- Kernel - mailweb.openeuler.org

[openeuler:OLK-6.6 3135/3135] mm/gmem_phys.c:245:19: warning: no previous prototype for function 'gm_evict_page_locked'
by kernel test robot 13 Nov '25

13 Nov '25

tree: https://gitee.com/openeuler/kernel.git OLK-6.6 head: fbfb361fe93de1a562df2ace860f64c4067abfdc commit: 8f4f66a9d851f366942e8ee319292e41ad2baa95 [3135/3135] gmem: support swapping heterogeneous memory config: arm64-randconfig-001-20251112 (https://download.01.org/0day-ci/archive/20251113/202511130834.0qWsUngh-lkp@…) compiler: clang version 22.0.0git (https://github.com/llvm/llvm-project 996639d6ebb86ff15a8c99b67f1c2e2117636ae7) reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20251113/202511130834.0qWsUngh-lkp@…) If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags | Reported-by: kernel test robot <lkp(a)intel.com> | Closes: https://lore.kernel.org/oe-kbuild-all/202511130834.0qWsUngh-lkp@intel.com/ All warnings (new ones prefixed by >>): In file included from <built-in>:3: In file included from include/linux/compiler_types.h:150: include/linux/compiler-clang.h:33:9: warning: '__SANITIZE_ADDRESS__' macro redefined [-Wmacro-redefined] 33 | #define __SANITIZE_ADDRESS__ | ^ <built-in>:368:9: note: previous definition is here 368 | #define __SANITIZE_ADDRESS__ 1 | ^ >> mm/gmem_phys.c:245:19: warning: no previous prototype for function 'gm_evict_page_locked' [-Wmissing-prototypes] 245 | enum gm_evict_ret gm_evict_page_locked(struct gm_page *gm_page) | ^ mm/gmem_phys.c:245:1: note: declare 'static' if the function is not intended to be used outside of this translation unit 245 | enum gm_evict_ret gm_evict_page_locked(struct gm_page *gm_page) | ^ | static >> mm/gmem_phys.c:345:19: warning: no previous prototype for function 'gm_evict_page' [-Wmissing-prototypes] 345 | enum gm_evict_ret gm_evict_page(struct gm_page *gm_page) | ^ mm/gmem_phys.c:345:1: note: declare 'static' if the function is not intended to be used outside of this translation unit 345 | enum gm_evict_ret gm_evict_page(struct gm_page *gm_page) | ^ | static 3 warnings generated. vim +/gm_evict_page_locked +245 mm/gmem_phys.c 244 > 245 enum gm_evict_ret gm_evict_page_locked(struct gm_page *gm_page) 246 { 247 struct gm_dev *gm_dev; 248 struct gm_mapping *gm_mapping; 249 struct vm_area_struct *vma; 250 struct mm_struct *mm; 251 struct page *page; 252 struct device *dma_dev; 253 unsigned long va; 254 struct folio *folio = NULL; 255 struct gm_fault_t gmf = { 256 .size = HPAGE_SIZE, 257 .copy = true 258 }; 259 enum gm_evict_ret ret = GM_EVICT_SUCCESS; 260 enum gm_ret gm_ret; 261 262 gm_dev = get_gm_dev(gm_page->hnid); 263 if (!gm_dev) 264 return GM_EVICT_DEVERR; 265 266 spin_lock(&gm_page->rmap_lock); 267 if (!gm_page->mm) { 268 /* Evicting gm_page conflicts with unmap.*/ 269 ret = GM_EVICT_UNMAP; 270 goto rmap_unlock; 271 } 272 273 mm = gm_page->mm; 274 va = gm_page->va; 275 vma = find_vma(mm, va); 276 if (!vma || !vma->vm_obj) { 277 gmem_err("%s: cannot find vma or vma->vm_obj is null for va %lx", __func__, va); 278 ret = GM_EVICT_UNMAP; 279 goto rmap_unlock; 280 } 281 282 gm_mapping = vm_object_lookup(vma->vm_obj, va); 283 if (!gm_mapping) { 284 gmem_err("%s: no gm_mapping for va %lx", __func__, va); 285 ret = GM_EVICT_UNMAP; 286 goto rmap_unlock; 287 } 288 289 spin_unlock(&gm_page->rmap_lock); 290 291 mutex_lock(&gm_mapping->lock); 292 if (!gm_mapping_device(gm_mapping)) { 293 /* Evicting gm_page conflicts with unmap.*/ 294 ret = GM_EVICT_UNMAP; 295 goto gm_mapping_unlock; 296 } 297 298 if (gm_mapping->gm_page != gm_page) { 299 /* gm_mapping maps to another gm_page. */ 300 ret = GM_EVICT_UNMAP; 301 goto gm_mapping_unlock; 302 } 303 304 folio = vma_alloc_folio(GFP_TRANSHUGE, HPAGE_PMD_ORDER, vma, va, true); 305 if (!folio) { 306 gmem_err("%s: allocate host page failed.", __func__); 307 ret = GM_EVICT_FALLBACK; 308 goto gm_mapping_unlock; 309 } 310 page = &folio->page; 311 312 gmf.mm = mm; 313 gmf.va = va; 314 gmf.dev = gm_dev; 315 gmf.pfn = gm_page->dev_pfn; 316 dma_dev = gm_dev->dma_dev; 317 gmf.dma_addr = dma_map_page(dma_dev, page, 0, HPAGE_SIZE, DMA_BIDIRECTIONAL); 318 if (dma_mapping_error(dma_dev, gmf.dma_addr)) { 319 gmem_err("%s: dma map failed.", __func__); 320 ret = GM_EVICT_FALLBACK; 321 goto gm_mapping_unlock; 322 } 323 324 gm_ret = gm_dev->mmu->peer_unmap(&gmf); 325 if (gm_ret != GM_RET_SUCCESS) { 326 gmem_err("%s: peer_unmap failed.", __func__); 327 ret = GM_EVICT_DEVERR; 328 goto dma_unmap; 329 } 330 331 gm_mapping_flags_set(gm_mapping, GM_MAPPING_CPU); 332 gm_page_remove_rmap(gm_page); 333 gm_mapping->page = page; 334 put_gm_page(gm_page); 335 dma_unmap: 336 dma_unmap_page(dma_dev, gmf.dma_addr, HPAGE_SIZE, DMA_BIDIRECTIONAL); 337 gm_mapping_unlock: 338 mutex_unlock(&gm_mapping->lock); 339 return ret; 340 rmap_unlock: 341 spin_unlock(&gm_page->rmap_lock); 342 return ret; 343 } 344 > 345 enum gm_evict_ret gm_evict_page(struct gm_page *gm_page) 346 { 347 struct mm_struct *mm = gm_page->mm; 348 enum gm_evict_ret ret; 349 350 mmap_read_lock(mm); 351 ret = gm_evict_page_locked(gm_page); 352 mmap_read_unlock(mm); 353 return ret; 354 } 355 -- 0-DAY CI Kernel Test Service https://github.com/intel/lkp-tests/wiki

