Message ID | 1656578288-11259-1-git-send-email-guanghuifeng@linux.alibaba.com (mailing list archive) |
---|---|
State | New |
Headers | show |
Series | [v2] arm64: mm: fix linear mapping mem access performance degradation | expand |
I'm very sorry! Please DO'T review this patch, I have rewrited it. 在 2022/6/30 16:38, Guanghui Feng 写道: > The arm64 can build 2M/1G block/sectiion mapping. When using DMA/DMA32 zone > (enable crashkernel, disable rodata full, disable kfence), the mem_map will > use non block/section mapping(for crashkernel requires to shrink the region > in page granularity). But it will degrade performance when doing larging > continuous mem access in kernel(memcpy/memmove, etc). > > There are many changes and discussions: > commit 031495635b46 ("arm64: Do not defer reserve_crashkernel() for > platforms with no DMA memory zones") > commit 0a30c53573b0 ("arm64: mm: Move reserve_crashkernel() into > mem_init()") > commit 2687275a5843 ("arm64: Force NO_BLOCK_MAPPINGS if crashkernel > reservation is required") > > This patch changes mem_map to use block/section mapping with crashkernel. > Firstly, do block/section mapping(normally 2M or 1G) for all avail mem at > mem_map, reserve crashkernel memory. And then walking pagetable to split > block/section mapping to non block/section mapping(normally 4K) [[[only]]] > for crashkernel mem. So the linear mem mapping use block/section mapping > as more as possible. We will reduce the cpu dTLB miss conspicuously, and > accelerate mem access about 10-20% performance improvement. > > I have tested it with pft(Page Fault Test) and fio, obtained great > performace improvement. > > For fio test: > 1.prepare ramdisk > modprobe -r brd > modprobe brd rd_nr=1 rd_size=67108864 > dmsetup remove_all > wipefs -a --force /dev/ram0 > mkfs -t ext4 -E lazy_itable_init=0,lazy_journal_init=0 -q -F /dev/ram0 > mkdir -p /fs/ram0 > mount -t ext4 /dev/ram0 /fs/ram0 > > 2.prepare fio paremeter in x.fio file: > [global] > bs=4k > ioengine=psync > iodepth=128 > size=32G > direct=1 > invalidate=1 > group_reporting > thread=1 > rw=read > directory=/fs/ram0 > numjobs=1 > > [task_0] > cpus_allowed=16 > stonewall=1 > > 3.run testcase: > perf stat -e dTLB-load-misses fio x.fio > > 4.contrast > ------------------------ > without patch with patch > fio READ aggrb=1493.2MB/s aggrb=1775.3MB/s > dTLB-load-misses 1,818,320,693 438,729,774 > time elapsed(s) 70.500326434 62.877316408 > user(s) 15.926332000 15.684721000 > sys(s) 54.211939000 47.046165000 > > 5.conclusion > Using this patch will reduce dTLB misses and improve performace greatly. > > Signed-off-by: Guanghui Feng <guanghuifeng@linux.alibaba.com> > --- > arch/arm64/include/asm/mmu.h | 1 + > arch/arm64/mm/init.c | 8 +-- > arch/arm64/mm/mmu.c | 163 ++++++++++++++++++++++++++++++------------- > 3 files changed, 116 insertions(+), 56 deletions(-) > > diff --git a/arch/arm64/include/asm/mmu.h b/arch/arm64/include/asm/mmu.h > index 48f8466..1a46b81 100644 > --- a/arch/arm64/include/asm/mmu.h > +++ b/arch/arm64/include/asm/mmu.h > @@ -63,6 +63,7 @@ static inline bool arm64_kernel_unmapped_at_el0(void) > extern void arm64_memblock_init(void); > extern void paging_init(void); > extern void bootmem_init(void); > +extern void map_crashkernel(void); > extern void __iomem *early_io_map(phys_addr_t phys, unsigned long virt); > extern void init_mem_pgprot(void); > extern void create_pgd_mapping(struct mm_struct *mm, phys_addr_t phys, > diff --git a/arch/arm64/mm/init.c b/arch/arm64/mm/init.c > index 339ee84..241d27e 100644 > --- a/arch/arm64/mm/init.c > +++ b/arch/arm64/mm/init.c > @@ -190,6 +190,7 @@ static void __init reserve_crashkernel(void) > crashk_res.start = crash_base; > crashk_res.end = crash_base + crash_size - 1; > insert_resource(&iomem_resource, &crashk_res); > + map_crashkernel(); > } > > /* > @@ -388,10 +389,6 @@ void __init arm64_memblock_init(void) > } > > early_init_fdt_scan_reserved_mem(); > - > - if (!IS_ENABLED(CONFIG_ZONE_DMA) && !IS_ENABLED(CONFIG_ZONE_DMA32)) > - reserve_crashkernel(); > - > high_memory = __va(memblock_end_of_DRAM() - 1) + 1; > } > > @@ -438,8 +435,7 @@ void __init bootmem_init(void) > * request_standard_resources() depends on crashkernel's memory being > * reserved, so