diff options
Diffstat (limited to 'arch/s390/mm')
-rw-r--r-- | arch/s390/mm/cmm.c | 2 | ||||
-rw-r--r-- | arch/s390/mm/fault.c | 36 | ||||
-rw-r--r-- | arch/s390/mm/hugetlbpage.c | 1 | ||||
-rw-r--r-- | arch/s390/mm/init.c | 3 | ||||
-rw-r--r-- | arch/s390/mm/pgalloc.c | 244 |
5 files changed, 199 insertions, 87 deletions
diff --git a/arch/s390/mm/cmm.c b/arch/s390/mm/cmm.c index 2203164b39da..9141ed4c52e9 100644 --- a/arch/s390/mm/cmm.c +++ b/arch/s390/mm/cmm.c @@ -90,7 +90,7 @@ static long cmm_alloc_pages(long nr, long *counter, } else free_page((unsigned long) npa); } - diag10_range(addr >> PAGE_SHIFT, 1); + diag10_range(virt_to_pfn(addr), 1); pa->pages[pa->index++] = addr; (*counter)++; spin_unlock(&cmm_lock); diff --git a/arch/s390/mm/fault.c b/arch/s390/mm/fault.c index d30f5986fa85..ff16ce0d04ee 100644 --- a/arch/s390/mm/fault.c +++ b/arch/s390/mm/fault.c @@ -115,7 +115,7 @@ static void dump_pagetable(unsigned long asce, unsigned long address) pr_cont("R1:%016lx ", *table); if (*table & _REGION_ENTRY_INVALID) goto out; - table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN); + table = __va(*table & _REGION_ENTRY_ORIGIN); fallthrough; case _ASCE_TYPE_REGION2: table += (address & _REGION2_INDEX) >> _REGION2_SHIFT; @@ -124,7 +124,7 @@ static void dump_pagetable(unsigned long asce, unsigned long address) pr_cont("R2:%016lx ", *table); if (*table & _REGION_ENTRY_INVALID) goto out; - table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN); + table = __va(*table & _REGION_ENTRY_ORIGIN); fallthrough; case _ASCE_TYPE_REGION3: table += (address & _REGION3_INDEX) >> _REGION3_SHIFT; @@ -133,7 +133,7 @@ static void dump_pagetable(unsigned long asce, unsigned long address) pr_cont("R3:%016lx ", *table); if (*table & (_REGION_ENTRY_INVALID | _REGION3_ENTRY_LARGE)) goto out; - table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN); + table = __va(*table & _REGION_ENTRY_ORIGIN); fallthrough; case _ASCE_TYPE_SEGMENT: table += (address & _SEGMENT_INDEX) >> _SEGMENT_SHIFT; @@ -142,7 +142,7 @@ static void dump_pagetable(unsigned long asce, unsigned long address) pr_cont("S:%016lx ", *table); if (*table & (_SEGMENT_ENTRY_INVALID | _SEGMENT_ENTRY_LARGE)) goto out; - table = (unsigned long *)(*table & _SEGMENT_ENTRY_ORIGIN); + table = __va(*table & _SEGMENT_ENTRY_ORIGIN); } table += (address & _PAGE_INDEX) >> _PAGE_SHIFT; if (bad_address(table)) @@ -452,21 +452,21 @@ retry: if (unlikely(fault & VM_FAULT_ERROR)) goto out_up; - if (flags & FAULT_FLAG_ALLOW_RETRY) { - if (fault & VM_FAULT_RETRY) { - if (IS_ENABLED(CONFIG_PGSTE) && gmap && - (flags & FAULT_FLAG_RETRY_NOWAIT)) { - /* FAULT_FLAG_RETRY_NOWAIT has been set, - * mmap_lock has not been released */ - current->thread.gmap_pfault = 1; - fault = VM_FAULT_PFAULT; - goto out_up; - } - flags &= ~FAULT_FLAG_RETRY_NOWAIT; - flags |= FAULT_FLAG_TRIED; - mmap_read_lock(mm); - goto retry; + if (fault & VM_FAULT_RETRY) { + if (IS_ENABLED(CONFIG_PGSTE) && gmap && + (flags & FAULT_FLAG_RETRY_NOWAIT)) { + /* + * FAULT_FLAG_RETRY_NOWAIT has been set, mmap_lock has + * not been released + */ + current->thread.gmap_pfault = 1; + fault = VM_FAULT_PFAULT; + goto out_up; } + flags &= ~FAULT_FLAG_RETRY_NOWAIT; + flags |= FAULT_FLAG_TRIED; + mmap_read_lock(mm); + goto retry; } if (IS_ENABLED(CONFIG_PGSTE) && gmap) { address = __gmap_link(gmap, current->thread.gmap_addr, diff --git a/arch/s390/mm/hugetlbpage.c b/arch/s390/mm/hugetlbpage.c index da36d13ffc16..082793d497ec 100644 --- a/arch/s390/mm/hugetlbpage.c +++ b/arch/s390/mm/hugetlbpage.c @@ -9,6 +9,7 @@ #define KMSG_COMPONENT "hugetlb" #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt +#include <asm/pgalloc.h> #include <linux/mm.h> #include <linux/hugetlb.h> #include <linux/mman.h> diff --git a/arch/s390/mm/init.c b/arch/s390/mm/init.c index 8c6f258a6183..86ffd0d51fd5 100644 --- a/arch/s390/mm/init.c +++ b/arch/s390/mm/init.c @@ -215,6 +215,9 @@ void free_initmem(void) __set_memory((unsigned long)_sinittext, (unsigned long)(_einittext - _sinittext) >> PAGE_SHIFT, SET_MEMORY_RW | SET_MEMORY_NX); + free_reserved_area(sclp_early_sccb, + sclp_early_sccb + EXT_SCCB_READ_SCP, + POISON_FREE_INITMEM, "unused early sccb"); free_initmem_default(POISON_FREE_INITMEM); } diff --git a/arch/s390/mm/pgalloc.c b/arch/s390/mm/pgalloc.c index 781965f7210e..fd35c1a0213b 100644 --- a/arch/s390/mm/pgalloc.c +++ b/arch/s390/mm/pgalloc.c @@ -176,7 +176,75 @@ void page_table_free_pgste(struct page *page) #endif /* CONFIG_PGSTE */ /* - * page table entry allocation/free routines. + * A 2KB-pgtable is either upper or lower half of a normal page. + * The second half of the page may be unused or used as another + * 2KB-pgtable. + * + * Whenever possible the parent page for a new 2KB-pgtable is picked + * from the list of partially allocated pages mm_context_t::pgtable_list. + * In case the list is empty a new parent page is allocated and added to + * the list. + * + * When a parent page gets fully allocated it contains 2KB-pgtables in both + * upper and lower halves and is removed from mm_context_t::pgtable_list. + * + * When 2KB-pgtable is freed from to fully allocated parent page that + * page turns partially allocated and added to mm_context_t::pgtable_list. + * + * If 2KB-pgtable is freed from the partially allocated parent page that + * page turns unused and gets removed from mm_context_t::pgtable_list. + * Furthermore, the unused parent page is released. + * + * As follows from the above, no unallocated or fully allocated parent + * pages are contained in mm_context_t::pgtable_list. + * + * The upper byte (bits 24-31) of the parent page _refcount is used + * for tracking contained 2KB-pgtables and has the following format: + * + * PP AA + * 01234567 upper byte (bits 24-31) of struct page::_refcount + * || || + * || |+--- upper 2KB-pgtable is allocated + * || +---- lower 2KB-pgtable is allocated + * |+------- upper 2KB-pgtable is pending for removal + * +-------- lower 2KB-pgtable is pending for removal + * + * (See commit 620b4e903179 ("s390: use _refcount for pgtables") on why + * using _refcount is possible). + * + * When 2KB-pgtable is allocated the corresponding AA bit is set to 1. + * The parent page is either: + * - added to mm_context_t::pgtable_list in case the second half of the + * parent page is still unallocated; + * - removed from mm_context_t::pgtable_list in case both hales of the + * parent page are allocated; + * These operations are protected with mm_context_t::lock. + * + * When 2KB-pgtable is deallocated the corresponding AA bit is set to 0 + * and the corresponding PP bit is set to 1 in a single atomic operation. + * Thus, PP and AA bits corresponding to the same 2KB-pgtable are mutually + * exclusive and may