1 0

[PATCH OLK-5.10 0/4] Backport kunpeng920G to OLK-5.10
by Lin Yujun 13 Nov '25

13 Nov '25

Cong Wang (1): iommu: avoid taking iova_rbtree_lock twice Xiang Chen (3): iommu/arm-smmu-v3: Add a check to avoid invalid iotlb sync dma-mapping: benchmark: Add support for multi-pages map/unmap iommu/iova: Put free_iova_mem() outside of spinlock iova_rbtree_lock drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c | 3 +++ drivers/iommu/iova.c | 24 ++++++++++++------- kernel/dma/map_benchmark.c | 21 ++++++++++------ .../testing/selftests/dma/dma_map_benchmark.c | 20 ++++++++++++---- 4 files changed, 49 insertions(+), 19 deletions(-) -- 2.34.1

2 5

[PATCH OLK-6.6 0/3] wakeup kswapd during node_reclaim
by Wupeng Ma 13 Nov '25

13 Nov '25

wakeup kswapd during node_reclaim and add new bits for zone recalim to enable this kind of swap. Wupeng Ma (3): arm64: support node_reclaim_distance adjust for arch mm: vmscan: wakeup kswapd during node_reclaim mm: node_reclaim: add wakeup kswapd mode Documentation/admin-guide/sysctl/vm.rst | 1 + arch/arm64/mm/init.c | 20 ++++++++++++++++++++ include/linux/swap.h | 3 ++- include/uapi/linux/mempolicy.h | 1 + mm/internal.h | 5 +++-- mm/page_alloc.c | 3 ++- mm/vmscan.c | 7 ++++++- 7 files changed, 35 insertions(+), 5 deletions(-) -- 2.43.0

2 4

[PATCH OLK-6.6 00/14] mm: add huge pfnmap support for remap_pfn_range()
by Yin Tirui 12 Nov '25