do it here. > */ > - if (IS_ENABLED(CONFIG_ZONE_DMA) || IS_ENABLED(CONFIG_ZONE_DMA32)) > - reserve_crashkernel(); > + reserve_crashkernel(); > > memblock_dump_all(); > } > diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c > index 626ec32..c832848 100644 > --- a/arch/arm64/mm/mmu.c > +++ b/arch/arm64/mm/mmu.c > @@ -42,6 +42,7 @@ > #define NO_BLOCK_MAPPINGS BIT(0) > #define NO_CONT_MAPPINGS BIT(1) > #define NO_EXEC_MAPPINGS BIT(2) /* assumes FEAT_HPDS is not used */ > +#define NO_SEC_REMAPPINGS BIT(3) /* rebuild with non block/sec mapping*/ > > u64 idmap_t0sz = TCR_T0SZ(VA_BITS_MIN); > u64 idmap_ptrs_per_pgd = PTRS_PER_PGD; > @@ -156,11 +157,12 @@ static bool pgattr_change_is_safe(u64 old, u64 new) > } > > static void init_pte(pmd_t *pmdp, unsigned long addr, unsigned long end, > - phys_addr_t phys, pgprot_t prot) > + phys_addr_t phys, pgprot_t prot, int flags) > { > pte_t *ptep; > > - ptep = pte_set_fixmap_offset(pmdp, addr); > + ptep = (flags & NO_SEC_REMAPPINGS) ? pte_offset_kernel(pmdp, addr) : > + pte_set_fixmap_offset(pmdp, addr); > do { > pte_t old_pte = READ_ONCE(*ptep); > > @@ -176,7 +178,8 @@ static void init_pte(pmd_t *pmdp, unsigned long addr, unsigned long end, > phys += PAGE_SIZE; > } while (ptep++, addr += PAGE_SIZE, addr != end); > > - pte_clear_fixmap(); > + if (!(flags & NO_SEC_REMAPPINGS)) > + pte_clear_fixmap(); > } > > static void alloc_init_cont_pte(pmd_t *pmdp, unsigned long addr, > @@ -208,11 +211,12 @@ static void alloc_init_cont_pte(pmd_t *pmdp, unsigned long addr, > next = pte_cont_addr_end(addr, end); > > /* use a contiguous mapping if the range is suitably aligned */ > - if ((((addr | next | phys) & ~CONT_PTE_MASK) == 0) && > + if (!(flags & NO_SEC_REMAPPINGS) && > + (((addr | next | phys) & ~CONT_PTE_MASK) == 0) && > (flags & NO_CONT_MAPPINGS) == 0) > __prot = __pgprot(pgprot_val(prot) | PTE_CONT); > > - init_pte(pmdp, addr, next, phys, __prot); > + init_pte(pmdp, addr, next, phys, __prot, flags); > > phys += next - addr; > } while (addr = next, addr != end); > @@ -224,15 +228,44 @@ static void init_pmd(pud_t *pudp, unsigned long addr, unsigned long end, > { > unsigned long next; > pmd_t *pmdp; > + phys_addr_t map_offset; > + pmdval_t pmdval; > > - pmdp = pmd_set_fixmap_offset(pudp, addr); > + pmdp = (flags & NO_SEC_REMAPPINGS) ? pmd_offset(pudp, addr) : > + pmd_set_fixmap_offset(pudp, addr); > do { > pmd_t old_pmd = READ_ONCE(*pmdp); > > next = pmd_addr_end(addr, end); > > - /* try section mapping first */ > - if (((addr | next | phys) & ~PMD_MASK) == 0 && > + if (flags & NO_SEC_REMAPPINGS) { > + if (!pmd_none(*pmdp) && pmd_sect(*pmdp)) { > + phys_addr_t pte_phys = pgtable_alloc(PAGE_SHIFT); > + pmd_clear(pmdp); > + pmdval = PMD_TYPE_TABLE | PMD_TABLE_UXN; > + if (flags & NO_EXEC_MAPPINGS) > + pmdval |= PMD_TABLE_PXN; > + __pmd_populate(pmdp, pte_phys, pmdval); > + flush_tlb_kernel_range(addr, addr + PAGE_SIZE); > + > + map_offset = addr - (addr & PMD_MASK); > + if (map_offset) > + alloc_init_cont_pte(pmdp, addr & PMD_MASK, addr, > + phys - map_offset, prot, > + pgtable_alloc, > + flags & (~NO_SEC_REMAPPINGS)); > + > + if (next < (addr & PMD_MASK) + PMD_SIZE) > + alloc_init_cont_pte(pmdp, next, > + (addr & PUD_MASK) + PUD_SIZE, > + next - addr + phys, > + prot, pgtable_alloc, > + flags & (~NO_SEC_REMAPPINGS)); > + } > + alloc_init_cont_pte(pmdp, addr, next, phys, prot, > + pgtable_alloc, flags); > + }/* try section mapping first */ > + else if (((addr | next | phys) & ~PMD_MASK) == 0 && > (flags & NO_BLOCK_MAPPINGS) == 0) { > pmd_set_huge(pmdp, phys, prot); > > @@ -252,7 +285,8 @@ static void init_pmd(pud_t *pudp, unsigned long addr, unsigned long end, > phys += next - addr; > } while (pmdp++, addr = next, addr != end); > > - pmd_clear_fixmap(); > + if (!(flags & NO_SEC_REMAPPINGS)) > + pmd_clear_fixmap(); > } > > static void alloc_init_cont_pmd(pud_t *pudp, unsigned long addr, > @@ -286,7 +320,8 @@ static void alloc_init_cont_pmd(pud_t *pudp, unsigned long addr, > next = pmd_cont_addr_end(addr, end); > > /* use a contiguous mapping if the range is suitably aligned */ > - if ((((addr | next | phys) & ~CONT_PMD_MASK) == 0) && > + if (!