linux_dsm_epyc7002/arch/x86/mm/kasan_init_64.c
Kirill A. Shutemov 39b9552281 x86/mm: Optimize boot-time paging mode switching cost
By this point we have functioning boot-time switching between 4- and
5-level paging mode. But naive approach comes with cost.

Numbers below are for kernel build, allmodconfig, 5 times.

CONFIG_X86_5LEVEL=n:

 Performance counter stats for 'sh -c make -j100 -B -k >/dev/null' (5 runs):

   17308719.892691      task-clock:u (msec)       #   26.772 CPUs utilized            ( +-  0.11% )
                 0      context-switches:u        #    0.000 K/sec
                 0      cpu-migrations:u          #    0.000 K/sec
       331,993,164      page-faults:u             #    0.019 M/sec                    ( +-  0.01% )
43,614,978,867,455      cycles:u                  #    2.520 GHz                      ( +-  0.01% )
39,371,534,575,126      stalled-cycles-frontend:u #   90.27% frontend cycles idle     ( +-  0.09% )
28,363,350,152,428      instructions:u            #    0.65  insn per cycle
                                                  #    1.39  stalled cycles per insn  ( +-  0.00% )
 6,316,784,066,413      branches:u                #  364.948 M/sec                    ( +-  0.00% )
   250,808,144,781      branch-misses:u           #    3.97% of all branches          ( +-  0.01% )

     646.531974142 seconds time elapsed                                          ( +-  1.15% )

CONFIG_X86_5LEVEL=y:

 Performance counter stats for 'sh -c make -j100 -B -k >/dev/null' (5 runs):

   17411536.780625      task-clock:u (msec)       #   26.426 CPUs utilized            ( +-  0.10% )
                 0      context-switches:u        #    0.000 K/sec
                 0      cpu-migrations:u          #    0.000 K/sec
       331,868,663      page-faults:u             #    0.019 M/sec                    ( +-  0.01% )
43,865,909,056,301      cycles:u                  #    2.519 GHz                      ( +-  0.01% )
39,740,130,365,581      stalled-cycles-frontend:u #   90.59% frontend cycles idle     ( +-  0.05% )
28,363,358,997,959      instructions:u            #    0.65  insn per cycle
                                                  #    1.40  stalled cycles per insn  ( +-  0.00% )
 6,316,784,937,460      branches:u                #  362.793 M/sec                    ( +-  0.00% )
   251,531,919,485      branch-misses:u           #    3.98% of all branches          ( +-  0.00% )

     658.886307752 seconds time elapsed                                          ( +-  0.92% )

The patch tries to fix the performance regression by using
cpu_feature_enabled(X86_FEATURE_LA57) instead of pgtable_l5_enabled in
all hot code paths. These will statically patch the target code for
additional performance.

CONFIG_X86_5LEVEL=y + the patch:

 Performance counter stats for 'sh -c make -j100 -B -k >/dev/null' (5 runs):

   17381990.268506      task-clock:u (msec)       #   26.907 CPUs utilized            ( +-  0.19% )
                 0      context-switches:u        #    0.000 K/sec
                 0      cpu-migrations:u          #    0.000 K/sec
       331,862,625      page-faults:u             #    0.019 M/sec                    ( +-  0.01% )
43,697,726,320,051      cycles:u                  #    2.514 GHz                      ( +-  0.03% )
39,480,408,690,401      stalled-cycles-frontend:u #   90.35% frontend cycles idle     ( +-  0.05% )
28,363,394,221,388      instructions:u            #    0.65  insn per cycle
                                                  #    1.39  stalled cycles per insn  ( +-  0.00% )
 6,316,794,985,573      branches:u                #  363.410 M/sec                    ( +-  0.00% )
   251,013,232,547      branch-misses:u           #    3.97% of all branches          ( +-  0.01% )

     645.991174661 seconds time elapsed                                          ( +-  1.19% )

Unfortunately, this approach doesn't help with text size:

  vmlinux.before .text size:	8190319
  vmlinux.after .text size:	8200623

The .text section is increased by about 4k. Not sure if we can do anything
about this.