never be both set to 1! + * The parent page is either: + * - added to mm_context_t::pgtable_list in case the second half of the + * parent page is still allocated; + * - removed from mm_context_t::pgtable_list in case the second half of + * the parent page is unallocated; + * These operations are protected with mm_context_t::lock. + * + * It is important to understand that mm_context_t::lock only protects + * mm_context_t::pgtable_list and AA bits, but not the parent page itself + * and PP bits. + * + * Releasing the parent page happens whenever the PP bit turns from 1 to 0, + * while both AA bits and the second PP bit are already unset. Then the + * parent page does not contain any 2KB-pgtable fragment anymore, and it has + * also been removed from mm_context_t::pgtable_list. It is safe to release + * the page therefore. + * + * PGSTE memory spaces use full 4KB-pgtables and do not need most of the + * logic described above. Both AA bits are set to 1 to denote a 4KB-pgtable + * while the PP bits are never used, nor such a page is added to or removed + * from mm_context_t::pgtable_list. */ unsigned long *page_table_alloc(struct mm_struct *mm) { @@ -192,14 +260,23 @@ unsigned long *page_table_alloc(struct mm_struct *mm) page = list_first_entry(&mm->context.pgtable_list, struct page, lru); mask = atomic_read(&page->_refcount) >> 24; - mask = (mask | (mask >> 4)) & 3; - if (mask != 3) { + /* + * The pending removal bits must also be checked. + * Failure to do so might lead to an impossible + * value of (i.e 0x13 or 0x23) written to _refcount. + * Such values violate the assumption that pending and + * allocation bits are mutually exclusive, and the rest + * of the code unrails as result. That could lead to + * a whole bunch of races and corruptions. + */ + mask = (mask | (mask >> 4)) & 0x03U; + if (mask != 0x03U) { table = (unsigned long *) page_to_virt(page); bit = mask & 1; /* =1 -> second 2K */ if (bit) table += PTRS_PER_PTE; atomic_xor_bits(&page->_refcount, - 1U << (bit + 24)); + 0x01U << (bit + 24)); list_del(&page->lru); } } @@ -220,12 +297,12 @@ unsigned long *page_table_alloc(struct mm_struct *mm) table = (unsigned long *) page_to_virt(page); if (mm_alloc_pgste(mm)) { /* Return 4K page table with PGSTEs */ - atomic_xor_bits(&page->_refcount, 3 << 24); + atomic_xor_bits(&page->_refcount, 0x03U << 24); memset64((u64 *)table, _PAGE_INVALID, PTRS_PER_PTE); memset64((u64 *)table + PTRS_PER_PTE, 0, PTRS_PER_PTE); } else { /* Return the first 2K fragment of the page */ - atomic_xor_bits(&page->_refcount, 1 << 24); + atomic_xor_bits(&page->_refcount, 0x01U << 24); memset64((u64 *)table, _PAGE_INVALID, 2 * PTRS_PER_PTE); spin_lock_bh(&mm->context.lock); list_add(&page->lru, &mm->context.pgtable_list); @@ -234,29 +311,53 @@ unsigned long *page_table_alloc(struct mm_struct *mm) return table; } +static void page_table_release_check(struct page *page, void *table, + unsigned int half, unsigned int mask) +{ + char msg[128]; + + if (!IS_ENABLED(CONFIG_DEBUG_VM) || !mask) + return; + snprintf(msg, sizeof(msg), + "Invalid pgtable %p release half 0x%02x mask 0x%02x", + table, half, mask); + dump_page(page, msg); +} + void page_table_free(struct mm_struct *mm, unsigned long *table) { + unsigned int mask, bit, half; struct page *page; - unsigned int bit, mask; page = virt_to_page(table); if (!mm_alloc_pgste(mm)) { /* Free 