(flags & NO_SEC_REMAPPINGS) && > + (((addr | next | phys) & ~CONT_PMD_MASK) == 0) && > (flags & NO_CONT_MAPPINGS) == 0) > __prot = __pgprot(pgprot_val(prot) | PTE_CONT); > > @@ -301,7 +336,9 @@ static void alloc_init_pud(pgd_t *pgdp, unsigned long addr, unsigned long end, > phys_addr_t (*pgtable_alloc)(int), > int flags) > { > + phys_addr_t map_offset; > unsigned long next; > + pudval_t pudval; > pud_t *pudp; > p4d_t *p4dp = p4d_offset(pgdp, addr); > p4d_t p4d = READ_ONCE(*p4dp); > @@ -325,18 +362,48 @@ static void alloc_init_pud(pgd_t *pgdp, unsigned long addr, unsigned long end, > */ > if (system_state != SYSTEM_BOOTING) > mutex_lock(&fixmap_lock); > - pudp = pud_set_fixmap_offset(p4dp, addr); > + > + pudp = (flags & NO_SEC_REMAPPINGS) ? pud_offset(p4dp, addr) : > + pud_set_fixmap_offset(p4dp, addr); > do { > pud_t old_pud = READ_ONCE(*pudp); > - > next = pud_addr_end(addr, end); > > + if (flags & NO_SEC_REMAPPINGS) { > + if (!pud_none(*pudp) && pud_sect(*pudp)) { > + phys_addr_t pmd_phys = pgtable_alloc(PMD_SHIFT); > + pud_clear(pudp); > + pudval = PUD_TYPE_TABLE | PUD_TABLE_UXN; > + if (flags & NO_EXEC_MAPPINGS) > + pudval |= PUD_TABLE_PXN; > + > + __pud_populate(pudp, pmd_phys, pudval); > + flush_tlb_kernel_range(addr, addr + PAGE_SIZE); > + > + map_offset = addr - (addr & PUD_MASK); > + if (map_offset) > + alloc_init_cont_pmd(pudp, addr & PUD_MASK, > + addr, phys - map_offset, > + prot, pgtable_alloc, > + flags & (~NO_SEC_REMAPPINGS)); > + > + if (next < (addr & PUD_MASK) + PUD_SIZE) > + alloc_init_cont_pmd(pudp, next, > + (addr & PUD_MASK) + > + PUD_SIZE, > + next - addr + phys, > + prot, pgtable_alloc, > + flags & (~NO_SEC_REMAPPINGS)); > + } > + alloc_init_cont_pmd(pudp, addr, next, phys, prot, > + pgtable_alloc, flags); > + } > /* > * For 4K granule only, attempt to put down a 1GB block > */ > - if (pud_sect_supported() && > - ((addr | next | phys) & ~PUD_MASK) == 0 && > - (flags & NO_BLOCK_MAPPINGS) == 0) { > + else if (pud_sect_supported() && > + ((addr | next | phys) & ~PUD_MASK) == 0 && > + (flags & NO_BLOCK_MAPPINGS) == 0) { > pud_set_huge(pudp, phys, prot); > > /* > @@ -355,7 +422,8 @@ static void alloc_init_pud(pgd_t *pgdp, unsigned long addr, unsigned long end, > phys += next - addr; > } while (pudp++, addr = next, addr != end); > > - pud_clear_fixmap(); > + if (!(flags & NO_SEC_REMAPPINGS)) > + pud_clear_fixmap(); > if (system_state != SYSTEM_BOOTING) > mutex_unlock(&fixmap_lock); > } > @@ -483,20 +551,39 @@ void __init mark_linear_text_alias_ro(void) > PAGE_KERNEL_RO); > } > > -static bool crash_mem_map __initdata; > +#ifdef CONFIG_KEXEC_CORE > +static phys_addr_t __init early_crashkernel_pgtable_alloc(int shift) > +{ > + phys_addr_t phys; > + void *ptr; > > -static int __init enable_crash_mem_map(char *arg) > + phys = memblock_phys_alloc_range(PAGE_SIZE, PAGE_SIZE, 0, > + MEMBLOCK_ALLOC_NOLEAKTRACE); > + if (!phys) > + panic("Failed to allocate page table page\n"); > + > + ptr = (void *)__phys_to_virt(phys); > + memset(ptr, 0, PAGE_SIZE); > + return phys; > +} > + > +void __init map_crashkernel(void) > { > - /* > - * Proper parameter parsing is done by reserve_crashkernel(). We only > - * need to know if the linear map has to avoid block mappings so that > - * the crashkernel reservations can be unmapped later. > - */ > - crash_mem_map = true; > + if (can_set_direct_map() || IS_ENABLED(CONFIG_KFENCE)) > + return; > > - return 0; > + if (!crashk_res.end) > + return; > + > + __create_pgd_mapping(swapper_pg_dir, crashk_res.start, > + __phys_to_virt(crashk_res.start), > + crashk_res.end + 1 - crashk_res.start, PAGE_KERNEL, > + early_crashkernel_pgtable_alloc, > + NO_EXEC_MAPPINGS | NO_SEC_REMAPPINGS); > } > -early_param("crashkernel", enable_crash_mem_map); > +#else > +void __init map_crashkernel(void) {} > +#endif > > static void __init map_mem(pgd_t *pgdp) > { > @@ -527,17 +614,6 @@ static void __init map_mem(pgd_t *pgdp) > */ > memblock_mark_nomap(kernel_start, kernel_end - kernel_start); > > -#ifdef CONFIG_KEXEC_CORE > - if (crash_mem_map) { > - if (IS_ENABLED(CONFIG_ZONE_DMA) || > - IS_ENABLED(CONFIG_ZONE_DMA32)) > - flags |= NO_BLOCK_MAPPINGS | NO_CONT_MAPPINGS; > - else if (crashk_res.end) > - memblock_mark_nomap(crashk_res.start, > - resource_size(&crashk_res)); > - } > -#endif > - > /* map all the memory banks */ > for_each_mem_range(i, &start, &end) { > if (start >= end) > @@ -570,19 +646,6 @@ static void __init map_mem(pgd_t *pgdp) > * in page granularity and put back unused memory to buddy system > * through /sys/kernel/kexec_crash_size interface. > */ > -#ifdef CONFIG_KEXEC_CORE > - if (crash_mem_map && > - !IS_ENABLED(CONFIG_ZONE_DMA) && !IS_ENABLED(CONFIG_ZONE_DMA32)) { > - if (crashk_res.end) { > - __map_memblock(pgdp, crashk_res.start, > - crashk_res.end + 1, > - PAGE_KERNEL, > - NO_BLOCK_MAPPINGS | NO_CONT_MAPPINGS); > - memblock_clear_nomap(crashk_res.start, > - resource_size(&crashk_res)); > - } > - } > -#endif > } > > void mark_rodata_ro(void)
Hi Guanghui,
Thank you for the patch! Perhaps something to improve:
[auto build test WARNING on v5.19-rc4]
[also build test WARNING on linus/master next-20220630]
[cannot apply to arm64/for-next/core]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]
url: https://github.com/intel-lab-lkp/linux/commits/Guanghui-Feng/arm64-mm-fix-linear-mapping-mem-access-performance-degradation/20220630-163924
base: 03c765b0e3b4cb5063276b086c76f7a612856a9a
config: arm64-allyesconfig (https://download.01.org/0day-ci/archive/20220701/202207010409.ajCVDXYO-lkp@intel.com/config)
compiler: aarch64-linux-gcc (GCC) 11.3.0
reproduce (this is a W=1 build):
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
chmod +x ~/bin/make.cross
# https://github.com/intel-lab-lkp/linux/commit/168f0d56f0a26c4f7a9470d6f1c398b4e1c1b5b9
git remote add linux-review https://github.com/intel-lab-lkp/linux
git fetch --no-tags linux-review Guanghui-Feng/arm64-mm-fix-linear-mapping-mem-access-performance-degradation/20220630-163924
git checkout 168f0d56f0a26c4f7a9470d6f1c398b4e1c1b5b9
# save the config file
mkdir build_dir && cp config build_dir/.config
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-11.3.0 make.cross W=1 O=build_dir ARCH=arm64 SHELL=/bin/bash arch/arm64/mm/
If you fix the issue, kindly add following tag where applicable
Reported-by: kernel test robot <lkp@intel.com>
All warnings (new ones prefixed by >>):
arch/arm64/mm/mmu.c: In function 'alloc_init_pud':
>> arch/arm64/mm/mmu.c:426:35: warning: suggest braces around empty body in an 'if' statement [-Wempty-body]
426 | pud_clear_fixmap();
| ^
vim +/if +426 arch/arm64/mm/mmu.c
d27cfa1fc823d3 Ard Biesheuvel 2017-03-09 333
20a004e7b017cc Will Deacon 2018-02-15 334 static void alloc_init_pud(pgd_t *pgdp, unsigned long addr, unsigned long end,
da141706aea52c Laura Abbott 2015-01-21 335 phys_addr_t phys, pgprot_t prot,
90292aca9854a2 Yu Zhao 2019-03-11 336 phys_addr_t (*pgtable_alloc)(int),
c0951366d4b7e0 Ard Biesheuvel 2017-03-09 337 int flags)
c1cc1552616d0f Catalin Marinas 2012-03-05 338 {
168f0d56f0a26c Guanghui Feng 2022-06-30 339 phys_addr_t map_offset;
c1cc1552616d0f Catalin Marinas 2012-03-05 340 unsigned long next;
168f0d56f0a26c Guanghui Feng 2022-06-30 341 pudval_t pudval;
20a004e7b017cc Will Deacon 2018-02-15 342 pud_t *pudp;
e9f6376858b979 Mike Rapoport 2020-06-04 343 p4d_t *p4dp = p4d_offset(pgdp, addr);
e9f6376858b979 Mike Rapoport 2020-06-04 344 p4d_t p4d = READ_ONCE(*p4dp);
c1cc1552616d0f Catalin Marinas 2012-03-05 345
e9f6376858b979 Mike Rapoport 2020-06-04 346 if (p4d_none(p4d)) {
87143f404f338d Ard Biesheuvel 2021-03-10 347 p4dval_t p4dval = P4D_TYPE_TABLE | P4D_TABLE_UXN;
132233a759580f Laura Abbott 2016-02-05 348 phys_addr_t pud_phys;
87143f404f338d Ard Biesheuvel 2021-03-10 349
87143f404f338d Ard Biesheuvel 2021-03-10 350 if (flags & NO_EXEC_MAPPINGS)
87143f404f338d Ard Biesheuvel 2021-03-10 351 p4dval |= P4D_TABLE_PXN;
132233a759580f Laura Abbott 2016-02-05 352 BUG_ON(!pgtable_alloc);
90292aca9854a2 Yu Zhao 2019-03-11 353 pud_phys = pgtable_alloc(PUD_SHIFT);
87143f404f338d Ard Biesheuvel 2021-03-10 354 __p4d_populate(p4dp, pud_phys, p4dval);
e9f6376858b979 Mike Rapoport 2020-06-04 355 p4d = READ_ONCE(*p4dp);
c79b954bf6c006 Jungseok Lee 2014-05-12 356 }
e9f6376858b979 Mike Rapoport 2020-06-04 357 BUG_ON(p4d_bad(p4d));
c79b954bf6c006 Jungseok Lee 2014-05-12 358
ee017ee353506f Jianyong Wu 2022-02-01 359 /*
ee017ee353506f Jianyong Wu 2022-02-01 360 * No need for locking during early boot. And it doesn't work as
ee017ee353506f Jianyong Wu 2022-02-01 361 * expected with KASLR enabled.
