kernel-ark/arch/x86/kernel/paravirt.c
Jeremy Fitzhardinge b4ecc12699 x86: Fix performance regression caused by paravirt_ops on native kernels
Xiaohui Xin and some other folks at Intel have been looking into what's
behind the performance hit of paravirt_ops when running native.

It appears that the hit is entirely due to the paravirtualized
spinlocks introduced by:

 | commit 8efcbab674
 | Date:   Mon Jul 7 12:07:51 2008 -0700
 |
 |     paravirt: introduce a "lock-byte" spinlock implementation

The extra call/return in the spinlock path is somehow
causing an increase in the cycles/instruction of somewhere around 2-7%
(seems to vary quite a lot from test to test).  The working theory is
that the CPU's pipeline is getting upset about the
call->call->locked-op->return->return, and seems to be failing to
speculate (though I haven't seen anything definitive about the precise
reasons).  This doesn't entirely make sense, because the performance
hit is also visible on unlock and other operations which don't involve
locked instructions.  But spinlock operations clearly swamp all the
other pvops operations, even though I can't imagine that they're
nearly as common (there's only a .05% increase in instructions
executed).

If I disable just the pv-spinlock calls, my tests show that pvops is
identical to non-pvops performance on native (my measurements show that
it is actually about .1% faster, but Xiaohui shows a .05% slowdown).

Summary of results, averaging 10 runs of the "mmperf" test, using a
no-pvops build as baseline:

		nopv		Pv-nospin	Pv-spin
CPU cycles	100.00%		99.89%		102.18%
instructions	100.00%		100.10%		100.15%
CPI		100.00%		99.79%		102.03%
cache ref	100.00%		100.84%		100.28%
cache miss	100.00%		90.47%		88.56%
cache miss rate	100.00%		89.72%		88.31%
branches	100.00%		99.93%		100.04%
branch miss	100.00%		103.66%		107.72%
branch miss rt	100.00%		103.73%		107.67%
wallclock	100.00%		99.90%		102.20%

The clear effect here is that the 2% increase in CPI is
directly reflected in the final wallclock time.

(The other interesting effect is that the more ops are
out of line calls via pvops, the lower the cache access
and miss rates.  Not too surprising, but it suggests that
the non-pvops kernel is over-inlined.  On the flipside,
the branch misses go up correspondingly...)

So, what's the fix?

Paravirt patching turns all the pvops calls into direct calls, so
_spin_lock etc do end up having direct calls.  For example, the compiler
generated code for paravirtualized _spin_lock is:

<_spin_lock+0>:		mov    %gs:0xb4c8,%rax
<_spin_lock+9>:		incl   0xffffffffffffe044(%rax)
<_spin_lock+15>:	callq  *0xffffffff805a5b30
<_spin_lock+22>:	retq

The indirect call will get patched to:
<_spin_lock+0>:		mov    %gs:0xb4c8,%rax
<_spin_lock+9>:		incl   0xffffffffffffe044(%rax)
<_spin_lock+15>:	callq <__ticket_spin_lock>
<_spin_lock+20>:	nop; nop		/* or whatever 2-byte nop */
<_spin_lock+22>:	retq

One possibility is to inline _spin_lock, etc, when building an
optimised kernel (ie, when there's no spinlock/preempt
instrumentation/debugging enabled).  That will remove the outer
call/return pair, returning the instruction stream to a single
call/return, which will presumably execute the same as the non-pvops
case.  The downsides arel 1) it will replicate the
preempt_disable/enable code at eack lock/unlock callsite; this code is
fairly small, but not nothing; and 2) the spinlock definitions are
already a very heavily tangled mass of #ifdefs and other preprocessor
magic, and making any changes will be non-trivial.

The other obvious answer is to disable pv-spinlocks.  Making them a
separate config option is fairly easy, and it would be trivial to
enable them only when Xen is enabled (as the only non-default user).
But it doesn't really address the common case of a distro build which
is going to have Xen support enabled, and leaves the open question of
whether the native performance cost of pv-spinlocks is worth the
performance improvement on a loaded Xen system (10% saving of overall
system CPU when guests block rather than spin).  Still it is a
reasonable short-term workaround.