12 Nov '25

Overview ======== This patch series adds huge page support for remap_pfn_range(), automatically creating huge mappings when prerequisites are satisfied (size, alignment, architecture support, etc.) and falling back to normal page mappings otherwise. This work builds on Peter Xu's previous efforts on huge pfnmap support [0]. TODO ==== - Add PUD-level huge page support. Currently, only PMD-level huge pages are supported. - Consider the logic related to vmap_page_range and extract reusable common code. Tests Done ========== - Cross-build tests. - Performance tests with custom device driver implementing mmap() with remap_pfn_range(): - lat_mem_rd benchmark modified to use mmap(device_fd) instead of malloc() shows around 40% improvement in memory access latency with huge page support compared to normal page mappings. numactl -C 0 lat_mem_rd -t 4096M (stride=64) Memory Size (MB) Without Huge Mapping With Huge Mapping Improvement ---------------- ----------------- -------------- ----------- 64.00 148.858 ns 100.780 ns 32.3% 128.00 164.745 ns 103.537 ns 37.2% 256.00 169.907 ns 103.179 ns 39.3% 512.00 171.285 ns 103.072 ns 39.8% 1024.00 173.054 ns 103.055 ns 40.4% 2048.00 172.820 ns 103.091 ns 40.3% 4096.00 172.877 ns 103.115 ns 40.4% - Custom memory copy operations on mmap(device_fd) show around 18% performance improvement with huge page support compared to normal page mappings. numactl -C 0 memcpy_test (memory copy performance test) Memory Size (MB) Without Huge Mapping With Huge Mapping Improvement ---------------- ----------------- -------------- ----------- 1024.00 95.76 ms 77.91 ms 18.6% 2048.00 190.87 ms 155.64 ms 18.5% 4096.00 380.84 ms 311.45 ms 18.2% [0] https://lore.kernel.org/all/20240826204353.2228736-2-peterx@redhat.com/T/#u David Hildenbrand (1): mm/huge_memory: check pmd_special() only after pmd_present() Peter Xu (11): mm: introduce ARCH_SUPPORTS_HUGE_PFNMAP and special bits to pmd/pud mm: drop is_huge_zero_pud() mm: mark special bits for huge pfn mappings when inject mm: allow THP orders for PFNMAPs mm/gup: detect huge pfnmap entries in gup-fast mm/pagewalk: check pfnmap for folio_walk_start() mm/fork: accept huge pfnmap entries mm: always define pxx_pgprot() mm/x86: support large pfn mappings mm/arm64: support large pfn mappings arm64: mm: Drop dead code for pud special bit handling Yin Tirui (2): pgtable: add pte_clrhuge() implementation for arm64 mm: introduce remap_pfn_range_try_pmd() for PMD-level hugepage mapping arch/arm64/Kconfig | 1 + arch/arm64/include/asm/pgtable.h | 39 ++++++ arch/powerpc/include/asm/book3s/64/pgtable.h | 2 +- arch/powerpc/include/asm/pgtable.h | 1 + arch/s390/include/asm/pgtable.h | 1 + arch/sparc/include/asm/pgtable_64.h | 1 + arch/x86/Kconfig | 1 + arch/x86/include/asm/pgtable.h | 81 +++++++----- fs/dax.c | 2 +- include/linux/huge_mm.h | 16 +-- include/linux/mm.h | 26 ++++ include/linux/pgtable.h | 20 ++- mm/Kconfig | 13 ++ mm/gup.c | 6 + mm/huge_memory.c | 93 ++++++++++---- mm/memory.c | 126 +++++++++++++++---- 16 files changed, 333 insertions(+), 96 deletions(-) -- 2.43.0