ee017ee353506f Jianyong Wu 2022-02-01 362 */
ee017ee353506f Jianyong Wu 2022-02-01 363 if (system_state != SYSTEM_BOOTING)
ee017ee353506f Jianyong Wu 2022-02-01 364 mutex_lock(&fixmap_lock);
168f0d56f0a26c Guanghui Feng 2022-06-30 365
168f0d56f0a26c Guanghui Feng 2022-06-30 366 pudp = (flags & NO_SEC_REMAPPINGS) ? pud_offset(p4dp, addr) :
168f0d56f0a26c Guanghui Feng 2022-06-30 367 pud_set_fixmap_offset(p4dp, addr);
c1cc1552616d0f Catalin Marinas 2012-03-05 368 do {
20a004e7b017cc Will Deacon 2018-02-15 369 pud_t old_pud = READ_ONCE(*pudp);
c1cc1552616d0f Catalin Marinas 2012-03-05 370 next = pud_addr_end(addr, end);
206a2a73a62d37 Steve Capper 2014-05-06 371
168f0d56f0a26c Guanghui Feng 2022-06-30 372 if (flags & NO_SEC_REMAPPINGS) {
168f0d56f0a26c Guanghui Feng 2022-06-30 373 if (!pud_none(*pudp) && pud_sect(*pudp)) {
168f0d56f0a26c Guanghui Feng 2022-06-30 374 phys_addr_t pmd_phys = pgtable_alloc(PMD_SHIFT);
168f0d56f0a26c Guanghui Feng 2022-06-30 375 pud_clear(pudp);
168f0d56f0a26c Guanghui Feng 2022-06-30 376 pudval = PUD_TYPE_TABLE | PUD_TABLE_UXN;
168f0d56f0a26c Guanghui Feng 2022-06-30 377 if (flags & NO_EXEC_MAPPINGS)
168f0d56f0a26c Guanghui Feng 2022-06-30 378 pudval |= PUD_TABLE_PXN;
168f0d56f0a26c Guanghui Feng 2022-06-30 379
168f0d56f0a26c Guanghui Feng 2022-06-30 380 __pud_populate(pudp, pmd_phys, pudval);
168f0d56f0a26c Guanghui Feng 2022-06-30 381 flush_tlb_kernel_range(addr, addr + PAGE_SIZE);
168f0d56f0a26c Guanghui Feng 2022-06-30 382
168f0d56f0a26c Guanghui Feng 2022-06-30 383 map_offset = addr - (addr & PUD_MASK);
168f0d56f0a26c Guanghui Feng 2022-06-30 384 if (map_offset)
168f0d56f0a26c Guanghui Feng 2022-06-30 385 alloc_init_cont_pmd(pudp, addr & PUD_MASK,
168f0d56f0a26c Guanghui Feng 2022-06-30 386 addr, phys - map_offset,
168f0d56f0a26c Guanghui Feng 2022-06-30 387 prot, pgtable_alloc,
168f0d56f0a26c Guanghui Feng 2022-06-30 388 flags & (~NO_SEC_REMAPPINGS));
168f0d56f0a26c Guanghui Feng 2022-06-30 389
168f0d56f0a26c Guanghui Feng 2022-06-30 390 if (next < (addr & PUD_MASK) + PUD_SIZE)
168f0d56f0a26c Guanghui Feng 2022-06-30 391 alloc_init_cont_pmd(pudp, next,
168f0d56f0a26c Guanghui Feng 2022-06-30 392 (addr & PUD_MASK) +
168f0d56f0a26c Guanghui Feng 2022-06-30 393 PUD_SIZE,
168f0d56f0a26c Guanghui Feng 2022-06-30 394 next - addr + phys,
168f0d56f0a26c Guanghui Feng 2022-06-30 395 prot, pgtable_alloc,
168f0d56f0a26c Guanghui Feng 2022-06-30 396 flags & (~NO_SEC_REMAPPINGS));
168f0d56f0a26c Guanghui Feng 2022-06-30 397 }
168f0d56f0a26c Guanghui Feng 2022-06-30 398 alloc_init_cont_pmd(pudp, addr, next, phys, prot,
168f0d56f0a26c Guanghui Feng 2022-06-30 399 pgtable_alloc, flags);
168f0d56f0a26c Guanghui Feng 2022-06-30 400 }
206a2a73a62d37 Steve Capper 2014-05-06 401 /*
206a2a73a62d37 Steve Capper 2014-05-06 402 * For 4K granule only, attempt to put down a 1GB block
206a2a73a62d37 Steve Capper 2014-05-06 403 */
168f0d56f0a26c Guanghui Feng 2022-06-30 404 else if (pud_sect_supported() &&
1310222c276b79 Anshuman Khandual 2022-02-16 405 ((addr | next | phys) & ~PUD_MASK) == 0 &&
c0951366d4b7e0 Ard Biesheuvel 2017-03-09 406 (flags & NO_BLOCK_MAPPINGS) == 0) {
20a004e7b017cc Will Deacon 2018-02-15 407 pud_set_huge(pudp, phys, prot);
206a2a73a62d37 Steve Capper 2014-05-06 408
206a2a73a62d37 Steve Capper 2014-05-06 409 /*
e98216b52176ba Ard Biesheuvel 2016-10-21 410 * After the PUD entry has been populated once, we
e98216b52176ba Ard Biesheuvel 2016-10-21 411 * only allow updates to the permission attributes.
