From: David Howells Here's a patch to allocate memory for big system hash tables with the bootmem allocator rather than with main page allocator. It is needed for three reasons: (1) So that the size can be bigger than MAX_ORDER. IBM have done some testing on their big PPC64 systems (64GB of RAM) with linux-2.4 and found that they get better performance if the sizes of the inode cache hash, dentry cache hash, buffer head hash and page cache hash are increased beyond MAX_ORDER (order 11). Now the main allocator can't allocate anything larger than MAX_ORDER, but the bootmem allocator can. In 2.6 it appears that only the inode and dentry hashes remain of those four, but there are other hash tables that could use this service. (2) Changing MAX_ORDER appears to have a number of effects beyond just limiting the maximum size that can be allocated in one go. (3) Should someone want a hash table in which each bucket isn't a power of two in size, memory will be wasted as the chunk of memory allocated will be a power of two in size (to hold a power of two number of buckets). On the other hand, using the bootmem allocator means the allocation will only take up sufficient pages to hold it, rather than the next power of two up. Admittedly, this point doesn't apply to the dentry and inode hashes, but it might to another hash table that might want to use this service. I've coelesced the meat of the inode and dentry allocation routines into one such routine in mm/page_alloc.c that the the respective initialisation functions now call before mem_init() is called. This routine gets it's approximation of memory size by counting up the ZONE_NORMAL and ZONE_DMA pages (and ZONE_HIGHMEM if requested) in all the nodes passed to the main allocator by paging_init() (or wherever the arch does it). It does not use max_low_pfn as that doesn't seem to be available on all archs, and it doesn't use num_physpages since that includes highmem pages not available to the kernel for allocating data structures upon - which may not be appropriate when calculating hash table size. On the off chance that the size of each hash bucket may not be exactly a power of two, the routine will only allocate as many pages as is necessary to ensure that the number of buckets is exactly a power of two, rather than allocating the smallest power-of-two sized chunk of memory that will hold the same array of buckets. The maximum size of any single hash table is given by MAX_SYS_HASH_TABLE_ORDER, as is now defined in linux/mmzone.h. DESC permit-inode-dentry-hash-tables-to-be-allocated-max_order-size-warning-fix EDESC Signed-off-by: Andrew Morton --- 25-akpm/fs/dcache.c | 74 ++++++++++++++---------------------- 25-akpm/fs/inode.c | 66 +++++++++++--------------------- 25-akpm/include/linux/bootmem.h | 8 +++ 25-akpm/include/linux/fs.h | 2 25-akpm/include/linux/mmzone.h | 12 +++++ 25-akpm/init/main.c | 1 25-akpm/mm/page_alloc.c | 82 ++++++++++++++++++++++++++++++++++++++++ 7 files changed, 159 insertions(+), 86 deletions(-) diff -puN fs/dcache.c~permit-inode-dentry-hash-tables-to-be-allocated-max_order-size fs/dcache.c --- 25/fs/dcache.c~permit-inode-dentry-hash-tables-to-be-allocated-max_order-size Fri Jun 18 15:41:02 2004 +++ 25-akpm/fs/dcache.c Fri Jun 18 15:41:02 2004 @@ -30,6 +30,7 @@ #include #include #include +#include #define DCACHE_PARANOIA 1 /* #define DCACHE_DEBUG 1 */ @@ -1561,13 +1562,31 @@ static int __init set_dhash_entries(char } __setup("dhash_entries=", set_dhash_entries); -static void __init dcache_init(unsigned long mempages) +static void __init dcache_init_early(void) { - struct hlist_head *d; - unsigned long order; - unsigned int nr_hash; - int i; + struct hlist_head *p; + int loop; + + dentry_hashtable = + alloc_large_system_hash("Dentry cache", + sizeof(struct hlist_head), + dhash_entries, + 13, + 0, + &d_hash_shift, + &d_hash_mask); + + p = dentry_hashtable; + loop = 1 << d_hash_shift; + do { + INIT_HLIST_HEAD(p); + p++; + loop--; + } while (loop); +} +static void __init dcache_init(unsigned long mempages) +{ /* * A constructor could be added for stable state like the lists, * but it is probably not worth it because of the cache nature @@ -1580,45 +1599,6 @@ static void __init dcache_init(unsigned NULL, NULL); set_shrinker(DEFAULT_SEEKS, shrink_dcache_memory); - - if (!dhash_entries) - dhash_entries = PAGE_SHIFT < 13 ? - mempages >> (13 - PAGE_SHIFT) : - mempages << (PAGE_SHIFT - 13); - - dhash_entries *= sizeof(struct hlist_head); - for (order = 0; ((1UL << order) << PAGE_SHIFT) < dhash_entries; order++) - ; - - do { - unsigned long tmp; - - nr_hash = (1UL << order) * PAGE_SIZE / - sizeof(struct hlist_head); - d_hash_mask = (nr_hash - 1); - - tmp = nr_hash; - d_hash_shift = 0; - while ((tmp >>= 1UL) != 0UL) - d_hash_shift++; - - dentry_hashtable = (struct hlist_head *) - __get_free_pages(GFP_ATOMIC, order); - } while (dentry_hashtable == NULL && --order >= 0); - - printk(KERN_INFO "Dentry cache hash table entries: %d (order: %ld, %ld bytes)\n", - nr_hash, order, (PAGE_SIZE << order)); - - if (!dentry_hashtable) - panic("Failed to allocate dcache hash table\n"); - - d = dentry_hashtable; - i = nr_hash; - do { - INIT_HLIST_HEAD(d); - d++; - i--; - } while (i); } /* SLAB cache for __getname() consumers */ @@ -1632,6 +1612,12 @@ EXPORT_SYMBOL(d_genocide); extern void bdev_cache_init(void); extern void chrdev_init(void); +void __init vfs_caches_init_early(void) +{ + dcache_init_early(); + inode_init_early(); +} + void __init vfs_caches_init(unsigned long mempages) { unsigned long reserve; diff -puN fs/inode.c~permit-inode-dentry-hash-tables-to-be-allocated-max_order-size fs/inode.c --- 25/fs/inode.c~permit-inode-dentry-hash-tables-to-be-allocated-max_order-size Fri Jun 18 15:41:02 2004 +++ 25-akpm/fs/inode.c Fri Jun 18 15:41:02 2004 @@ -20,6 +20,7 @@ #include #include #include +#include /* * This is needed for the following functions: @@ -1334,54 +1335,35 @@ __setup("ihash_entries=", set_ihash_entr /* * Initialize the waitqueues and inode hash table. */ -void __init inode_init(unsigned long mempages) +void __init inode_init_early(void) { - struct hlist_head *head; - unsigned long order; - unsigned int nr_hash; - int i; - - for (i = 0; i < ARRAY_SIZE(i_wait_queue_heads); i++) - init_waitqueue_head(&i_wait_queue_heads[i].wqh); + struct hlist_head *p; + int loop; - if (!ihash_entries) - ihash_entries = PAGE_SHIFT < 14 ? - mempages >> (14 - PAGE_SHIFT) : - mempages << (PAGE_SHIFT - 14); - - ihash_entries *= sizeof(struct hlist_head); - for (order = 0; ((1UL << order) << PAGE_SHIFT) < ihash_entries; order++) - ; + inode_hashtable = + alloc_large_system_hash("Inode-cache", + sizeof(struct