 @@ void __dummy__(void) OFFSET(CPUINFO_shadow_spec_ctrl, struct cpu_info, shadow_spec_ctrl); OFFSET(CPUINFO_use_shadow_spec_ctrl, struct cpu_info, use_shadow_spec_ctrl); OFFSET(CPUINFO_bti_ist_info, struct cpu_info, bti_ist_info); + OFFSET(CPUINFO_root_pgt_changed, struct cpu_info, root_pgt_changed); DEFINE(CPUINFO_sizeof, sizeof(struct cpu_info)); BLANK(); diff --git a/xen/arch/x86/x86_64/entry.S b/xen/arch/x86/x86_64/entry.S index 3b578b4392..828f9ccfe8 100644 --- a/xen/arch/x86/x86_64/entry.S +++ b/xen/arch/x86/x86_64/entry.S @@ -48,10 +48,13 @@ restore_all_guest: .Lrag_cr3_start: mov VCPU_cr3(%rbx), %r9 GET_STACK_END(dx) - mov STACK_CPUINFO_FIELD(pv_cr3)(%rdx), %rdi + mov STACK_CPUINFO_FIELD(pv_cr3)(%rdx), %rax + cmpb $0, STACK_CPUINFO_FIELD(root_pgt_changed)(%rdx) + je .Lrag_copy_done + movb $0, STACK_CPUINFO_FIELD(root_pgt_changed)(%rdx) movabs $PADDR_MASK & PAGE_MASK, %rsi movabs $DIRECTMAP_VIRT_START, %rcx - mov %rdi, %rax + mov %rax, %rdi and %rsi, %rdi and %r9, %rsi add %rcx, %rdi @@ -67,6 +70,7 @@ restore_all_guest: sub $(ROOT_PAGETABLE_FIRST_XEN_SLOT - \ ROOT_PAGETABLE_LAST_XEN_SLOT - 1) * 8, %rdi rep movsq +.Lrag_copy_done: mov STACK_CPUINFO_FIELD(cr4)(%rdx), %rdi mov %r9, STACK_CPUINFO_FIELD(xen_cr3)(%rdx) mov %rdi, %rsi diff --git a/xen/include/asm-x86/current.h b/xen/include/asm-x86/current.h index 4678a0fcf5..3c96c173c2 100644 --- a/xen/include/asm-x86/current.h +++ b/xen/include/asm-x86/current.h @@ -59,6 +59,14 @@ struct cpu_info { bool use_shadow_spec_ctrl; uint8_t bti_ist_info; + /* + * The following field controls copying of the L4 page table of 64-bit + * PV guests to the per-cpu root page table on entering the guest context. + * If set the L4 page table is being copied to the root page table and + * the field will be reset. + */ + bool root_pgt_changed; + unsigned long __pad; /* get_stack_bottom() must be 16-byte aligned */ }; diff --git a/xen/include/asm-x86/flushtlb.h b/xen/include/asm-x86/flushtlb.h index 2cade9cbfb..052f0fa403 100644 --- a/xen/include/asm-x86/flushtlb.h +++ b/xen/include/asm-x86/flushtlb.h @@ -103,6 +103,8 @@ void write_cr3(unsigned long cr3); #define FLUSH_VA_VALID 0x800 /* Flush CPU state */ #define FLUSH_VCPU_STATE 0x1000 + /* Flush the per-cpu root page table */ +#define FLUSH_ROOT_PGTBL 0x2000 /* Flush local TLBs/caches. */ unsigned int flush_area_local(const void *va, unsigned int flags); -- 2.13.6 _______________________________________________ Xen-devel mailing list Xen-devel@xxxxxxxxxxxxxxxxxxxx https://lists.xenproject.org/mailman/listinfo/xen-devel
|
Lists.xenproject.org is hosted with RackSpace, monitoring our |