206a2a73a62d37 Steve Capper 2014-05-06 412 */
e98216b52176ba Ard Biesheuvel 2016-10-21 413 BUG_ON(!pgattr_change_is_safe(pud_val(old_pud),
20a004e7b017cc Will Deacon 2018-02-15 414 READ_ONCE(pud_val(*pudp))));
206a2a73a62d37 Steve Capper 2014-05-06 415 } else {
20a004e7b017cc Will Deacon 2018-02-15 416 alloc_init_cont_pmd(pudp, addr, next, phys, prot,
c0951366d4b7e0 Ard Biesheuvel 2017-03-09 417 pgtable_alloc, flags);
e98216b52176ba Ard Biesheuvel 2016-10-21 418
e98216b52176ba Ard Biesheuvel 2016-10-21 419 BUG_ON(pud_val(old_pud) != 0 &&
20a004e7b017cc Will Deacon 2018-02-15 420 pud_val(old_pud) != READ_ONCE(pud_val(*pudp)));
206a2a73a62d37 Steve Capper 2014-05-06 421 }
c1cc1552616d0f Catalin Marinas 2012-03-05 422 phys += next - addr;
20a004e7b017cc Will Deacon 2018-02-15 423 } while (pudp++, addr = next, addr != end);
f4710445458c0a Mark Rutland 2016-01-25 424
168f0d56f0a26c Guanghui Feng 2022-06-30 425 if (!(flags & NO_SEC_REMAPPINGS))
f4710445458c0a Mark Rutland 2016-01-25 @426 pud_clear_fixmap();
ee017ee353506f Jianyong Wu 2022-02-01 427 if (system_state != SYSTEM_BOOTING)
ee017ee353506f Jianyong Wu 2022-02-01 428 mutex_unlock(&fixmap_lock);
c1cc1552616d0f Catalin Marinas 2012-03-05 429 }
c1cc1552616d0f Catalin Marinas 2012-03-05 430
diff --git a/arch/arm64/include/asm/mmu.h b/arch/arm64/include/asm/mmu.h index 48f8466..1a46b81 100644 --- a/arch/arm64/include/asm/mmu.h +++ b/arch/arm64/include/asm/mmu.h @@ -63,6 +63,7 @@ static inline bool arm64_kernel_unmapped_at_el0(void) extern void arm64_memblock_init(void); extern void paging_init(void); extern void bootmem_init(void); +extern void map_crashkernel(void); extern void __iomem *early_io_map(phys_addr_t phys, unsigned long virt); extern void init_mem_pgprot(void); extern void create_pgd_mapping(struct mm_struct *mm, phys_addr_t phys, diff --git a/arch/arm64/mm/init.c b/arch/arm64/mm/init.c index 339ee84..241d27e 100644 --- a/arch/arm64/mm/init.c +++ b/arch/arm64/mm/init.c @@ -190,6 +190,7 @@ static void __init reserve_crashkernel(void) crashk_res.start = crash_base; crashk_res.end = crash_base + crash_size - 1; insert_resource(&iomem_resource, &crashk_res); + map_crashkernel(); } /* @@ -388,10 +389,6 @@ void __init arm64_memblock_init(void) } early_init_fdt_scan_reserved_mem(); - - if (!IS_ENABLED(CONFIG_ZONE_DMA) && !IS_ENABLED(CONFIG_ZONE_DMA32)) - reserve_crashkernel(); - high_memory = __va(memblock_end_of_DRAM() - 1) + 1; } @@ -438,8 +435,7 @@ void __init bootmem_init(void) * request_standard_resources() depends on crashkernel's memory being * reserved, so do it here. */ - if (IS_ENABLED(CONFIG_ZONE_DMA) || IS_ENABLED(CONFIG_ZONE_DMA32)) - reserve_crashkernel(); + reserve_crashkernel(); memblock_dump_all(); } diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c index 626ec32..c832848 100644 --- a/arch/arm64/mm/mmu.c +++ b/arch/arm64/mm/mmu.c @@ -42,6 +42,7 @@ #define NO_BLOCK_MAPPINGS BIT(0) #define NO_CONT_MAPPINGS BIT(1) #define NO_EXEC_MAPPINGS BIT(2) /* assumes FEAT_HPDS is not used */ +#define NO_SEC_REMAPPINGS BIT(3) /* rebuild with non block/sec mapping*/ u64 idmap_t0sz = TCR_T0SZ(VA_BITS_MIN); u64 idmap_ptrs_per_pgd = PTRS_PER_PGD; @@ -156,11 +157,12 @@ static bool pgattr_change_is_safe(u64 old, u64 new) } static void init_pte(pmd_t *pmdp, unsigned long addr, unsigned long end, - phys_addr_t phys, pgprot_t prot) + phys_addr_t phys, pgprot_t prot, int flags) { pte_t *ptep; - ptep = pte_set_fixmap_offset(pmdp, addr); + ptep = (flags & NO_SEC_REMAPPINGS) ? pte_offset_kernel(pmdp, addr) : + pte_set_fixmap_offset(pmdp, addr); do { pte_t old_pte = READ_ONCE(*ptep); @@ -176,7 +178,8 @@ static void init_pte(pmd_t *pmdp, unsigned long addr, unsigned long end, phys += PAGE_SIZE; } while (ptep++, addr += PAGE_SIZE, addr != end); - pte_clear_fixmap(); + if (!(flags & NO_SEC_REMAPPINGS)) + pte_clear_fixmap(); } static void alloc_init_cont_pte(pmd_t *pmdp, unsigned long addr, @@ -208,11 +211,12 @@ static void alloc_init_cont_pte(pmd_t *pmdp, unsigned long addr, next = pte_cont_addr_end(addr, end); /* use a contiguous mapping if the range is suitably aligned */ - if ((((addr | next | phys) & ~CONT_PTE_MASK) == 0) && + if (!