/** * \file xpmem_private.h * License details are found in the file LICENSE. * \brief * Private Cross Partition Memory (XPMEM) structures and macros. * \author Yoichi Umezawa \par * Copyright (C) 2016 Yoichi Umezawa * * Original Copyright follows: * * This file is subject to the terms and conditions of the GNU General Public * License. See the file "COPYING" in the main directory of this archive * for more details. * * Copyright (c) 2004-2007 Silicon Graphics, Inc. All Rights Reserved. * Copyright 2009, 2010, 2014 Cray Inc. All Rights Reserved * Copyright (c) 2014-2016 Los Alamos National Security, LCC. All rights * reserved. */ /* * HISTORY */ #ifndef _XPMEM_PRIVATE_H #define _XPMEM_PRIVATE_H #include #include #define XPMEM_CURRENT_VERSION 0x00026003 //#define DEBUG_PRINT_XPMEM #ifdef DEBUG_PRINT_XPMEM #define dkprintf(...) kprintf(__VA_ARGS__) #define ekprintf(...) kprintf(__VA_ARGS__) #define XPMEM_DEBUG(format, a...) kprintf("[%d] %s: "format"\n", cpu_local_var(current)->proc->rgid, __func__, ##a) #else #define dkprintf(...) do { if (0) kprintf(__VA_ARGS__); } while (0) #define ekprintf(...) kprintf(__VA_ARGS__) #define XPMEM_DEBUG(format, a...) do { if (0) kprintf("\n"); } while (0) #endif //#define USE_DBUG_ON #ifdef USE_DBUG_ON #define DBUG_ON(condition) do { if (condition) kprintf("[%d] BUG: func=%s\n", cpu_local_var(current)->proc->rgid, __func__); } while (0) #else #define DBUG_ON(condition) #endif #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) #define min(x, y) ({ \ __typeof__(x) _min1 = (x); \ __typeof__(y) _min2 = (y); \ (void) (&_min1 == &_min2); \ _min1 < _min2 ? _min1 : _min2;}) #define max(x, y) ({ \ __typeof__(x) _max1 = (x); \ __typeof__(y) _max2 = (y); \ (void) (&_max1 == &_max2); \ _max1 > _max2 ? _max1 : _max2;}) #define MAX_ERRNO 4095 #define IS_ERR_VALUE(x) ((x) >= (unsigned long)-MAX_ERRNO) static inline void * ERR_PTR(long error) { return (void *)error; } static inline long PTR_ERR(const void *ptr) { return (long)ptr; } static inline long IS_ERR(const void *ptr) { return IS_ERR_VALUE((unsigned long)ptr); } static inline long IS_ERR_OR_NULL(const void *ptr) { return !ptr || IS_ERR_VALUE((unsigned long)ptr); } /* * Both the xpmem_segid_t and xpmem_apid_t are of type __s64 and designed * to be opaque to the user. Both consist of the same underlying fields. * * The 'uniq' field is designed to give each segid or apid a unique value. * Each type is only unique with respect to itself. * * An ID is never less than or equal to zero. */ struct xpmem_id { pid_t tgid; /* thread group that owns ID */ unsigned int uniq; /* this value makes the ID unique */ }; typedef union { struct xpmem_id xpmem_id; xpmem_segid_t segid; xpmem_apid_t apid; } xpmem_id_t; /* Shift INT_MAX by one so we can tell when we overflow. */ #define XPMEM_MAX_UNIQ_ID (INT_MAX >> 1) static inline pid_t xpmem_segid_to_tgid(xpmem_segid_t segid) { DBUG_ON(segid <= 0); return ((xpmem_id_t *)&segid)->xpmem_id.tgid; } static inline pid_t xpmem_apid_to_tgid(xpmem_apid_t apid) { DBUG_ON(apid <= 0); return ((xpmem_id_t *)&apid)->xpmem_id.tgid; } /* * Hash Tables * * XPMEM utilizes hash tables to enable faster lookups of list entries. * These hash tables are implemented as arrays. A simple modulus of the hash * key yields the appropriate array index. A hash table's array