[ Impact: fix pvops performance regression when running native ]

Analysed-by: "Xin Xiaohui" <xiaohui.xin@intel.com>
Analysed-by: "Li Xin" <xin.li@intel.com>
Analysed-by: "Nakajima Jun" <jun.nakajima@intel.com>
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Acked-by: H. Peter Anvin <hpa@zytor.com>
Cc: Nick Piggin <npiggin@suse.de>
Cc: Xen-devel <xen-devel@lists.xensource.com>
LKML-Reference: <4A0B62F7.5030802@goop.org>
[ fixed the help text ]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-05-15 20:07:42 +02:00

515 lines
12 KiB
C

/* Paravirtualization interfaces
Copyright (C) 2006 Rusty Russell IBM Corporation
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
2007 - x86_64 support added by Glauber de Oliveira Costa, Red Hat Inc
*/
#include <linux/errno.h>
#include <linux/module.h>
#include <linux/efi.h>
#include <linux/bcd.h>
#include <linux/highmem.h>
#include <asm/bug.h>
#include <asm/paravirt.h>
#include <asm/desc.h>
#include <asm/setup.h>
#include <asm/pgtable.h>
#include <asm/time.h>
#include <asm/pgalloc.h>
#include <asm/irq.h>
#include <asm/delay.h>
#include <asm/fixmap.h>
#include <asm/apic.h>
#include <asm/tlbflush.h>
#include <asm/timer.h>
/* nop stub */
void _paravirt_nop(void)
{
}
/* identity function, which can be inlined */
u32 _paravirt_ident_32(u32 x)
{
return x;
}
u64 _paravirt_ident_64(u64 x)
{
return x;
}
static void __init default_banner(void)
{
printk(KERN_INFO "Booting paravirtualized kernel on %s\n",
pv_info.name);
}
char *memory_setup(void)
{
return pv_init_ops.memory_setup();
}
/* Simple instruction patching code. */
#define DEF_NATIVE(ops, name, code) \
extern const char start_##ops##_##name[], end_##ops##_##name[]; \
asm("start_" #ops "_" #name ": " code "; end_" #ops "_" #name ":")
/* Undefined instruction for dealing with missing ops pointers. */
static const unsigned char ud2a[] = { 0x0f, 0x0b };
unsigned paravirt_patch_nop(void)
{
return 0;
}
unsigned paravirt_patch_ignore(unsigned len)
{
return len;
}
struct branch {
unsigned char opcode;
u32 delta;
} __attribute__((packed));
unsigned paravirt_patch_call(void *insnbuf,
const void *target, u16 tgt_clobbers,
unsigned long addr, u16 site_clobbers,
unsigned len)
{
struct branch *b = insnbuf;
unsigned long delta = (unsigned long)target - (addr+5);
if (tgt_clobbers & ~site_clobbers)
return len; /* target would clobber too much for this site */
if (len < 5)
return len; /* call too long for patch site */
b->opcode = 0xe8; /* call */
b->delta = delta;
BUILD_BUG_ON(sizeof(*b) != 5);
return 5;
}
unsigned paravirt_patch_jmp(void *insnbuf, const void *target,
unsigned long addr, unsigned len)
{
struct branch *b = insnbuf;
unsigned long delta = (unsigned long)target - (addr+5);
if (len < 5)
return len; /* call too long for patch site */
b->opcode = 0xe9; /* jmp */
b->delta = delta;
return 5;
}
/* Neat trick to map patch type back to the call within the
* corresponding structure. */
static void *get_call_destination(u8 type)
{
struct paravirt_patch_template tmpl = {
.pv_init_ops = pv_init_ops,