(flags & NO_SEC_REMAPPINGS) && + (((addr | next | phys) & ~CONT_PTE_MASK) == 0) && (flags & NO_CONT_MAPPINGS) == 0) __prot = __pgprot(pgprot_val(prot) | PTE_CONT); - init_pte(pmdp, addr, next, phys, __prot); + init_pte(pmdp, addr, next, phys, __prot, flags); phys += next - addr; } while (addr = next, addr != end); @@ -224,15 +228,44 @@ static void init_pmd(pud_t *pudp, unsigned long addr, unsigned long end, { unsigned long next; pmd_t *pmdp; + phys_addr_t map_offset; + pmdval_t pmdval; - pmdp = pmd_set_fixmap_offset(pudp, addr); + pmdp = (flags & NO_SEC_REMAPPINGS) ? pmd_offset(pudp, addr) : + pmd_set_fixmap_offset(pudp, addr); do { pmd_t old_pmd = READ_ONCE(*pmdp); next = pmd_addr_end(addr, end); - /* try section mapping first */ - if (((addr | next | phys) & ~PMD_MASK) == 0 && + if (flags & NO_SEC_REMAPPINGS) { + if (!pmd_none(*pmdp) && pmd_sect(*pmdp)) { + phys_addr_t pte_phys = pgtable_alloc(PAGE_SHIFT); + pmd_clear(pmdp); + pmdval = PMD_TYPE_TABLE | PMD_TABLE_UXN; + if (flags & NO_EXEC_MAPPINGS) + pmdval |= PMD_TABLE_PXN; + __pmd_populate(pmdp, pte_phys, pmdval); + flush_tlb_kernel_range(addr, addr + PAGE_SIZE); + + map_offset = addr - (addr & PMD_MASK); + if (map_offset) + alloc_init_cont_pte(pmdp, addr & PMD_MASK, addr, + phys - map_offset, prot, + pgtable_alloc, + flags & (~NO_SEC_REMAPPINGS)); + + if (next < (addr & PMD_MASK) + PMD_SIZE) + alloc_init_cont_pte(pmdp, next, + (addr & PUD_MASK) + PUD_SIZE, + next - addr + phys, + prot, pgtable_alloc, + flags & (~NO_SEC_REMAPPINGS)); + } + alloc_init_cont_pte(pmdp, addr, next, phys, prot, + pgtable_alloc, flags); + }/* try section mapping first */ + else if (((addr | next | phys) & ~PMD_MASK) == 0 && (flags & NO_BLOCK_MAPPINGS) == 0) { pmd_set_huge(pmdp, phys, prot); @@ -252,7 +285,8 @@ static void init_pmd(pud_t *pudp, unsigned long addr, unsigned long end, phys += next - addr; } while (pmdp++, addr = next, addr != end); - pmd_clear_fixmap(); + if (!(flags & NO_SEC_REMAPPINGS)) + pmd_clear_fixmap(); } static void alloc_init_cont_pmd(pud_t *pudp, unsigned long addr, @@ -286,7 +320,8 @@ static void alloc_init_cont_pmd(pud_t *pudp, unsigned long addr, next = pmd_cont_addr_end(addr, end); /* use a contiguous mapping if the range is suitably aligned */ - if ((((addr | next | phys) & ~CONT_PMD_MASK) == 0) && + if (!(flags & NO_SEC_REMAPPINGS) && + (((addr | next | phys) & ~CONT_PMD_MASK) == 0) && (flags & NO_CONT_MAPPINGS) == 0) __prot = __pgprot(pgprot_val(prot) | PTE_CONT); @@ -301,7 +336,9 @@ static void alloc_init_pud(pgd_t *pgdp, unsigned long addr, unsigned long end, phys_addr_t (*pgtable_alloc)(int), int flags) { + phys_addr_t map_offset; unsigned long next; + pudval_t pudval; pud_t *pudp; p4d_t *p4dp = p4d_offset(pgdp, addr); p4d_t p4d = READ_ONCE(*p4dp); @@ -325,18 +362,48 @@ static void alloc_init_pud(pgd_t *pgdp, unsigned long addr, unsigned long end, */ if (system_state != SYSTEM_BOOTING) mutex_lock(&fixmap_lock); - pudp = pud_set_fixmap_offset(p4dp, addr); + + pudp = (flags & NO_SEC_REMAPPINGS) ? pud_offset(p4dp, addr) : + pud_set_fixmap_offset(p4dp, addr); do { pud_t old_pud = READ_ONCE(*pudp); - next = pud_addr_end(addr, end); + if (flags & NO_SEC_REMAPPINGS) { + if (!pud_none(*pudp) && pud_sect(*pudp)) { + phys_addr_t pmd_phys = pgtable_alloc(PMD_SHIFT); + pud_clear(pudp); + pudval = PUD_TYPE_TABLE | PUD_TABLE_UXN; + if (flags & NO_EXEC_MAPPINGS) + pudval |= PUD_TABLE_PXN; + + __pud_populate(pudp, pmd_phys, pudval); + flush_tlb_kernel_range(addr, addr + PAGE_SIZE); + + map_offset = addr - (addr & PUD_MASK); + if (map_offset) + alloc_init_cont_pmd(pudp, addr & PUD_MASK, + addr, phys - map_offset, + prot, pgtable_alloc, + flags & (~NO_SEC_REMAPPINGS)); + + if (next < (addr & PUD_MASK) + PUD_SIZE) + alloc_init_cont_pmd(pudp, next, + (addr & PUD_MASK) + + PUD_SIZE, + next - addr + phys, + prot, pgtable_alloc, + flags & (~NO_SEC_REMAPPINGS)); + } + alloc_init_cont_pmd(pudp, addr, next, phys, prot, + pgtable_alloc, flags); + } /* * For 4K granule only, attempt to put down a 1GB block */ - if (pud_sect_supported() && - ((addr | next | phys) & ~PUD_MASK) == 0 && - (flags & NO_BLOCK_MAPPINGS) == 0) { + else if (pud_sect_supported() && + ((addr | next | phys) & ~PUD_MASK) == 0 && + (flags & NO_BLOCK_MAPPINGS) == 0) { pud_set_huge(pudp, phys, prot); /* @@ -355,7 +422,8 @@ static void alloc_init_pud(pgd_t *pgdp, unsigned long addr, unsigned long end, phys += next - addr; } while (pudp++, addr = next, addr != end); - pud_clear_fixmap(); + if (!