element (i.e., * hash table bucket) consists of a hash list and the lock that protects it. * * XPMEM has the following two hash tables: * * table bucket key * part->tg_hashtable list of struct xpmem_thread_group tgid * tg->ap_hashtable list of struct xpmem_access_permit apid.uniq */ struct xpmem_hashlist { mcs_rwlock_lock_t lock; /* lock for hash list */ struct list_head list; /* hash list */ }; #define XPMEM_TG_HASHTABLE_SIZE 8 #define XPMEM_AP_HASHTABLE_SIZE 8 static inline int xpmem_tg_hashtable_index(pid_t tgid) { int index; index = (unsigned int)tgid % XPMEM_TG_HASHTABLE_SIZE; XPMEM_DEBUG("return: tgid=%lu, index=%d", tgid, index); return index; } static inline int xpmem_ap_hashtable_index(xpmem_apid_t apid) { int index; DBUG_ON(apid <= 0); index = ((xpmem_id_t *)&apid)->xpmem_id.uniq % XPMEM_AP_HASHTABLE_SIZE; XPMEM_DEBUG("return: apid=0x%lx, index=%d", apid, index); return index; } /* * general internal driver structures */ struct xpmem_thread_group { ihk_spinlock_t lock; /* tg lock */ pid_t tgid; /* tg's tgid */ uid_t uid; /* tg's uid */ gid_t gid; /* tg's gid */ volatile int flags; /* tg attributes and state */ ihk_atomic_t uniq_segid; /* segid uniq */ ihk_atomic_t uniq_apid; /* apid uniq */ mcs_rwlock_lock_t seg_list_lock; /* tg's list of segs lock */ struct list_head seg_list; /* tg's list of segs */ ihk_atomic_t refcnt; /* references to tg */ ihk_atomic_t n_pinned; /* #of pages pinned by this tg */ struct list_head tg_hashlist; /* tg hash list */ struct thread *group_leader; /* thread group leader */ struct process_vm *vm; /* tg's process_vm */ struct xpmem_hashlist ap_hashtable[]; /* locks + ap hash lists */ }; struct xpmem_segment { ihk_spinlock_t lock; /* seg lock */ xpmem_segid_t segid; /* unique segid */ unsigned long vaddr; /* starting address */ size_t size; /* size of seg */ int permit_type; /* permission scheme */ void *permit_value; /* permission data */ volatile int flags; /* seg attributes and state */ ihk_atomic_t refcnt; /* references to seg */ struct xpmem_thread_group *tg; /* creator tg */ struct list_head ap_list; /* local access permits of seg */ struct list_head seg_list; /* tg's list of segs */ }; struct xpmem_access_permit { ihk_spinlock_t lock; /* access permit lock */ xpmem_apid_t apid; /* unique apid */ int mode; /* read/write mode */ volatile int flags; /* access permit attributes and state */ ihk_atomic_t refcnt; /* references to access permit */ struct xpmem_segment *seg; /* seg permitted to be accessed */ struct xpmem_thread_group *tg; /* access permit's tg */ struct list_head att_list; /* atts of this access permit's seg */ struct list_head ap_list; /* access permits linked to seg */ struct list_head ap_hashlist; /* access permit hash list */ }; struct xpmem_attachment { mcs_rwlock_lock_t at_lock; /* att lock */ unsigned long vaddr; /* starting address of seg attached */ unsigned long at_vaddr; /* address where seg is attached */ size_t at_size; /* size of seg attachment */ struct vm_range *at_vmr; /* vm_range where seg is attachment */ volatile int flags; /* att attributes and state */ ihk_atomic_t refcnt; /* references to att */ struct xpmem_access_permit *ap; /* associated access permit */ struct list_head att_list; /* atts linked to access permit */ struct process_vm *vm; /* process_vm attached to */ }; struct