Signed-off-by: Kirill A. Shuemov <kirill.shutemov@linux.intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Borislav Petkov <bp@suse.de>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-mm@kvack.org
Link: http://lkml.kernel.org/r/20180216114948.68868-4-kirill.shutemov@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2018-02-21 10:19:18 +01:00

383 lines
9.8 KiB
C

// SPDX-License-Identifier: GPL-2.0
#define DISABLE_BRANCH_PROFILING
#define pr_fmt(fmt) "kasan: " fmt
#ifdef CONFIG_X86_5LEVEL
/* Too early to use cpu_feature_enabled() */
#define pgtable_l5_enabled __pgtable_l5_enabled
#endif
#include <linux/bootmem.h>
#include <linux/kasan.h>
#include <linux/kdebug.h>
#include <linux/memblock.h>
#include <linux/mm.h>
#include <linux/sched.h>
#include <linux/sched/task.h>
#include <linux/vmalloc.h>
#include <asm/e820/types.h>
#include <asm/pgalloc.h>
#include <asm/tlbflush.h>
#include <asm/sections.h>
#include <asm/pgtable.h>
#include <asm/cpu_entry_area.h>
extern struct range pfn_mapped[E820_MAX_ENTRIES];
static p4d_t tmp_p4d_table[MAX_PTRS_PER_P4D] __initdata __aligned(PAGE_SIZE);
static __init void *early_alloc(size_t size, int nid, bool panic)
{
if (panic)
return memblock_virt_alloc_try_nid(size, size,
__pa(MAX_DMA_ADDRESS), BOOTMEM_ALLOC_ACCESSIBLE, nid);
else
return memblock_virt_alloc_try_nid_nopanic(size, size,
__pa(MAX_DMA_ADDRESS), BOOTMEM_ALLOC_ACCESSIBLE, nid);
}
static void __init kasan_populate_pmd(pmd_t *pmd, unsigned long addr,
unsigned long end, int nid)
{
pte_t *pte;
if (pmd_none(*pmd)) {
void *p;
if (boot_cpu_has(X86_FEATURE_PSE) &&
((end - addr) == PMD_SIZE) &&
IS_ALIGNED(addr, PMD_SIZE)) {
p = early_alloc(PMD_SIZE, nid, false);
if (p && pmd_set_huge(pmd, __pa(p), PAGE_KERNEL))
return;
else if (p)
memblock_free(__pa(p), PMD_SIZE);
}
p = early_alloc(PAGE_SIZE, nid, true);
pmd_populate_kernel(&init_mm, pmd, p);
}
pte = pte_offset_kernel(pmd, addr);
do {
pte_t entry;
void *p;
if (!pte_none(*pte))
continue;
p = early_alloc(PAGE_SIZE, nid, true);
entry = pfn_pte(PFN_DOWN(__pa(p)), PAGE_KERNEL);
set_pte_at(&init_mm, addr, pte, entry);
} while (pte++, addr += PAGE_SIZE, addr != end);
}
static void __init kasan_populate_pud(pud_t *pud, unsigned long addr,
unsigned long end, int nid)
{
pmd_t *pmd;
unsigned long next;
if (pud_none(*pud)) {
void *p;
if (boot_cpu_has(X86_FEATURE_GBPAGES) &&
((end - addr) == PUD_SIZE) &&
IS_ALIGNED(addr, PUD_SIZE)) {
p = early_alloc(PUD_SIZE, nid, false);
if (p && pud_set_huge(pud, __pa(p), PAGE_KERNEL))
return;
else if (p)
memblock_free(__pa(p), PUD_SIZE);
}
p = early_alloc(PAGE_SIZE, nid, true);
pud_populate(&init_mm, pud, p);
}
pmd = pmd_offset(pud, addr);
do {
next = pmd_addr_end(addr, end);
if (!pmd_large(*pmd))
kasan_populate_pmd(pmd, addr, next, nid);
} while (pmd++, addr = next, addr != end);
}
static void __init kasan_populate_p4d(p4d_t *p4d, unsigned long addr,
unsigned long end, int nid)
{
pud_t *pud;
unsigned long next;
if (p4d_none(*p4d)) {
void *p = early_alloc(PAGE_SIZE, nid, true);
p4d_populate(&init_mm, p4d, p);
}
pud = pud_offset(p4d, addr);
do {
next = pud_addr_end(addr, end);
if (!pud_large(*pud))
kasan_populate_pud(pud, addr, next, nid);
} while (pud++, addr = next, addr != end);
}
static void __init kasan_populate_pgd(pgd_t *pgd, unsigned long addr,