(flags & NO_SEC_REMAPPINGS)) + pud_clear_fixmap(); if (system_state != SYSTEM_BOOTING) mutex_unlock(&fixmap_lock); } @@ -483,20 +551,39 @@ void __init mark_linear_text_alias_ro(void) PAGE_KERNEL_RO); } -static bool crash_mem_map __initdata; +#ifdef CONFIG_KEXEC_CORE +static phys_addr_t __init early_crashkernel_pgtable_alloc(int shift) +{ + phys_addr_t phys; + void *ptr; -static int __init enable_crash_mem_map(char *arg) + phys = memblock_phys_alloc_range(PAGE_SIZE, PAGE_SIZE, 0, + MEMBLOCK_ALLOC_NOLEAKTRACE); + if (!phys) + panic("Failed to allocate page table page\n"); + + ptr = (void *)__phys_to_virt(phys); + memset(ptr, 0, PAGE_SIZE); + return phys; +} + +void __init map_crashkernel(void) { - /* - * Proper parameter parsing is done by reserve_crashkernel(). We only - * need to know if the linear map has to avoid block mappings so that - * the crashkernel reservations can be unmapped later. - */ - crash_mem_map = true; + if (can_set_direct_map() || IS_ENABLED(CONFIG_KFENCE)) + return; - return 0; + if (!crashk_res.end) + return; + + __create_pgd_mapping(swapper_pg_dir, crashk_res.start, + __phys_to_virt(crashk_res.start), + crashk_res.end + 1 - crashk_res.start, PAGE_KERNEL, + early_crashkernel_pgtable_alloc, + NO_EXEC_MAPPINGS | NO_SEC_REMAPPINGS); } -early_param("crashkernel", enable_crash_mem_map); +#else +void __init map_crashkernel(void) {} +#endif static void __init map_mem(pgd_t *pgdp) { @@ -527,17 +614,6 @@ static void __init map_mem(pgd_t *pgdp) */ memblock_mark_nomap(kernel_start, kernel_end - kernel_start); -#ifdef CONFIG_KEXEC_CORE - if (crash_mem_map) { - if (IS_ENABLED(CONFIG_ZONE_DMA) || - IS_ENABLED(CONFIG_ZONE_DMA32)) - flags |= NO_BLOCK_MAPPINGS | NO_CONT_MAPPINGS; - else if (crashk_res.end) - memblock_mark_nomap(crashk_res.start, - resource_size(&crashk_res)); - } -#endif - /* map all the memory banks */ for_each_mem_range(i, &start, &end) { if (start >= end) @@ -570,19 +646,6 @@ static void __init map_mem(pgd_t *pgdp) * in page granularity and put back unused memory to buddy system * through /sys/kernel/kexec_crash_size interface. */ -#ifdef CONFIG_KEXEC_CORE - if (crash_mem_map && - !IS_ENABLED(CONFIG_ZONE_DMA) && !IS_ENABLED(CONFIG_ZONE_DMA32)) { - if (crashk_res.end) { - __map_memblock(pgdp, crashk_res.start, - crashk_res.end + 1, - PAGE_KERNEL, - NO_BLOCK_MAPPINGS | NO_CONT_MAPPINGS); - memblock_clear_nomap(crashk_res.start, - resource_size(&crashk_res)); - } - } -#endif } void mark_rodata_ro(void)
The arm64 can build 2M/1G block/sectiion mapping. When using DMA/DMA32 zone (enable crashkernel, disable rodata full, disable kfence), the mem_map will use non block/section mapping(for crashkernel requires to shrink the region in page granularity). But it will degrade performance when doing larging continuous mem access in kernel(memcpy/memmove, etc). There are many changes and discussions: commit 031495635b46 ("arm64: Do not defer reserve_crashkernel() for platforms with no DMA memory zones") commit 0a30c53573b0 ("arm64: mm: Move reserve_crashkernel() into mem_init()") commit 2687275a5843 ("arm64: Force NO_BLOCK_MAPPINGS if crashkernel reservation is required") This patch changes mem_map to use block/section mapping with crashkernel. Firstly, do block/section mapping(normally 2M or 1G) for all avail mem at mem_map, reserve crashkernel memory. And then walking pagetable to split block/section mapping to non block/section mapping(normally 4K) [[[only]]] for crashkernel mem. So the linear mem mapping use block/section mapping as more as possible. We will reduce the cpu dTLB miss conspicuously, and accelerate mem access about 10-20% performance improvement. I have tested it with pft(Page Fault Test) and fio, obtained great performace improvement. For fio test: 1.prepare ramdisk modprobe -r brd modprobe brd rd_nr=1 rd_size=67108864 dmsetup remove_all wipefs -a --force /dev/ram0 mkfs -t ext4 -E lazy_itable_init=0,lazy_journal_init=0 -q -F /dev/ram0 mkdir -p /fs/ram0 mount -t ext4 /dev/ram0 /fs/ram0 2.prepare fio paremeter in x.fio file: [global] bs=4k ioengine=psync iodepth=128 size=32G direct=1 invalidate=1 group_reporting thread=1 rw=read directory=/fs/ram0 numjobs=1 [task_0] cpus_allowed=16 stonewall=1 3.run testcase: perf stat -e dTLB-load-misses fio x.fio 4.contrast ------------------------ without patch with patch fio READ aggrb=1493.2MB/s aggrb=1775.3MB/s dTLB-load-misses 1,818,320,693 438,729,774 time elapsed(s) 70.500326434 62.877316408 user(s) 15.926332000 15.684721000 sys(s) 54.211939000 47.046165000 5.conclusion Using this patch will reduce dTLB misses and improve performace greatly. Signed-off-by: Guanghui Feng <guanghuifeng@linux.alibaba.com> --- arch/arm64/include/asm/mmu.h | 1 + arch/arm64/mm/init.c | 8 +-- arch/arm64/mm/mmu.c | 163 ++++++++++++++++++++++++++++++------------- 3 files changed, 116 insertions(+), 56 deletions(-)