.pv_time_ops = pv_time_ops,
.pv_cpu_ops = pv_cpu_ops,
.pv_irq_ops = pv_irq_ops,
.pv_apic_ops = pv_apic_ops,
.pv_mmu_ops = pv_mmu_ops,
#ifdef CONFIG_PARAVIRT_SPINLOCKS
.pv_lock_ops = pv_lock_ops,
#endif
};
return *((void **)&tmpl + type);
}
unsigned paravirt_patch_default(u8 type, u16 clobbers, void *insnbuf,
unsigned long addr, unsigned len)
{
void *opfunc = get_call_destination(type);
unsigned ret;
if (opfunc == NULL)
/* If there's no function, patch it with a ud2a (BUG) */
ret = paravirt_patch_insns(insnbuf, len, ud2a, ud2a+sizeof(ud2a));
else if (opfunc == _paravirt_nop)
/* If the operation is a nop, then nop the callsite */
ret = paravirt_patch_nop();
/* identity functions just return their single argument */
else if (opfunc == _paravirt_ident_32)
ret = paravirt_patch_ident_32(insnbuf, len);
else if (opfunc == _paravirt_ident_64)
ret = paravirt_patch_ident_64(insnbuf, len);
else if (type == PARAVIRT_PATCH(pv_cpu_ops.iret) ||
type == PARAVIRT_PATCH(pv_cpu_ops.irq_enable_sysexit) ||
type == PARAVIRT_PATCH(pv_cpu_ops.usergs_sysret32) ||
type == PARAVIRT_PATCH(pv_cpu_ops.usergs_sysret64))
/* If operation requires a jmp, then jmp */
ret = paravirt_patch_jmp(insnbuf, opfunc, addr, len);
else
/* Otherwise call the function; assume target could
clobber any caller-save reg */
ret = paravirt_patch_call(insnbuf, opfunc, CLBR_ANY,
addr, clobbers, len);
return ret;
}
unsigned paravirt_patch_insns(void *insnbuf, unsigned len,
const char *start, const char *end)
{
unsigned insn_len = end - start;
if (insn_len > len || start == NULL)
insn_len = len;
else
memcpy(insnbuf, start, insn_len);
return insn_len;
}
void init_IRQ(void)
{
pv_irq_ops.init_IRQ();
}
static void native_flush_tlb(void)
{
__native_flush_tlb();
}
/*
* Global pages have to be flushed a bit differently. Not a real
* performance problem because this does not happen often.
*/
static void native_flush_tlb_global(void)
{
__native_flush_tlb_global();
}
static void native_flush_tlb_single(unsigned long addr)
{
__native_flush_tlb_single(addr);
}
/* These are in entry.S */
extern void native_iret(void);
extern void native_irq_enable_sysexit(void);
extern void native_usergs_sysret32(void);
extern void native_usergs_sysret64(void);
static int __init print_banner(void)
{
pv_init_ops.banner();
return 0;
}
core_initcall(print_banner);
static struct resource reserve_ioports = {
.start = 0,
.end = IO_SPACE_LIMIT,
.name = "paravirt-ioport",
.flags = IORESOURCE_IO | IORESOURCE_BUSY,
};
/*
* Reserve the whole legacy IO space to prevent any legacy drivers
* from wasting time probing for their hardware. This is a fairly
* brute-force approach to disabling all non-virtual drivers.
*
* Note that this must be called very early to have any effect.
*/
int paravirt_disable_iospace(void)
{
return request_resource(&ioport_resource, &reserve_ioports);
}
static DEFINE_PER_CPU(enum paravirt_lazy_mode, paravirt_lazy_mode) = PARAVIRT_LAZY_NONE;
static inline void enter_lazy(enum paravirt_lazy_mode mode)
{