2 15

[PATCH OLK-6.6 v7 00/10] Add xcall 2.0 support
by Jinjie Ruan 12 Nov '25

12 Nov '25

Add xcall 2.0 support, which is compatible with xcall1.0, and can be used to dynamic instruction replacement for hardware xcall. Changes in v7: - Fix alternative bug in sync ventry. - Add "xcall" limit for xcall2.0 and hardware xcall. Changes in v6: - Add "xcall=debug" debug mode to use the "fast syscall" for every system call. - Use alternative to reduce the baseline performance noise floor when not set "xcall" on the kernel cmdline - Move binary_path out from xcall struct, and add to xcall_comm struct. - Fix1. insert_xcall_locked() check both "name" and "binary" to avoid repeated attach same name of xcall. - Fix2. Check if it is an executable file for the binary path. - Fix3. Check if it is an absolute path for the binary path. - Fix4. Remove the extra free_xcall_comm(info) in default branch in proc_xcall_command(). - Fix5. Add zero syscall hijack check for xcall_prog_register(). - Use __arm64_sys_ni_syscall() instead of define inv_xcall(). - Replace the -ENOSYS return value with -EINVAL, to avoid the warning. Changes in v5: - Fix build problem. - el0t_64_sync_ventry jump to kernel_ventry to do VMAP_STACK. - Some cleanups. - Merge the three refactor patch. Changes in v4: - Reuse uprobe_write_opcode() instead of use __replace_page() and copy_to_page(). - Handle error for uprobe_write_opcode(). - Support xcall1.0 before 920G. - Change xcall1.0 bitmap to byte array to save check insns. - Use percpu variable to save check insns for xcall1.0. - A little cleanup. Changes in v3: - Add validity checks during xcall registration to prevent out-of-bounds access and null pointer dereferences. - Extract __replace_page() and copy_to_page() from uprobe, to avoid code duplication. - Replace with new compatible refactored code. - Remove unnecessary member, such as old_name. Changes in v2: - Remove duplicates or unnecessary ~300 LOC, such as link_slot, init_task. - Fix kabi. - Some cleanups. Jinjie Ruan (2): xcall2.0: Support xcall1.0 for hardware xcall openeuler_defconfig: Enable CONFIG_DYNAMIC_XCALL by default Liao Chang (6): xcall2.0: Add userspace proc interface xcall2.0: Add xcall module register interface xcall2.0: Add xcall_area xcall2.0: Hajack syscall with dynamic instruciton replace xcall: Rework the early exception vector of XCALL and SYNC xcall2.0: Add a basic testcase Yipeng Zou (1): xcall: Add debug mode to force fast syscall for all threads Yuntao Liu (1): xcall2.0: Intruduce xcall2.0 prefetch kernel module arch/arm64/Kconfig.turbo | 13 + arch/arm64/configs/openeuler_defconfig | 1 + arch/arm64/include/asm/exception.h | 4 +- arch/arm64/include/asm/mmu_context.h | 7 - arch/arm64/include/asm/xcall.h | 154 ++++---- arch/arm64/kernel/cpufeature.c | 79 +++-- arch/arm64/kernel/entry-common.c | 28 +- arch/arm64/kernel/entry.S | 86 +---- arch/arm64/kernel/probes/uprobes.c | 6 + arch/arm64/kernel/process.c | 5 + arch/arm64/kernel/syscall.c | 14 + arch/arm64/kernel/xcall/Makefile | 3 +- arch/arm64/kernel/xcall/core.c | 390 +++++++++++++++++++++ arch/arm64/kernel/xcall/entry.S | 139 ++++++-- arch/arm64/kernel/xcall/proc.c | 226 ++++++++++++ arch/arm64/kernel/xcall/xcall.c | 66 ++-- arch/arm64/kvm/sys_regs.c | 1 + drivers/staging/Kconfig | 2 + drivers/staging/Makefile | 1 + drivers/staging/xcall/Kconfig | 19 + drivers/staging/xcall/Makefile | 1 + drivers/staging/xcall/dynamic_xcall_test.c | 90 +++++ drivers/staging/xcall/prefetch.c | 265 ++++++++++++++ fs/proc/proc_xcall.c | 144 ++------ include/linux/mm_types.h | 4 + include/linux/xcall.h | 51 +++ kernel/events/uprobes.c | 23 ++ kernel/fork.c | 3 + mm/mmap.c | 14 +- 29 files changed, 1465 insertions(+), 374 deletions(-) create mode 100644 arch/arm64/kernel/xcall/core.c create mode 100644 arch/arm64/kernel/xcall/proc.c create mode 100644 drivers/staging/xcall/Kconfig create mode 100644 drivers/staging/xcall/Makefile create mode 100644 drivers/staging/xcall/dynamic_xcall_test.c create mode 100644 drivers/staging/xcall/prefetch.c create mode 100644 include/linux/xcall.h -- 2.34.1