hlist_head), + ihash_entries, + 14, + 0, + &i_hash_shift, + &i_hash_mask); + p = inode_hashtable; + loop = 1 << i_hash_shift; do { - unsigned long tmp; - - nr_hash = (1UL << order) * PAGE_SIZE / - sizeof(struct hlist_head); - i_hash_mask = (nr_hash - 1); - - tmp = nr_hash; - i_hash_shift = 0; - while ((tmp >>= 1UL) != 0UL) - i_hash_shift++; - - inode_hashtable = (struct hlist_head *) - __get_free_pages(GFP_ATOMIC, order); - } while (inode_hashtable == NULL && --order >= 0); - - printk("Inode-cache hash table entries: %d (order: %ld, %ld bytes)\n", - nr_hash, order, (PAGE_SIZE << order)); + INIT_HLIST_HEAD(p); + p++; + loop--; + } while (loop); +} - if (!inode_hashtable) - panic("Failed to allocate inode hash table\n"); +void __init inode_init(unsigned long mempages) +{ + int i; - head = inode_hashtable; - i = nr_hash; - do { - INIT_HLIST_HEAD(head); - head++; - i--; - } while (i); + for (i = 0; i < ARRAY_SIZE(i_wait_queue_heads); i++) + init_waitqueue_head(&i_wait_queue_heads[i].wqh); /* inode slab cache */ inode_cachep = kmem_cache_create("inode_cache", sizeof(struct inode), diff -puN include/linux/bootmem.h~permit-inode-dentry-hash-tables-to-be-allocated-max_order-size include/linux/bootmem.h --- 25/include/linux/bootmem.h~permit-inode-dentry-hash-tables-to-be-allocated-max_order-size Fri Jun 18 15:41:02 2004 +++ 25-akpm/include/linux/bootmem.h Fri Jun 18 15:41:02 2004 @@ -67,4 +67,12 @@ extern void * __init __alloc_bootmem_nod __alloc_bootmem_node((pgdat), (x), PAGE_SIZE, 0) #endif /* !CONFIG_HAVE_ARCH_BOOTMEM_NODE */ +extern void *__init alloc_large_system_hash(const char *tablename, + unsigned long bucketsize, + unsigned long numentries, + int scale, + int consider_highmem, + unsigned int *_hash_shift, + unsigned int *_hash_mask); + #endif /* _LINUX_BOOTMEM_H */ diff -puN include/linux/fs.h~permit-inode-dentry-hash-tables-to-be-allocated-max_order-size include/linux/fs.h --- 25/include/linux/fs.h~permit-inode-dentry-hash-tables-to-be-allocated-max_order-size Fri Jun 18 15:41:02 2004 +++ 25-akpm/include/linux/fs.h Fri Jun 18 15:41:02 2004 @@ -222,6 +222,7 @@ extern int leases_enable, dir_notify_ena extern void update_atime (struct inode *); extern void inode_init(unsigned long); +extern void inode_init_early(void); extern void mnt_init(unsigned long); extern void files_init(unsigned long); @@ -1203,6 +1204,7 @@ extern int filp_close(struct file *, fl_ extern char * getname(const char __user *); /* fs/dcache.c */ +extern void vfs_caches_init_early(void); extern void vfs_caches_init(unsigned long); #define __getname() kmem_cache_alloc(names_cachep, SLAB_KERNEL) diff -puN include/linux/mmzone.h~permit-inode-dentry-hash-tables-to-be-allocated-max_order-size include/linux/mmzone.h --- 25/include/linux/mmzone.h~permit-inode-dentry-hash-tables-to-be-allocated-max_order-size Fri Jun 18 15:41:02 2004 +++ 25-akpm/include/linux/mmzone.h Fri Jun 18 15:41:02 2004 @@ -20,6 +20,18 @@ #define MAX_ORDER CONFIG_FORCE_MAX_ZONEORDER #endif +/* + * system hash table size limits + * - on large memory machines, we may want to allocate a bigger hash than that + * permitted by MAX_ORDER, so we allocate with the bootmem allocator, and are + * limited to this size + */ +#if MAX_ORDER > 14 +#define MAX_SYS_HASH_TABLE_ORDER MAX_ORDER +#else +#define