BUG_ON(__get_cpu_var(paravirt_lazy_mode) != PARAVIRT_LAZY_NONE);
BUG_ON(preemptible());
__get_cpu_var(paravirt_lazy_mode) = mode;
}
void paravirt_leave_lazy(enum paravirt_lazy_mode mode)
{
BUG_ON(__get_cpu_var(paravirt_lazy_mode) != mode);
BUG_ON(preemptible());
__get_cpu_var(paravirt_lazy_mode) = PARAVIRT_LAZY_NONE;
}
void paravirt_enter_lazy_mmu(void)
{
enter_lazy(PARAVIRT_LAZY_MMU);
}
void paravirt_leave_lazy_mmu(void)
{
paravirt_leave_lazy(PARAVIRT_LAZY_MMU);
}
void paravirt_enter_lazy_cpu(void)
{
enter_lazy(PARAVIRT_LAZY_CPU);
}
void paravirt_leave_lazy_cpu(void)
{
paravirt_leave_lazy(PARAVIRT_LAZY_CPU);
}
enum paravirt_lazy_mode paravirt_get_lazy_mode(void)
{
return __get_cpu_var(paravirt_lazy_mode);
}
void arch_flush_lazy_mmu_mode(void)
{
preempt_disable();
if (paravirt_get_lazy_mode() == PARAVIRT_LAZY_MMU) {
WARN_ON(preempt_count() == 1);
arch_leave_lazy_mmu_mode();
arch_enter_lazy_mmu_mode();
}
preempt_enable();
}
void arch_flush_lazy_cpu_mode(void)
{
preempt_disable();
if (paravirt_get_lazy_mode() == PARAVIRT_LAZY_CPU) {
WARN_ON(preempt_count() == 1);
arch_leave_lazy_cpu_mode();
arch_enter_lazy_cpu_mode();
}
preempt_enable();
}
struct pv_info pv_info = {
.name = "bare hardware",
.paravirt_enabled = 0,
.kernel_rpl = 0,
.shared_kernel_pmd = 1, /* Only used when CONFIG_X86_PAE is set */
};
struct pv_init_ops pv_init_ops = {
.patch = native_patch,
.banner = default_banner,
.arch_setup = paravirt_nop,
.memory_setup = machine_specific_memory_setup,
};
struct pv_time_ops pv_time_ops = {
.time_init = hpet_time_init,
.get_wallclock = native_get_wallclock,
.set_wallclock = native_set_wallclock,
.sched_clock = native_sched_clock,
.get_tsc_khz = native_calibrate_tsc,
};
struct pv_irq_ops pv_irq_ops = {
.init_IRQ = native_init_IRQ,
.save_fl = __PV_IS_CALLEE_SAVE(native_save_fl),
.restore_fl = __PV_IS_CALLEE_SAVE(native_restore_fl),
.irq_disable = __PV_IS_CALLEE_SAVE(native_irq_disable),
.irq_enable = __PV_IS_CALLEE_SAVE(native_irq_enable),
.safe_halt = native_safe_halt,
.halt = native_halt,
#ifdef CONFIG_X86_64
.adjust_exception_frame = paravirt_nop,
#endif
};
struct pv_cpu_ops pv_cpu_ops = {
.cpuid = native_cpuid,
.get_debugreg = native_get_debugreg,
.set_debugreg = native_set_debugreg,
.clts = native_clts,
.read_cr0 = native_read_cr0,
.write_cr0 = native_write_cr0,
.read_cr4 = native_read_cr4,
.read_cr4_safe = native_read_cr4_safe,
.write_cr4 = native_write_cr4,
#ifdef CONFIG_X86_64
.read_cr8 = native_read_cr8,
.write_cr8 = native_write_cr8,
#endif
.wbinvd = native_wbinvd,
.read_msr = native_read_msr_safe,
.read_msr_amd = native_read_msr_amd_safe,
.write_msr = native_write_msr_safe,
.read_tsc = native_read_tsc,
.read_pmc = native_read_pmc,
.read_tscp = native_read_tscp,
.load_tr_desc = native_load_tr_desc,
.set_ldt = native_set_ldt,
.load_gdt = native_load_gdt,
.load_idt = native_load_idt,
.store_gdt = native_store_gdt,
.store_idt = native_store_idt,
.store_tr = native_store_tr,
.load_tls = native_load_tls,
#ifdef CONFIG_X86_64
.load_gs_index = native_load_gs_index,
#endif
.write_ldt_entry = native_write_ldt_entry,