2K page table fragment of a 4K page */ bit = ((unsigned long) table & ~PAGE_MASK)/(PTRS_PER_PTE*sizeof(pte_t)); spin_lock_bh(&mm->context.lock); - mask = atomic_xor_bits(&page->_refcount, 1U << (bit + 24)); + /* + * Mark the page for delayed release. The actual release + * will happen outside of the critical section from this + * function or from __tlb_remove_table() + */ + mask = atomic_xor_bits(&page->_refcount, 0x11U << (bit + 24)); mask >>= 24; - if (mask & 3) + if (mask & 0x03U) list_add(&page->lru, &mm->context.pgtable_list); else list_del(&page->lru); spin_unlock_bh(&mm->context.lock); - if (mask != 0) + mask = atomic_xor_bits(&page->_refcount, 0x10U << (bit + 24)); + mask >>= 24; + if (mask != 0x00U) return; + half = 0x01U << bit; } else { - atomic_xor_bits(&page->_refcount, 3U << 24); + half = 0x03U; + mask = atomic_xor_bits(&page->_refcount, 0x03U << 24); + mask >>= 24; } + page_table_release_check(page, table, half, mask); pgtable_pte_page_dtor(page); __free_page(page); } @@ -272,47 +373,54 @@ void page_table_free_rcu(struct mmu_gather *tlb, unsigned long *table, page = virt_to_page(table); if (mm_alloc_pgste(mm)) { gmap_unlink(mm, table, vmaddr); - table = (unsigned long *) ((unsigned long)table | 3); + table = (unsigned long *) ((unsigned long)table | 0x03U); tlb_remove_table(tlb, table); return; } bit = ((unsigned long) table & ~PAGE_MASK) / (PTRS_PER_PTE*sizeof(pte_t)); spin_lock_bh(&mm->context.lock); + /* + * Mark the page for delayed release. The actual release will happen + * outside of the critical section from __tlb_remove_table() or from + * page_table_free() + */ mask = atomic_xor_bits(&page->_refcount, 0x11U << (bit + 24)); mask >>= 24; - if (mask & 3) + if (mask & 0x03U) list_add_tail(&page->lru, &mm->context.pgtable_list); else list_del(&page->lru); spin_unlock_bh(&mm->context.lock); - table = (unsigned long *) ((unsigned long) table | (1U << bit)); + table = (unsigned long *) ((unsigned long) table | (0x01U << bit)); tlb_remove_table(tlb, table); } void __tlb_remove_table(void *_table) { - unsigned int mask = (unsigned long) _table & 3; + unsigned int mask = (unsigned long) _table & 0x03U, half = mask; void *table = (void *)((unsigned long) _table ^ mask); struct page *page = virt_to_page(table); - switch (mask) { - case 0: /* pmd, pud, or p4d */ + switch (half) { + case 0x00U: /* pmd, pud, or p4d */ free_pages((unsigned long) table, 2); - break; - case 1: /* lower 2K of a 4K page table */ - case 2: /* higher 2K of a 4K page table */ + return; + case 0x01U: /* lower 2K of a 4K page table */ + case 0x02U: /* higher 2K of a 4K page table */ mask = atomic_xor_bits(&page->_refcount, mask << (4 + 24)); mask >>= 24; - if (mask != 0) - break; - fallthrough; - case 3: /* 4K page table with pgstes */ - if (mask & 3) - atomic_xor_bits(&page->_refcount, 3 << 24); - pgtable_pte_page_dtor(page); - __free_page(page); + if (mask != 0x00U) + return; + break; + case 0x03U: /* 4K page table with pgstes */ + mask = atomic_xor_bits(&page->_refcount, 0x03U << 24); + mask >>= 24; break; } + + page_table_release_check(page, table, half, mask); + pgtable_pte_page_dtor(page); + __free_page(page); } /* @@ -322,34 +430,34 @@ void __tlb_remove_table(void *_table) static struct kmem_cache *base_pgt_cache; -static unsigned long base_pgt_alloc(void) +static unsigned long *base_pgt_alloc(void) { - u64 *table; + unsigned long *table; table = kmem_cache_alloc(base_pgt_cache, GFP_KERNEL); if (table) - memset64(table, _PAGE_INVALID, PTRS_PER_PTE); - return (unsigned long) table; + memset64((u64 *)table, _PAGE_INVALID, PTRS_PER_PTE); + return table; } -static void base_pgt_free(unsigned long table) +static void base_pgt_free(unsigned long *table) { - kmem_cache_free(base_pgt_cache, (void *) table); + kmem_cache_free(base_pgt_cache, table); } -static unsigned long base_crst_alloc(unsigned long val) +static unsigned long *base_crst_alloc(unsigned long val) { - unsigned long table; + unsigned long *table; - table = __get_free_pages(GFP_KERNEL, CRST_ALLOC_ORDER); + table = (unsigned long *)__get_free_pages(GFP_KERNEL, CRST_ALLOC_ORDER); if (table) - crst_table_init((unsigned long *)table, val); + crst_table_init(table, val); return table; } -static void base_crst_free(unsigned long table) +static void base_crst_free(unsigned long *table) { - free_pages(table, CRST_ALLOC_ORDER); + free_pages((unsigned long)table, CRST_ALLOC_ORDER); } #define BASE_ADDR_END_FUNC(NAME, SIZE) \ @@ -377,14 +485,14 @@ static inline unsigned long base_lra(unsigned long address) return real; } -static int base_page_walk(unsigned long origin, unsigned long addr, +static int base_page_walk(unsigned long *origin, unsigned long addr, unsigned long end, int alloc) { unsigned long *pte, next; if (!alloc) return 0; - pte = (unsigned long *) origin; + pte = origin; pte += (addr & _PAGE_INDEX) >> _PAGE_SHIFT; do { next = base_page_addr_end(addr, end); @@ -393,13 +501,13 @@ static int base_page_walk(unsigned long origin, unsigned long addr, return 0; } -static int base_segment_walk(unsigned long origin, unsigned long addr, +static int base_segment_walk(unsigned long *origin, unsigned long addr, unsigned long end, int alloc) { - unsigned long *ste, next, table; + unsigned long *ste, next, *table; int rc; - ste = (unsigned long *) origin; + ste = origin; ste += (addr & _SEGMENT_INDEX) >> _SEGMENT_SHIFT; do { next = base_segment_addr_end(addr, end); @@ -409,9 +517,9 @@ static int base_segment_walk(unsigned long origin, unsigned long addr, table = base_pgt_alloc(); if (!table) return -ENOMEM; - *ste = table | _SEGMENT_ENTRY; + *ste = __pa(table) | _SEGMENT_ENTRY; } - table = *ste & _SEGMENT_ENTRY_ORIGIN; + table = __va(*ste & _SEGMENT_ENTRY_ORIGIN); rc = base_page_walk(table, addr, next, alloc); if (rc) return rc; @@ -422,13 +530,13 @@ static int base_segment_walk(unsigned long origin, unsigned long addr, return 0; } -static int base_region3_walk(unsigned long origin, unsigned long addr, +static int base_region3_walk(unsigned long *origin, unsigned long addr, unsigned long end, int alloc) { - unsigned long *rtte, next, table; + unsigned long *rtte, next, *table; int rc; - rtte = (unsigned long *) origin; + rtte = origin; rtte += (addr & _REGION3_INDEX) >> _REGION3_SHIFT; do { next = base_region3_addr_end(addr, end); @@ -438,9 +546,9 @@ static int base_region3_walk(unsigned long origin, unsigned long addr, table = base_crst_alloc(_SEGMENT_ENTRY_EMPTY); if (!table) return -ENOMEM; - *rtte = table | _REGION3_ENTRY; + *rtte = __pa(table) | _REGION3_ENTRY; } - table = *rtte & _REGION_ENTRY_ORIGIN; + table = __va(*rtte & _REGION_ENTRY_ORIGIN); rc = base_segment_walk(table, addr, next, alloc); if (rc) return rc; @@ -450,13 +558,13 @@ static int base_region3_walk(unsigned