2 11

[PATCH OLK-6.6 v7 00/10] Add xcall 2.0 support
by Jinjie Ruan 12 Nov '25

12 Nov '25

Add xcall 2.0 support, which is compatible with xcall1.0, and can be used to dynamic instruction replacement for hardware xcall. Changes in v7: - Fix alternative bug in sync ventry. - Add "xcall" limit for xcall2.0. Changes in v6: - Add "xcall=debug" debug mode to use the "fast syscall" for every system call. - Use alternative to reduce the baseline performance noise floor when not set "xcall" on the kernel cmdline - Move binary_path out from xcall struct, and add to xcall_comm struct. - Fix1. insert_xcall_locked() check both "name" and "binary" to avoid repeated attach same name of xcall. - Fix2. Check if it is an executable file for the binary path. - Fix3. Check if it is an absolute path for the binary path. - Fix4. Remove the extra free_xcall_comm(info) in default branch in proc_xcall_command(). - Fix5. Add zero syscall hijack check for xcall_prog_register(). - Use __arm64_sys_ni_syscall() instead of define inv_xcall(). - Replace the -ENOSYS return value with -EINVAL, to avoid the warning. Changes in v5: - Fix build problem. - el0t_64_sync_ventry jump to kernel_ventry to do VMAP_STACK. - Some cleanups. - Merge the three refactor patch. Changes in v4: - Reuse uprobe_write_opcode() instead of use __replace_page() and copy_to_page(). - Handle error for uprobe_write_opcode(). - Support xcall1.0 before 920G. - Change xcall1.0 bitmap to byte array to save check insns. - Use percpu variable to save check insns for xcall1.0. - A little cleanup. Changes in v3: - Add validity checks during xcall registration to prevent out-of-bounds access and null pointer dereferences. - Extract __replace_page() and copy_to_page() from uprobe, to avoid code duplication. - Replace with new compatible refactored code. - Remove unnecessary member, such as old_name. Changes in v2: - Remove duplicates or unnecessary ~300 LOC, such as link_slot, init_task. - Fix kabi. - Some cleanups. Jinjie Ruan (2): xcall2.0: Support xcall1.0 for hardware xcall openeuler_defconfig: Enable CONFIG_DYNAMIC_XCALL by default Liao Chang (6): xcall2.0: Add userspace proc interface xcall2.0: Add xcall module register interface xcall2.0: Add xcall_area xcall2.0: Hajack syscall with dynamic instruciton replace xcall: Rework the early exception vector of XCALL and SYNC xcall2.0: Add a basic testcase Yipeng Zou (1): xcall: Add debug mode to force fast syscall for all threads Yuntao Liu (1): xcall2.0: Intruduce xcall2.0 prefetch kernel module arch/arm64/Kconfig.turbo | 13 + arch/arm64/configs/openeuler_defconfig | 1 + arch/arm64/include/asm/exception.h | 4 +- arch/arm64/include/asm/mmu_context.h | 7 - arch/arm64/include/asm/xcall.h | 154 ++++---- arch/arm64/kernel/cpufeature.c | 76 ++-- arch/arm64/kernel/entry-common.c | 28 +- arch/arm64/kernel/entry.S | 86 +---- arch/arm64/kernel/probes/uprobes.c | 6 + arch/arm64/kernel/process.c | 5 + arch/arm64/kernel/syscall.c | 14 + arch/arm64/kernel/xcall/Makefile | 3 +- arch/arm64/kernel/xcall/core.c | 390 +++++++++++++++++++++ arch/arm64/kernel/xcall/entry.S | 139 ++++++-- arch/arm64/kernel/xcall/proc.c | 226 ++++++++++++ arch/arm64/kernel/xcall/xcall.c | 66 ++-- arch/arm64/kvm/sys_regs.c | 1 + drivers/staging/Kconfig | 2 + drivers/staging/Makefile | 1 + drivers/staging/xcall/Kconfig | 19 + drivers/staging/xcall/Makefile | 1 + drivers/staging/xcall/dynamic_xcall_test.c | 90 +++++ drivers/staging/xcall/prefetch.c | 265 ++++++++++++++ fs/proc/proc_xcall.c | 144 ++------ include/linux/mm_types.h | 4 + include/linux/xcall.h | 51 +++ kernel/events/uprobes.c | 23 ++ kernel/fork.c | 3 + mm/mmap.c | 14 +- 29 files changed, 1462 insertions(+), 374 deletions(-) create mode 100644 arch/arm64/kernel/xcall/core.c create mode 100644 arch/arm64/kernel/xcall/proc.c create mode 100644 drivers/staging/xcall/Kconfig create mode 100644 drivers/staging/xcall/Makefile create mode 100644 drivers/staging/xcall/dynamic_xcall_test.c create mode 100644 drivers/staging/xcall/prefetch.c create mode 100644 include/linux/xcall.h -- 2.34.1

2 11

[PATCH OLK-6.6] mm/thp: Add a new sysfs knob for zero COW
by Zhang Qilong 12 Nov '25