MAX_SYS_HASH_TABLE_ORDER 14 +#endif + struct free_area { struct list_head free_list; unsigned long *map; diff -puN init/main.c~permit-inode-dentry-hash-tables-to-be-allocated-max_order-size init/main.c --- 25/init/main.c~permit-inode-dentry-hash-tables-to-be-allocated-max_order-size Fri Jun 18 15:41:02 2004 +++ 25-akpm/init/main.c Fri Jun 18 15:41:02 2004 @@ -499,6 +499,7 @@ asmlinkage void __init start_kernel(void initrd_start = 0; } #endif + vfs_caches_init_early(); mem_init(); kmem_cache_init(); numa_policy_init(); diff -puN mm/page_alloc.c~permit-inode-dentry-hash-tables-to-be-allocated-max_order-size mm/page_alloc.c --- 25/mm/page_alloc.c~permit-inode-dentry-hash-tables-to-be-allocated-max_order-size Fri Jun 18 15:41:02 2004 +++ 25-akpm/mm/page_alloc.c Fri Jun 18 15:41:02 2004 @@ -55,6 +55,9 @@ EXPORT_SYMBOL(zone_table); static char *zone_names[MAX_NR_ZONES] = { "DMA", "Normal", "HighMem" }; int min_free_kbytes = 1024; +static unsigned long __initdata nr_kernel_pages; +static unsigned long __initdata nr_all_pages; + /* * Temporary debugging check for pages not lying within a given zone. */ @@ -1430,6 +1433,10 @@ static void __init free_area_init_core(s if (zholes_size) realsize -= zholes_size[j]; + if (j == ZONE_DMA || j == ZONE_NORMAL) + nr_kernel_pages += realsize; + nr_all_pages += realsize; + zone->spanned_pages = size; zone->present_pages = realsize; zone->name = zone_names[j]; @@ -1970,3 +1977,78 @@ int lower_zone_protection_sysctl_handler setup_per_zone_protection(); return 0; } + +/* + * allocate a large system hash table from bootmem + * - it is assumed that the hash table must contain an exact power-of-2 + * quantity of entries + */ +static inline int calc_log2(unsigned long x) __attribute_pure__; +static inline int calc_log2(unsigned long x) +{ + int r = 0; + for (x >>= 1; x > 0; x >>= 1) + r++; + return r; +} + +void *__init alloc_large_system_hash(const char *tablename, + unsigned long bucketsize, + unsigned long numentries, + int scale, + int consider_highmem, + unsigned int *_hash_shift, + unsigned int *_hash_mask) +{ + unsigned long mem, max, log2qty, size; + void *table; + + /* round applicable memory size up to nearest megabyte */ + mem = consider_highmem ? nr_all_pages : nr_kernel_pages; + mem += (1UL << (20 - PAGE_SHIFT)) - 1; + mem >>= 20 - PAGE_SHIFT; + mem <<= 20 - PAGE_SHIFT; + + /* limit to 1 bucket per 2^scale bytes of low memory (rounded up to + * nearest power of 2 in size) */ + if (scale > PAGE_SHIFT) + mem >>= (scale - PAGE_SHIFT); + else + mem <<= (PAGE_SHIFT - scale); + + mem = 1UL << (calc_log2(mem) + 1); + + /* limit allocation size */ + max = (1UL << (PAGE_SHIFT + MAX_SYS_HASH_TABLE_ORDER)) / bucketsize; + if (max > mem) + max = mem; + + /* allow the kernel cmdline to have a say */ + if (!numentries || numentries > max) + numentries = max; + + log2qty = calc_log2(numentries); + + do { + size = bucketsize << log2qty; + + table = alloc_bootmem(size); + + } while (!table && size > PAGE_SIZE && --log2qty); + + if (!table) + panic("Failed to allocate %s hash table\n", tablename); + + printk("%s hash table entries: %d (order: %d, %lu bytes)\n", + tablename, + (1U << log2qty), + calc_log2(size) - PAGE_SHIFT, + size); + + if (_hash_shift) + *_hash_shift = log2qty; + if (_hash_mask) + *_hash_mask = (1 << log2qty) - 1; + + return table; +} _