unsigned long end, int nid)
{
void *p;
p4d_t *p4d;
unsigned long next;
if (pgd_none(*pgd)) {
p = early_alloc(PAGE_SIZE, nid, true);
pgd_populate(&init_mm, pgd, p);
}
p4d = p4d_offset(pgd, addr);
do {
next = p4d_addr_end(addr, end);
kasan_populate_p4d(p4d, addr, next, nid);
} while (p4d++, addr = next, addr != end);
}
static void __init kasan_populate_shadow(unsigned long addr, unsigned long end,
int nid)
{
pgd_t *pgd;
unsigned long next;
addr = addr & PAGE_MASK;
end = round_up(end, PAGE_SIZE);
pgd = pgd_offset_k(addr);
do {
next = pgd_addr_end(addr, end);
kasan_populate_pgd(pgd, addr, next, nid);
} while (pgd++, addr = next, addr != end);
}
static void __init map_range(struct range *range)
{
unsigned long start;
unsigned long end;
start = (unsigned long)kasan_mem_to_shadow(pfn_to_kaddr(range->start));
end = (unsigned long)kasan_mem_to_shadow(pfn_to_kaddr(range->end));
kasan_populate_shadow(start, end, early_pfn_to_nid(range->start));
}
static void __init clear_pgds(unsigned long start,
unsigned long end)
{
pgd_t *pgd;
/* See comment in kasan_init() */
unsigned long pgd_end = end & PGDIR_MASK;
for (; start < pgd_end; start += PGDIR_SIZE) {
pgd = pgd_offset_k(start);
/*
* With folded p4d, pgd_clear() is nop, use p4d_clear()
* instead.
*/
if (pgtable_l5_enabled)
pgd_clear(pgd);
else
p4d_clear(p4d_offset(pgd, start));
}
pgd = pgd_offset_k(start);
for (; start < end; start += P4D_SIZE)
p4d_clear(p4d_offset(pgd, start));
}
static inline p4d_t *early_p4d_offset(pgd_t *pgd, unsigned long addr)
{
unsigned long p4d;
if (!pgtable_l5_enabled)
return (p4d_t *)pgd;
p4d = __pa_nodebug(pgd_val(*pgd)) & PTE_PFN_MASK;
p4d += __START_KERNEL_map - phys_base;
return (p4d_t *)p4d + p4d_index(addr);
}
static void __init kasan_early_p4d_populate(pgd_t *pgd,
unsigned long addr,
unsigned long end)
{
pgd_t pgd_entry;
p4d_t *p4d, p4d_entry;
unsigned long next;
if (pgd_none(*pgd)) {
pgd_entry = __pgd(_KERNPG_TABLE | __pa_nodebug(kasan_zero_p4d));
set_pgd(pgd, pgd_entry);
}
p4d = early_p4d_offset(pgd, addr);
do {
next = p4d_addr_end(addr, end);
if (!p4d_none(*p4d))
continue;
p4d_entry = __p4d(_KERNPG_TABLE | __pa_nodebug(kasan_zero_pud));
set_p4d(p4d, p4d_entry);
} while (p4d++, addr = next, addr != end && p4d_none(*p4d));
}
static void __init kasan_map_early_shadow(pgd_t *pgd)
{
/* See comment in kasan_init() */
unsigned long addr = KASAN_SHADOW_START & PGDIR_MASK;
unsigned long end = KASAN_SHADOW_END;
unsigned long next;
pgd += pgd_index(addr);
do {
next = pgd_addr_end(addr, end);
kasan_early_p4d_populate(pgd, addr, next);
} while (pgd++, addr = next, addr != end);
}
#ifdef CONFIG_KASAN_INLINE
static int kasan_die_handler(struct notifier_block *self,
unsigned long val,
void *data)
{
if (val == DIE_GPF) {
pr_emerg("CONFIG_KASAN_INLINE enabled\n");
pr_emerg("GPF could be caused by NULL-ptr deref or user memory access\n");
}
return NOTIFY_OK;
}
static struct notifier_block kasan_die_notifier = {
.notifier_call = kasan_die_handler,
};
#endif
void __init kasan_early_init(void)
{
int i;
pteval_t pte_val = __pa_nodebug(kasan_zero_page) | __PAGE_KERNEL | _PAGE_ENC;
pmdval_t pmd_val = __pa_nodebug(kasan_zero_pte) | _KERNPG_TABLE;
pudval_t pud_val = __pa_nodebug(kasan_zero_pmd) | _KERNPG_TABLE;
p4dval_t p4d_val = __pa_nodebug(kasan_zero_pud) | _KERNPG_TABLE;
