From mboxrd@z Thu Jan 1 00:00:00 1970 Return-path: Received: from tyo201.gate.nec.co.jp ([210.143.35.51]) by bombadil.infradead.org with esmtps (Exim 4.80.1 #2 (Red Hat Linux)) id 1XS08g-0003B5-9g for kexec@lists.infradead.org; Thu, 11 Sep 2014 08:54:04 +0000 From: Atsushi Kumagai Subject: RE: [PATCH v4] makedumpfile: Exclude unnecessary hugepages. Date: Thu, 11 Sep 2014 08:52:26 +0000 Message-ID: <0910DD04CBD6DE4193FCF86B9C00BE97254CF9@BPXM01GP.gisp.nec.co.jp> References: <0910DD04CBD6DE4193FCF86B9C00BE9724E000@BPXM01GP.gisp.nec.co.jp> <20140910073424.GD25867@dhcp-16-116.nay.redhat.com> In-Reply-To: <20140910073424.GD25867@dhcp-16-116.nay.redhat.com> Content-Language: ja-JP MIME-Version: 1.0 List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Sender: "kexec" Errors-To: kexec-bounces+dwmw2=infradead.org@lists.infradead.org To: "bhe@redhat.com" Cc: "kexec@lists.infradead.org" >Hi Atsushi, > >Since huge pages are included in user pages, I can't think of a way to >make test cases for huge page exclusion. Could you give some suggestions >on this or how did you test it? Before I posted this patch, I tested as below. This idea came from the fact that old makedumpfile can't exclude huge pages except the first page(PG_head). 1. Get the number of hugepages from /proc/meminfo 2. Calculate the number of PG_tail pages 3. Capture the dumpfile without filtering 4. Run makedumpfile and compare the report message between v1.5.6 and v1.5.7(rc) to get how many user pages become excludable with this patch. 5. The result of Step2's and Step4's must be same, confirm it. The way above is for THP but you can apply it also for hugetlbfs if you take care of the things that old makedumpfile can't exclude *any* hugetlbfs pages. I recommend to separate the two cases completely by enabling either THP or hugetlbfs explicitly since it's easier to confirm the results. Thanks Atsushi Kumagai > >Thanks >Baoquan > > >On 08/20/14 at 07:27am, Atsushi Kumagai wrote: >> There are 2 types of hugepages in the kernel, the both should be >> excluded as user pages. >> >> 1. Transparent huge pages (THP) >> All the pages are anonymous pages (at least for now), so we should >> just get how many pages are in the corresponding hugepage. >> It can be gotten from the page->lru.prev of the second page in the >> hugepage. >> >> 2. Hugetlbfs pages >> The pages aren't anonymous pages but kind of user pages, we should >> exclude also these pages in any way. >> Luckily, it's possible to detect these pages by looking the >> page->lru.next of the second page in the hugepage. This idea came >> from the kernel's PageHuge(). >> The number of pages can be gotten in the same way as THP. >> >> Changelog: >> v4: >> - Cleaned up according to Petr's and Baoquan's comments. >> v3: >> - Cleaned up according to Petr's comments. >> - Fix misdetection of hugetlb pages. >> v2: >> - Rebased to "Generic multi-page exclusion". >> >> Signed-off-by: Atsushi Kumagai >> --- >> makedumpfile.c | 86 ++++++++++++++++++++++++++++++++++++++++++++++++---------- >> makedumpfile.h | 7 +++++ >> 2 files changed, 78 insertions(+), 15 deletions(-) >> >> diff --git a/makedumpfile.c b/makedumpfile.c >> index 11cd473..b4b6eca 100644 >> --- a/makedumpfile.c >> +++ b/makedumpfile.c >> @@ -1180,6 +1180,7 @@ get_symbol_info(void) >> SYMBOL_INIT(vmemmap_list, "vmemmap_list"); >> SYMBOL_INIT(mmu_psize_defs, "mmu_psize_defs"); >> SYMBOL_INIT(mmu_vmemmap_psize, "mmu_vmemmap_psize"); >> + SYMBOL_INIT(free_huge_page, "free_huge_page"); >> >> SYMBOL_INIT(cpu_pgd, "cpu_pgd"); >> SYMBOL_INIT(demote_segment_4k, "demote_segment_4k"); >> @@ -1296,6 +1297,15 @@ get_structure_info(void) >> ENUM_NUMBER_INIT(PG_slab, "PG_slab"); >> ENUM_NUMBER_INIT(PG_hwpoison, "PG_hwpoison"); >> >> + ENUM_NUMBER_INIT(PG_head_mask, "PG_head_mask"); >> + if (NUMBER(PG_head_mask) == NOT_FOUND_NUMBER) { >> + ENUM_NUMBER_INIT(PG_head, "PG_head"); >> + if (NUMBER(PG_head) == NOT_FOUND_NUMBER) >> + ENUM_NUMBER_INIT(PG_head, "PG_compound"); >> + if (NUMBER(PG_head) != NOT_FOUND_NUMBER) >> + NUMBER(PG_head_mask) = 1UL << NUMBER(PG_head); >> + } >> + >> ENUM_TYPE_SIZE_INIT(pageflags, "pageflags"); >> >> TYPEDEF_SIZE_INIT(nodemask_t, "nodemask_t"); >> @@ -1530,6 +1540,9 @@ get_value_for_old_linux(void) >> NUMBER(PG_swapcache) = PG_swapcache_ORIGINAL; >> if (NUMBER(PG_slab) == NOT_FOUND_NUMBER) >> NUMBER(PG_slab) = PG_slab_ORIGINAL; >> + if (NUMBER(PG_head_mask) == NOT_FOUND_NUMBER) >> + NUMBER(PG_head_mask) = 1L << PG_compound_ORIGINAL; >> + >> /* >> * The values from here are for free page filtering based on >> * mem_map array. These are minimum effort to cover old >> @@ -1699,6 +1712,7 @@ write_vmcoreinfo_data(void) >> WRITE_SYMBOL("mmu_vmemmap_psize", mmu_vmemmap_psize); >> WRITE_SYMBOL("cpu_pgd", cpu_pgd); >> WRITE_SYMBOL("demote_segment_4k", demote_segment_4k); >> + WRITE_SYMBOL("free_huge_page", free_huge_page); >> >> /* >> * write the structure size of 1st kernel >> @@ -1788,6 +1802,7 @@ write_vmcoreinfo_data(void) >> >> WRITE_NUMBER("PG_lru", PG_lru); >> WRITE_NUMBER("PG_private", PG_private); >> + WRITE_NUMBER("PG_head_mask", PG_head_mask); >> WRITE_NUMBER("PG_swapcache", PG_swapcache); >> WRITE_NUMBER("PG_buddy", PG_buddy); >> WRITE_NUMBER("PG_slab", PG_slab); >> @@ -2040,6 +2055,7 @@ read_vmcoreinfo(void) >> READ_SYMBOL("mmu_vmemmap_psize", mmu_vmemmap_psize); >> READ_SYMBOL("cpu_pgd", cpu_pgd); >> READ_SYMBOL("demote_segment_4k", demote_segment_4k); >> + READ_SYMBOL("free_huge_page", free_huge_page); >> >> READ_STRUCTURE_SIZE("page", page); >> READ_STRUCTURE_SIZE("mem_section", mem_section); >> @@ -2116,6 +2132,7 @@ read_vmcoreinfo(void) >> >> READ_NUMBER("PG_lru", PG_lru); >> READ_NUMBER("PG_private", PG_private); >> + READ_NUMBER("PG_head_mask", PG_head_mask); >> READ_NUMBER("PG_swapcache", PG_swapcache); >> READ_NUMBER("PG_slab", PG_slab); >> READ_NUMBER("PG_buddy", PG_buddy); >> @@ -4643,13 +4660,16 @@ __exclude_unnecessary_pages(unsigned long mem_map, >> mdf_pfn_t pfn_start, mdf_pfn_t pfn_end, struct cycle *cycle) >> { >> mdf_pfn_t pfn; >> + mdf_pfn_t *pfn_counter; >> + mdf_pfn_t nr_pages; >> unsigned long index_pg, pfn_mm; >> unsigned long long maddr; >> mdf_pfn_t pfn_read_start, pfn_read_end; >> unsigned char page_cache[SIZE(page) * PGMM_CACHED]; >> unsigned char *pcache; >> - unsigned int _count, _mapcount = 0; >> + unsigned int _count, _mapcount = 0, compound_order = 0; >> unsigned long flags, mapping, private = 0; >> + unsigned long compound_dtor; >> >> /* >> * If a multi-page exclusion is pending, do it first >> @@ -4715,11 +4735,36 @@ __exclude_unnecessary_pages(unsigned long mem_map, >> flags = ULONG(pcache + OFFSET(page.flags)); >> _count = UINT(pcache + OFFSET(page._count)); >> mapping = ULONG(pcache + OFFSET(page.mapping)); >> + >> + if ((index_pg < PGMM_CACHED - 1) && >> + isCompoundHead(flags)) { >> + compound_order = ULONG(pcache + SIZE(page) + OFFSET(page.lru) >> + + OFFSET(list_head.prev)); >> + compound_dtor = ULONG(pcache + SIZE(page) + OFFSET(page.lru) >> + + OFFSET(list_head.next)); >> + >> + if ((compound_order >= sizeof(unsigned long) * 8) >> + || ((pfn & ((1UL << compound_order) - 1)) != 0)) { >> + /* Invalid order */ >> + compound_order = 0; >> + } >> + } else { >> + /* >> + * The last pfn of the mem_map cache must not be compound page >> + * since all compound pages are aligned to its page order and >> + * PGMM_CACHED is a power of 2. >> + */ >> + compound_order = 0; >> + compound_dtor = 0; >> + } >> + >> if (OFFSET(page._mapcount) != NOT_FOUND_STRUCTURE) >> _mapcount = UINT(pcache + OFFSET(page._mapcount)); >> if (OFFSET(page.private) != NOT_FOUND_STRUCTURE) >> private = ULONG(pcache + OFFSET(page.private)); >> >> + nr_pages = 1 << compound_order; >> + pfn_counter = NULL; >> /* >> * Exclude the free page managed by a buddy >> * Use buddy identification of free pages whether cyclic or not. >> @@ -4727,12 +4772,8 @@ __exclude_unnecessary_pages(unsigned long mem_map, >> if ((info->dump_level & DL_EXCLUDE_FREE) >> && info->page_is_buddy >> && info->page_is_buddy(flags, _mapcount, private, _count)) { >> - int nr_pages = 1 << private; >> - >> - exclude_range(&pfn_free, pfn, pfn + nr_pages, cycle); >> - >> - pfn += nr_pages - 1; >> - mem_map += (nr_pages - 1) * SIZE(page); >> + nr_pages = 1 << private; >> + pfn_counter = &pfn_free; >> } >> /* >> * Exclude the cache page without the private page. >> @@ -4740,8 +4781,7 @@ __exclude_unnecessary_pages(unsigned long mem_map, >> else if ((info->dump_level & DL_EXCLUDE_CACHE) >> && (isLRU(flags) || isSwapCache(flags)) >> && !isPrivate(flags) && !isAnon(mapping)) { >> - if (clear_bit_on_2nd_bitmap_for_kernel(pfn, cycle)) >> - pfn_cache++; >> + pfn_counter = &pfn_cache; >> } >> /* >> * Exclude the cache page with the private page. >> @@ -4749,23 +4789,39 @@ __exclude_unnecessary_pages(unsigned long mem_map, >> else if ((info->dump_level & DL_EXCLUDE_CACHE_PRI) >> && (isLRU(flags) || isSwapCache(flags)) >> && !isAnon(mapping)) { >> - if (clear_bit_on_2nd_bitmap_for_kernel(pfn, cycle)) >> - pfn_cache_private++; >> + pfn_counter = &pfn_cache_private; >> } >> /* >> * Exclude the data page of the user process. >> + * - anonymous pages >> + * - hugetlbfs pages >> */ >> else if ((info->dump_level & DL_EXCLUDE_USER_DATA) >> - && isAnon(mapping)) { >> - if (clear_bit_on_2nd_bitmap_for_kernel(pfn, cycle)) >> - pfn_user++; >> + && (isAnon(mapping) || isHugetlb(compound_dtor))) { >> + pfn_counter = &pfn_user; >> } >> /* >> * Exclude the hwpoison page. >> */ >> else if (isHWPOISON(flags)) { >> + pfn_counter = &pfn_hwpoison; >> + } >> + /* >> + * Unexcludable page >> + */ >> + else >> + continue; >> + >> + /* >> + * Execute exclusion >> + */ >> + if (nr_pages == 1) { >> if (clear_bit_on_2nd_bitmap_for_kernel(pfn, cycle)) >> - pfn_hwpoison++; >> + (*pfn_counter)++; >> + } else { >> + exclude_range(pfn_counter, pfn, pfn + nr_pages, cycle); >> + pfn += nr_pages - 1; >> + mem_map += (nr_pages - 1) * SIZE(page); >> } >> } >> return TRUE; >> diff --git a/makedumpfile.h b/makedumpfile.h >> index eba9798..9f90b53 100644 >> --- a/makedumpfile.h >> +++ b/makedumpfile.h >> @@ -74,6 +74,7 @@ int get_mem_type(void); >> #define PG_lru_ORIGINAL (5) >> #define PG_slab_ORIGINAL (7) >> #define PG_private_ORIGINAL (11) /* Has something at ->private */ >> +#define PG_compound_ORIGINAL (14) /* Is part of a compound page */ >> #define PG_swapcache_ORIGINAL (15) /* Swap page: swp_entry_t in private */ >> >> #define PAGE_BUDDY_MAPCOUNT_VALUE_v2_6_38 (-2) >> @@ -148,6 +149,9 @@ test_bit(int nr, unsigned long addr) >> >> #define isLRU(flags) test_bit(NUMBER(PG_lru), flags) >> #define isPrivate(flags) test_bit(NUMBER(PG_private), flags) >> +#define isCompoundHead(flags) (!!((flags) & NUMBER(PG_head_mask))) >> +#define isHugetlb(dtor) ((SYMBOL(free_huge_page) != NOT_FOUND_SYMBOL) \ >> + && (SYMBOL(free_huge_page) == dtor)) >> #define isSwapCache(flags) test_bit(NUMBER(PG_swapcache), flags) >> #define isHWPOISON(flags) (test_bit(NUMBER(PG_hwpoison), flags) \ >> && (NUMBER(PG_hwpoison) != NOT_FOUND_NUMBER)) >> @@ -1218,6 +1222,7 @@ struct symbol_table { >> unsigned long long node_remap_start_vaddr; >> unsigned long long node_remap_end_vaddr; >> unsigned long long node_remap_start_pfn; >> + unsigned long long free_huge_page; >> >> /* >> * for Xen extraction >> @@ -1509,6 +1514,8 @@ struct number_table { >> */ >> long PG_lru; >> long PG_private; >> + long PG_head; >> + long PG_head_mask; >> long PG_swapcache; >> long PG_buddy; >> long PG_slab; >> -- >> 1.9.0 >> >> _______________________________________________ >> kexec mailing list >> kexec@lists.infradead.org >> http://lists.infradead.org/mailman/listinfo/kexec _______________________________________________ kexec mailing list kexec@lists.infradead.org http://lists.infradead.org/mailman/listinfo/kexec