.write_gdt_entry = native_write_gdt_entry,
.write_idt_entry = native_write_idt_entry,
.alloc_ldt = paravirt_nop,
.free_ldt = paravirt_nop,
.load_sp0 = native_load_sp0,
#if defined(CONFIG_X86_32) || defined(CONFIG_IA32_EMULATION)
.irq_enable_sysexit = native_irq_enable_sysexit,
#endif
#ifdef CONFIG_X86_64
#ifdef CONFIG_IA32_EMULATION
.usergs_sysret32 = native_usergs_sysret32,
#endif
.usergs_sysret64 = native_usergs_sysret64,
#endif
.iret = native_iret,
.swapgs = native_swapgs,
.set_iopl_mask = native_set_iopl_mask,
.io_delay = native_io_delay,
.lazy_mode = {
.enter = paravirt_nop,
.leave = paravirt_nop,
},
};
struct pv_apic_ops pv_apic_ops = {
#ifdef CONFIG_X86_LOCAL_APIC
.setup_boot_clock = setup_boot_APIC_clock,
.setup_secondary_clock = setup_secondary_APIC_clock,
.startup_ipi_hook = paravirt_nop,
#endif
};
#if defined(CONFIG_X86_32) && !defined(CONFIG_X86_PAE)
/* 32-bit pagetable entries */
#define PTE_IDENT __PV_IS_CALLEE_SAVE(_paravirt_ident_32)
#else
/* 64-bit pagetable entries */
#define PTE_IDENT __PV_IS_CALLEE_SAVE(_paravirt_ident_64)
#endif
struct pv_mmu_ops pv_mmu_ops = {
#ifndef CONFIG_X86_64
.pagetable_setup_start = native_pagetable_setup_start,
.pagetable_setup_done = native_pagetable_setup_done,
#else
.pagetable_setup_start = paravirt_nop,
.pagetable_setup_done = paravirt_nop,
#endif
.read_cr2 = native_read_cr2,
.write_cr2 = native_write_cr2,
.read_cr3 = native_read_cr3,
.write_cr3 = native_write_cr3,
.flush_tlb_user = native_flush_tlb,
.flush_tlb_kernel = native_flush_tlb_global,
.flush_tlb_single = native_flush_tlb_single,
.flush_tlb_others = native_flush_tlb_others,
.pgd_alloc = __paravirt_pgd_alloc,
.pgd_free = paravirt_nop,
.alloc_pte = paravirt_nop,
.alloc_pmd = paravirt_nop,
.alloc_pmd_clone = paravirt_nop,
.alloc_pud = paravirt_nop,
.release_pte = paravirt_nop,
.release_pmd = paravirt_nop,
.release_pud = paravirt_nop,
.set_pte = native_set_pte,
.set_pte_at = native_set_pte_at,
.set_pmd = native_set_pmd,
.pte_update = paravirt_nop,
.pte_update_defer = paravirt_nop,
.ptep_modify_prot_start = __ptep_modify_prot_start,
.ptep_modify_prot_commit = __ptep_modify_prot_commit,
#ifdef CONFIG_HIGHPTE
.kmap_atomic_pte = kmap_atomic,
#endif
#if PAGETABLE_LEVELS >= 3
#ifdef CONFIG_X86_PAE
.set_pte_atomic = native_set_pte_atomic,
.pte_clear = native_pte_clear,
.pmd_clear = native_pmd_clear,
#endif
.set_pud = native_set_pud,
.pmd_val = PTE_IDENT,
.make_pmd = PTE_IDENT,
#if PAGETABLE_LEVELS == 4
.pud_val = PTE_IDENT,
.make_pud = PTE_IDENT,
.set_pgd = native_set_pgd,
#endif
#endif /* PAGETABLE_LEVELS >= 3 */
.pte_val = PTE_IDENT,
.pgd_val = PTE_IDENT,
.make_pte = PTE_IDENT,
.make_pgd = PTE_IDENT,
.dup_mmap = paravirt_nop,
.exit_mmap = paravirt_nop,
.activate_mm = paravirt_nop,
.lazy_mode = {
.enter = paravirt_nop,
.leave = paravirt_nop,
},
.set_fixmap = native_set_fixmap,
};
EXPORT_SYMBOL_GPL(pv_time_ops);
EXPORT_SYMBOL (pv_cpu_ops);
EXPORT_SYMBOL (pv_mmu_ops);
EXPORT_SYMBOL_GPL(pv_apic_ops);
EXPORT_SYMBOL_GPL(pv_info);
EXPORT_SYMBOL (pv_irq_ops);