12 Nov '25

hulk inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/ID65ME ---------------------------------------- Add zero_page_full_cow sysfs knob for zero hugepage, and the value is '0' by default. When it's enabled, kernel will try to allocate PMD page for PMD-mappable zero anonymous mapping on WP mode. Signed-off-by: Zhang Qilong <zhangqilong3(a)huawei.com> --- Documentation/admin-guide/mm/transhuge.rst | 7 +++++++ include/linux/huge_mm.h | 1 + mm/huge_memory.c | 21 +++++++++++++++++++++ 3 files changed, 29 insertions(+) diff --git a/Documentation/admin-guide/mm/transhuge.rst b/Documentation/admin-guide/mm/transhuge.rst index 70176fc3f94c..55f4ede1f8f8 100644 --- a/Documentation/admin-guide/mm/transhuge.rst +++ b/Documentation/admin-guide/mm/transhuge.rst @@ -201,10 +201,17 @@ page fault to anonymous mapping. It's possible to disable huge zero page by writing 0 or enable it back by writing 1:: echo 0 >/sys/kernel/mm/transparent_hugepage/use_zero_page echo 1 >/sys/kernel/mm/transparent_hugepage/use_zero_page +By default kernel disable zero page full CoW. When it's enabled, kernel +will try to allocate PMD page for PMD-mappable zero anonymous mapping on +WP mode. We can enable it by writing 1 or disable it back by writing 0:: + + echo 0 >/sys/kernel/mm/transparent_hugepage/zero_page_full_cow + echo 1 >/sys/kernel/mm/transparent_hugepage/zero_page_full_cow + Some userspace (such as a test program, or an optimized memory allocation library) may want to know the size (in bytes) of a PMD-mappable transparent hugepage:: cat /sys/kernel/mm/transparent_hugepage/hpage_pmd_size diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h index cfe42c43b55b..47a6634e74c4 100644 --- a/include/linux/huge_mm.h +++ b/include/linux/huge_mm.h @@ -55,10 +55,11 @@ enum transparent_hugepage_flag { TRANSPARENT_HUGEPAGE_FILE_EXEC_THP_FLAG, TRANSPARENT_HUGEPAGE_FILE_EXEC_MTHP_FLAG, TRANSPARENT_HUGEPAGE_FILE_MAPPING_ALIGN_FLAG, TRANSPARENT_HUGEPAGE_ANON_MAPPING_ALIGN_FLAG, TRANSPARENT_HUGEPAGE_ANON_MAPPING_PMD_ALIGN_FLAG, + TRANSPARENT_HUGEPAGE_ZERO_PAGE_FULL_COW_FLAG, }; struct kobject; struct kobj_attribute; diff --git a/mm/huge_memory.c b/mm/huge_memory.c index a28dda799978..11162cf63f44 100644 --- a/mm/huge_memory.c +++ b/mm/huge_memory.c @@ -601,10 +601,26 @@ static ssize_t thp_mapping_align_store(struct kobject *kobj, return count; } static struct kobj_attribute thp_mapping_align_attr = __ATTR_RW(thp_mapping_align); +static ssize_t zero_page_full_cow_show(struct kobject *kobj, + struct kobj_attribute *attr, char *buf) +{ + return single_hugepage_flag_show(kobj, attr, buf, + TRANSPARENT_HUGEPAGE_ZERO_PAGE_FULL_COW_FLAG); +} +static ssize_t zero_page_full_cow_store(struct kobject *kobj, + struct kobj_attribute *attr, const char *buf, size_t count) +{ + return single_hugepage_flag_store(kobj, attr, buf, count, + TRANSPARENT_HUGEPAGE_ZERO_PAGE_FULL_COW_FLAG); +} + +static struct kobj_attribute zero_page_full_cow_attr = + __ATTR_RW(zero_page_full_cow); + static struct attribute *hugepage_attr[] = { &enabled_attr.attr, &defrag_attr.attr, &use_zero_page_attr.attr, &pcp_allow_high_order_attr.attr, @@ -613,10 +629,11 @@ static struct attribute *hugepage_attr[] = { &shmem_enabled_attr.attr, #endif &file_enabled_attr.attr, &thp_exec_enabled_attr.attr, &thp_mapping_align_attr.attr, + &zero_page_full_cow_attr.attr, NULL, }; static const struct attribute_group hugepage_attr_group = { .attrs = hugepage_attr, @@ -2012,10 +2029,14 @@ static vm_fault_t do_huge_zero_wp_pmd(struct vm_fault *vmf) struct vm_area_struct *vma = vmf->vma; struct mmu_notifier_range range; struct folio *folio; vm_fault_t ret = 0; + if (!test_bit(TRANSPARENT_HUGEPAGE_ZERO_PAGE_FULL_COW_FLAG, + &transparent_hugepage_flags)) + return VM_FAULT_FALLBACK; + folio = vma_alloc_anon_folio_pmd(vma, vmf->address); if (unlikely(!folio)) return VM_FAULT_FALLBACK; mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma->vm_mm, haddr, -- 2.43.0

2 1

[PATCH OLK-6.6 00/14] mm: add huge pfnmap support for remap_pfn_range()
by Yin Tirui 12 Nov '25