for (i = 0; i < PTRS_PER_PTE; i++)
kasan_zero_pte[i] = __pte(pte_val);
for (i = 0; i < PTRS_PER_PMD; i++)
kasan_zero_pmd[i] = __pmd(pmd_val);
for (i = 0; i < PTRS_PER_PUD; i++)
kasan_zero_pud[i] = __pud(pud_val);
for (i = 0; pgtable_l5_enabled && i < PTRS_PER_P4D; i++)
kasan_zero_p4d[i] = __p4d(p4d_val);
kasan_map_early_shadow(early_top_pgt);
kasan_map_early_shadow(init_top_pgt);
}
void __init kasan_init(void)
{
int i;
void *shadow_cpu_entry_begin, *shadow_cpu_entry_end;
#ifdef CONFIG_KASAN_INLINE
register_die_notifier(&kasan_die_notifier);
#endif
memcpy(early_top_pgt, init_top_pgt, sizeof(early_top_pgt));
/*
* We use the same shadow offset for 4- and 5-level paging to
* facilitate boot-time switching between paging modes.
* As result in 5-level paging mode KASAN_SHADOW_START and
* KASAN_SHADOW_END are not aligned to PGD boundary.
*
* KASAN_SHADOW_START doesn't share PGD with anything else.
* We claim whole PGD entry to make things easier.
*
* KASAN_SHADOW_END lands in the last PGD entry and it collides with
* bunch of things like kernel code, modules, EFI mapping, etc.
* We need to take extra steps to not overwrite them.
*/
if (pgtable_l5_enabled) {
void *ptr;
ptr = (void *)pgd_page_vaddr(*pgd_offset_k(KASAN_SHADOW_END));
memcpy(tmp_p4d_table, (void *)ptr, sizeof(tmp_p4d_table));
set_pgd(&early_top_pgt[pgd_index(KASAN_SHADOW_END)],
__pgd(__pa(tmp_p4d_table) | _KERNPG_TABLE));
}
load_cr3(early_top_pgt);
__flush_tlb_all();
clear_pgds(KASAN_SHADOW_START & PGDIR_MASK, KASAN_SHADOW_END);
kasan_populate_zero_shadow((void *)(KASAN_SHADOW_START & PGDIR_MASK),
kasan_mem_to_shadow((void *)PAGE_OFFSET));
for (i = 0; i < E820_MAX_ENTRIES; i++) {
if (pfn_mapped[i].end == 0)
break;
map_range(&pfn_mapped[i]);
}
shadow_cpu_entry_begin = (void *)CPU_ENTRY_AREA_BASE;
shadow_cpu_entry_begin = kasan_mem_to_shadow(shadow_cpu_entry_begin);
shadow_cpu_entry_begin = (void *)round_down((unsigned long)shadow_cpu_entry_begin,
PAGE_SIZE);
shadow_cpu_entry_end = (void *)(CPU_ENTRY_AREA_BASE +
CPU_ENTRY_AREA_MAP_SIZE);
shadow_cpu_entry_end = kasan_mem_to_shadow(shadow_cpu_entry_end);
shadow_cpu_entry_end = (void *)round_up((unsigned long)shadow_cpu_entry_end,
PAGE_SIZE);
kasan_populate_zero_shadow(
kasan_mem_to_shadow((void *)PAGE_OFFSET + MAXMEM),
shadow_cpu_entry_begin);
kasan_populate_shadow((unsigned long)shadow_cpu_entry_begin,
(unsigned long)shadow_cpu_entry_end, 0);
kasan_populate_zero_shadow(shadow_cpu_entry_end,
kasan_mem_to_shadow((void *)__START_KERNEL_map));
kasan_populate_shadow((unsigned long)kasan_mem_to_shadow(_stext),
(unsigned long)kasan_mem_to_shadow(_end),
early_pfn_to_nid(__pa(_stext)));
kasan_populate_zero_shadow(kasan_mem_to_shadow((void *)MODULES_END),
(void *)KASAN_SHADOW_END);
load_cr3(init_top_pgt);
__flush_tlb_all();
/*
* kasan_zero_page has been used as early shadow memory, thus it may
* contain some garbage. Now we can clear and write protect it, since
* after the TLB flush no one should write to it.
*/
memset(kasan_zero_page, 0, PAGE_SIZE);
for (i = 0; i < PTRS_PER_PTE; i++) {
pte_t pte = __pte(__pa(kasan_zero_page) | __PAGE_KERNEL_RO | _PAGE_ENC);
set_pte(&kasan_zero_pte[i], pte);
}
/* Flush TLBs again to be sure that write protection applied. */
__flush_tlb_all();
init_task.kasan_depth = 0;
pr_info("KernelAddressSanitizer initialized\n");
}