long origin, unsigned long addr, return 0; } -static int base_region2_walk(unsigned long origin, unsigned long addr, +static int base_region2_walk(unsigned long *origin, unsigned long addr, unsigned long end, int alloc) { - unsigned long *rste, next, table; + unsigned long *rste, next, *table; int rc; - rste = (unsigned long *) origin; + rste = origin; rste += (addr & _REGION2_INDEX) >> _REGION2_SHIFT; do { next = base_region2_addr_end(addr, end); @@ -466,9 +574,9 @@ static int base_region2_walk(unsigned long origin, unsigned long addr, table = base_crst_alloc(_REGION3_ENTRY_EMPTY); if (!table) return -ENOMEM; - *rste = table | _REGION2_ENTRY; + *rste = __pa(table) | _REGION2_ENTRY; } - table = *rste & _REGION_ENTRY_ORIGIN; + table = __va(*rste & _REGION_ENTRY_ORIGIN); rc = base_region3_walk(table, addr, next, alloc); if (rc) return rc; @@ -478,13 +586,13 @@ static int base_region2_walk(unsigned long origin, unsigned long addr, return 0; } -static int base_region1_walk(unsigned long origin, unsigned long addr, +static int base_region1_walk(unsigned long *origin, unsigned long addr, unsigned long end, int alloc) { - unsigned long *rfte, next, table; + unsigned long *rfte, next, *table; int rc; - rfte = (unsigned long *) origin; + rfte = origin; rfte += (addr & _REGION1_INDEX) >> _REGION1_SHIFT; do { next = base_region1_addr_end(addr, end); @@ -494,9 +602,9 @@ static int base_region1_walk(unsigned long origin, unsigned long addr, table = base_crst_alloc(_REGION2_ENTRY_EMPTY); if (!table) return -ENOMEM; - *rfte = table | _REGION1_ENTRY; + *rfte = __pa(table) | _REGION1_ENTRY; } - table = *rfte & _REGION_ENTRY_ORIGIN; + table = __va(*rfte & _REGION_ENTRY_ORIGIN); rc = base_region2_walk(table, addr, next, alloc); if (rc) return rc; @@ -515,7 +623,7 @@ static int base_region1_walk(unsigned long origin, unsigned long addr, */ void base_asce_free(unsigned long asce) { - unsigned long table = asce & _ASCE_ORIGIN; + unsigned long *table = __va(asce & _ASCE_ORIGIN); if (!asce) return; @@ -567,7 +675,7 @@ static int base_pgt_cache_init(void) */ unsigned long base_asce_alloc(unsigned long addr, unsigned long num_pages) { - unsigned long asce, table, end; + unsigned long asce, *table, end; int rc; if (base_pgt_cache_init()) @@ -578,25 +686,25 @@ unsigned long base_asce_alloc(unsigned long addr, unsigned long num_pages) if (!table) return 0; rc = base_segment_walk(table, addr, end, 1); - asce = table | _ASCE_TYPE_SEGMENT | _ASCE_TABLE_LENGTH; + asce = __pa(table) | _ASCE_TYPE_SEGMENT | _ASCE_TABLE_LENGTH; } else if (end <= _REGION2_SIZE) { table = base_crst_alloc(_REGION3_ENTRY_EMPTY); if (!table) return 0; rc = base_region3_walk(table, addr, end, 1); - asce = table | _ASCE_TYPE_REGION3 | _ASCE_TABLE_LENGTH; + asce = __pa(table) | _ASCE_TYPE_REGION3 | _ASCE_TABLE_LENGTH; } else if (end <= _REGION1_SIZE) { table = base_crst_alloc(_REGION2_ENTRY_EMPTY); if (!table) return 0; rc = base_region2_walk(table, addr, end, 1); - asce = table | _ASCE_TYPE_REGION2 | _ASCE_TABLE_LENGTH; + asce = __pa(table) | _ASCE_TYPE_REGION2 | _ASCE_TABLE_LENGTH; } else { table = base_crst_alloc(_REGION1_ENTRY_EMPTY); if (!table) return 0; rc = base_region1_walk(table, addr, end, 1); - asce = table | _ASCE_TYPE_REGION1 | _ASCE_TABLE_LENGTH; + asce = __pa(table) | _ASCE_TYPE_REGION1 | _ASCE_TABLE_LENGTH; } if (rc) { base_asce_free(asce); |