12 Nov '25

Overview ======== This patch series adds huge page support for remap_pfn_range(), automatically creating huge mappings when prerequisites are satisfied (size, alignment, architecture support, etc.) and falling back to normal page mappings otherwise. This work builds on Peter Xu's previous efforts on huge pfnmap support [0]. TODO ==== - Add PUD-level huge page support. Currently, only PMD-level huge pages are supported. - Consider the logic related to vmap_page_range and extract reusable common code. Tests Done ========== - Cross-build tests. - Performance tests with custom device driver implementing mmap() with remap_pfn_range(): - lat_mem_rd benchmark modified to use mmap(device_fd) instead of malloc() shows around 40% improvement in memory access latency with huge page support compared to normal page mappings. numactl -C 0 lat_mem_rd -t 4096M (stride=64) Memory Size (MB) Without Huge Mapping With Huge Mapping Improvement ---------------- ----------------- -------------- ----------- 64.00 148.858 ns 100.780 ns 32.3% 128.00 164.745 ns 103.537 ns 37.2% 256.00 169.907 ns 103.179 ns 39.3% 512.00 171.285 ns 103.072 ns 39.8% 1024.00 173.054 ns 103.055 ns 40.4% 2048.00 172.820 ns 103.091 ns 40.3% 4096.00 172.877 ns 103.115 ns 40.4% - Custom memory copy operations on mmap(device_fd) show around 18% performance improvement with huge page support compared to normal page mappings. numactl -C 0 memcpy_test (memory copy performance test) Memory Size (MB) Without Huge Mapping With Huge Mapping Improvement ---------------- ----------------- -------------- ----------- 1024.00 95.76 ms 77.91 ms 18.6% 2048.00 190.87 ms 155.64 ms 18.5% 4096.00 380.84 ms 311.45 ms 18.2% [0] https://lore.kernel.org/all/20240826204353.2228736-2-peterx@redhat.com/T/#u David Hildenbrand (1): mm/huge_memory: check pmd_special() only after pmd_present() Peter Xu (11): mm: introduce ARCH_SUPPORTS_HUGE_PFNMAP and special bits to pmd/pud mm: drop is_huge_zero_pud() mm: mark special bits for huge pfn mappings when inject mm: allow THP orders for PFNMAPs mm/gup: detect huge pfnmap entries in gup-fast mm/pagewalk: check pfnmap for folio_walk_start() mm/fork: accept huge pfnmap entries mm: always define pxx_pgprot() mm/x86: support large pfn mappings mm/arm64: support large pfn mappings arm64: mm: Drop dead code for pud special bit handling Yin Tirui (2): pgtable: add pte_clrhuge() implementation for arm64 mm: introduce remap_pfn_range_try_pmd() for PMD-level hugepage mapping arch/arm64/Kconfig | 1 + arch/arm64/include/asm/pgtable.h | 39 ++++++ arch/powerpc/include/asm/book3s/64/pgtable.h | 2 +- arch/powerpc/include/asm/pgtable.h | 1 + arch/s390/include/asm/pgtable.h | 1 + arch/sparc/include/asm/pgtable_64.h | 1 + arch/x86/Kconfig | 1 + arch/x86/include/asm/pgtable.h | 81 +++++++----- fs/dax.c | 2 +- include/linux/huge_mm.h | 16 +-- include/linux/mm.h | 26 ++++ include/linux/pgtable.h | 20 ++- mm/Kconfig | 13 ++ mm/gup.c | 6 + mm/huge_memory.c | 93 ++++++++++---- mm/memory.c | 126 +++++++++++++++---- 16 files changed, 333 insertions(+), 96 deletions(-) -- 2.43.0

2 15

[openeuler:OLK-5.10 3282/3282] mm/memcontrol.c:5402:26: error: implicit declaration of function 'ksm_process_profit'
by kernel test robot 12 Nov '25

12 Nov '25

Hi Nanyong, FYI, the error/warning still remains. tree: https://gitee.com/openeuler/kernel.git OLK-5.10 head: 36c423f81858a53a2f8b9ac653a4ee48a6b4a3ae commit: 0f6fb3570be40b783d8c756d64297be1a8e3d2b3 [3282/3282] memcg: support ksm merge any mode per cgroup config: x86_64-buildonly-randconfig-004-20251112 (https://download.01.org/0day-ci/archive/20251112/202511121630.TEi1Z7Hc-lkp@…) compiler: clang version 20.1.8 (https://github.com/llvm/llvm-project 87f0227cb60147a26a1eeb4fb06e3b505e9c7261) reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20251112/202511121630.TEi1Z7Hc-lkp@…) If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags | Reported-by: kernel test robot <lkp(a)intel.com> | Closes: https://lore.kernel.org/oe-kbuild-all/202511121630.TEi1Z7Hc-lkp@intel.com/ All errors (new ones prefixed by >>): >> mm/memcontrol.c:5402:26: error: implicit declaration of function 'ksm_process_profit' [-Werror,-Wimplicit-function-declaration] 5402 | ksm_process_profits += ksm_process_profit(mm); | ^ 1 error generated. vim +/ksm_process_profit +5402 mm/memcontrol.c 5379 5380 static int memory_ksm_show(struct seq_file *m, void *v) 5381 { 5382 unsigned long ksm_merging_pages = 0; 5383 unsigned long ksm_rmap_items = 0; 5384 long ksm_process_profits = 0; 5385 unsigned int tasks = 0; 5386 struct task_struct *task; 5387 struct mm_struct *mm; 5388 struct css_task_iter it; 5389 struct mem_cgroup *memcg = mem_cgroup_from_seq(m); 5390 5391 css_task_iter_start(&memcg->css, CSS_TASK_ITER_PROCS, &it); 5392 while ((task = css_task_iter_next(&it))) { 5393 mm = get_task_mm(task); 5394 if (!mm) 5395 continue; 5396 5397 if (test_bit(MMF_VM_MERGE_ANY, &mm->flags)) 5398 tasks++; 5399 5400 ksm_rmap_items += mm->ksm_rmap_items; 5401 ksm_merging_pages += mm->ksm_merging_pages; > 5402 ksm_process_profits += ksm_process_profit(mm); 5403 mmput(mm); 5404 } 5405 css_task_iter_end(&it); 5406 5407 seq_printf(m, "merge any state: %d\n", READ_ONCE(memcg->ksm_merge_any)); 5408 seq_printf(m, "merge any tasks: %u\n", tasks); 5409 seq_printf(m, "ksm_rmap_items %lu\n", ksm_rmap_items); 5410 seq_printf(m, "ksm_merging_pages %lu\n", ksm_merging_pages); 5411 seq_printf(m, "ksm_process_profits %ld\n", ksm_process_profits); 5412 return 0; 5413 } 5414 -- 0-DAY CI Kernel Test Service https://github.com/intel/lkp-tests/wiki

1 0

[PATCH OLK-6.6] mm: add nr_free_highatomic in show_free_areas
by Jinjiang Tu 12 Nov '25

12 Nov '25

From: gaoxu <gaoxu2(a)honor.com> mainline inclusion from mainline-v6.16-rc1 commit 4f219913c1368a372b07832451cb86ea3a1df31c category: other bugzilla: https://gitee.com/openeuler/kernel/issues/ID65M0 Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?… -------------------------------- commit c928807f6f6b6("mm/page_alloc: keep track of free highatomic") adds a new variable nr_free_highatomic, which is useful for analyzing low mem issues. add nr_free_highatomic in show_free_areas. Signed-off-by: gao xu <gaoxu2(a)honor.com> Link: https://lkml.kernel.org/r/d92eeff74f7a4578a14ac777cfe3603a@honor.com Acked-by: David Hildenbrand <david(a)redhat.com> Reviewed-by: Barry Song <baohua(a)kernel.org> Acked-by: David Rientjes <rientjes(a)google.com> Reviewed-by: Anshuman Khandual <anshuman.khandual(a)arm.com> Cc: Mike Rapoport <rppt(a)kernel.org> Cc: Suren Baghdasaryan <surenb(a)google.com> Cc: Yu Zhao <yuzhao(a)google.com> Signed-off-by: Andrew Morton <akpm(a)linux-foundation.org> Signed-off-by: Jinjiang Tu <tujinjiang(a)huawei.com> --- mm/show_mem.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/mm/show_mem.c b/mm/show_mem.c index a5608c697d15..23cb66ac712c 100644 --- a/mm/show_mem.c +++ b/mm/show_mem.c @@ -314,6 +314,7 @@ static void show_free_areas(unsigned int filter, nodemask_t *nodemask, int max_z " low:%lukB" " high:%lukB" " reserved_highatomic:%luKB" + " free_highatomic:%luKB" " active_anon:%lukB" " inactive_anon:%lukB" " active_file:%lukB" @@ -335,6 +336,7 @@ static void show_free_areas(unsigned int filter, nodemask_t *nodemask, int max_z K(low_wmark_pages(zone)), K(high_wmark_pages(zone)), K(zone->nr_reserved_highatomic), + K(zone->nr_free_highatomic), K(zone_page_state(zone, NR_ZONE_ACTIVE_ANON)), K(zone_page_state(zone, NR_ZONE_INACTIVE_ANON)), K(zone_page_state(zone, NR_ZONE_ACTIVE_FILE)), -- 2.43.0

2 1