xpmem_partition { ihk_atomic_t n_opened; /* # of /dev/xpmem opened */ struct xpmem_hashlist tg_hashtable[]; /* locks + tg hash lists */ }; #define XPMEM_FLAG_DESTROYING 0x00040 /* being destroyed */ #define XPMEM_FLAG_DESTROYED 0x00080 /* 'being destroyed' finished */ #define XPMEM_FLAG_VALIDPTEs 0x00200 /* valid PTEs exist */ struct xpmem_perm { uid_t uid; gid_t gid; unsigned long mode; }; #define XPMEM_PERM_IRUSR 00400 #define XPMEM_PERM_IWUSR 00200 extern struct xpmem_partition *xpmem_my_part; static int xpmem_ioctl(struct mckfd *mckfd, ihk_mc_user_context_t *ctx); static int xpmem_close(struct mckfd *mckfd, ihk_mc_user_context_t *ctx); static int xpmem_init(void); static void xpmem_exit(void); static int __xpmem_open(void); static void xpmem_destroy_tg(struct xpmem_thread_group *); static int xpmem_make(unsigned long, size_t, int, void *, xpmem_segid_t *); static xpmem_segid_t xpmem_make_segid(struct xpmem_thread_group *); static int xpmem_remove(xpmem_segid_t); static void xpmem_remove_seg(struct xpmem_thread_group *, struct xpmem_segment *); static void xpmem_remove_segs_of_tg(struct xpmem_thread_group *seg_tg); static int xpmem_get(xpmem_segid_t, int, int, void *, xpmem_apid_t *); static int xpmem_check_permit_mode(int, struct xpmem_segment *); static int xpmem_perms(struct xpmem_perm *, short); static xpmem_apid_t xpmem_make_apid(struct xpmem_thread_group *); static int xpmem_release(xpmem_apid_t); static void xpmem_release_ap(struct xpmem_thread_group *, struct xpmem_access_permit *); static void xpmem_release_aps_of_tg(struct xpmem_thread_group *ap_tg); static int xpmem_attach(struct mckfd *, xpmem_apid_t, off_t, size_t, unsigned long, int, int, unsigned long *); static int xpmem_detach(unsigned long); static int xpmem_vm_munmap(struct process_vm *vm, void *addr, size_t len); static int xpmem_remove_process_range(struct process_vm *vm, unsigned long start, unsigned long end, int *ro_freedp); static int xpmem_free_process_memory_range(struct process_vm *vm, struct vm_range *range); static void xpmem_detach_att(struct xpmem_access_permit *, struct xpmem_attachment *); static void xpmem_clear_PTEs(struct xpmem_segment *); static void xpmem_clear_PTEs_range(struct xpmem_segment *, unsigned long, unsigned long); static void xpmem_clear_PTEs_of_ap(struct xpmem_access_permit *, unsigned long, unsigned long); static void xpmem_clear_PTEs_of_att(struct xpmem_attachment *, unsigned long, unsigned long); static int xpmem_remap_pte(struct process_vm *, struct vm_range *, unsigned long, uint64_t, struct xpmem_segment *, unsigned long); static int xpmem_ensure_valid_page(struct xpmem_segment *, unsigned long); static pte_t * xpmem_vaddr_to_pte(struct process_vm *, unsigned long, size_t *pgsize); static int xpmem_pin_page(struct xpmem_thread_group *, struct thread *, struct process_vm *, unsigned long); static void xpmem_unpin_pages(struct xpmem_segment *, struct process_vm *, unsigned long, size_t); static struct xpmem_thread_group * __xpmem_tg_ref_by_tgid_nolock_internal( pid_t, int, int); static inline struct xpmem_thread_group *__xpmem_tg_ref_by_tgid( pid_t tgid, int return_destroying) { struct xpmem_thread_group *tg; int index; struct mcs_rwlock_node_irqsave lock; XPMEM_DEBUG("call: tgid=%d, return_destroying=%d", tgid, return_destroying); index = xpmem_tg_hashtable_index(tgid); XPMEM_DEBUG("xpmem_my_part=%p\n", xpmem_my_part); XPMEM_DEBUG("xpmem_my_part->tg_hashtable=%p\n", xpmem_my_part->tg_hashtable); mcs_rwlock_reader_lock(&xpmem_my_part->tg_hashtable[index].lock, &lock); tg = __xpmem_tg_ref_by_tgid_nolock_internal(tgid, index, return_destroying); mcs_rwlock_reader_unlock(&xpmem_my_part->tg_hashtable[index].lock, &lock); XPMEM_DEBUG("return: tg=0x%p", tg); return tg; } static inline struct xpmem_thread_group *__xpmem_tg_ref_by_tgid_nolock( pid_t tgid, int return_destroying) { struct xpmem_thread_group *tg; XPMEM_DEBUG("call: tgid=%d, return_destroying=%d", tgid, return_destroying); tg = __xpmem_tg_ref_by_tgid_nolock_internal(tgid, xpmem_tg_hashtable_index(tgid), return_destroying); XPMEM_DEBUG("return: tg=0x%p", tg); return tg; } #define xpmem_tg_ref_by_tgid(t) __xpmem_tg_ref_by_tgid(t, 0) #define xpmem_tg_ref_by_tgid_all(t) __xpmem_tg_ref_by_tgid(t, 1) #define xpmem_tg_ref_by_tgid_nolock(t) __xpmem_tg_ref_by_tgid_nolock(t, 0) #define xpmem_tg_ref_by_tgid_all_nolock(t) __xpmem_tg_ref_by_tgid_nolock(t, 1) static struct xpmem_thread_group * xpmem_tg_ref_by_segid(xpmem_segid_t); static struct xpmem_thread_group * xpmem_tg_ref_by_apid(xpmem_apid_t); static void xpmem_tg_deref(struct xpmem_thread_group *); static struct xpmem_segment *xpmem_seg_ref_by_segid(struct xpmem_thread_group *, xpmem_segid_t); static void xpmem_seg_deref(struct xpmem_segment *); static struct xpmem_access_permit * xpmem_ap_ref_by_apid( struct xpmem_thread_group *, xpmem_apid_t); static void xpmem_ap_deref(struct xpmem_access_permit *); static void xpmem_att_deref(struct xpmem_attachment *); static int xpmem_validate_access(struct xpmem_access_permit *, off_t, size_t, int, unsigned long *); /* * Inlines that mark an internal driver structure as being destroyable or not. * The idea is to set the refcnt to 1 at structure creation time and then * drop that reference at the time the structure is to be destroyed. */ static inline void xpmem_tg_not_destroyable( struct xpmem_thread_group *tg) { ihk_atomic_set(&tg->refcnt, 1); XPMEM_DEBUG("return: tg->refcnt=%d", tg->refcnt); } static inline void xpmem_tg_destroyable( struct xpmem_thread_group *tg) { XPMEM_DEBUG("call: "); xpmem_tg_deref(tg); XPMEM_DEBUG("return: "); } static inline void xpmem_seg_not_destroyable( struct xpmem_segment *seg) { ihk_atomic_set(&seg->refcnt, 1); XPMEM_DEBUG("return: seg->refcnt=%d", seg->refcnt); } static inline void xpmem_seg_destroyable( struct xpmem_segment *seg) { XPMEM_DEBUG("call: "); xpmem_seg_deref(seg); XPMEM_DEBUG("return: "); } static inline void xpmem_ap_not_destroyable( struct xpmem_access_permit *ap) { ihk_atomic_set(&ap->refcnt, 1); XPMEM_DEBUG("return: ap->refcnt=%d", ap->refcnt); } static inline void xpmem_ap_destroyable( struct xpmem_access_permit *ap) { XPMEM_DEBUG("call: "); xpmem_ap_deref(ap); XPMEM_DEBUG("return: "); } static inline void xpmem_att_not_destroyable( struct xpmem_attachment *att) { ihk_atomic_set(&att->refcnt, 1); XPMEM_DEBUG("return: att->refcnt=%d", att->refcnt); } static inline void xpmem_att_destroyable( struct xpmem_attachment *att) { XPMEM_DEBUG("call: "); xpmem_att_deref(att); XPMEM_DEBUG("return: "); } /* * Inlines that increment the refcnt for the specified structure. */ static inline void xpmem_tg_ref( struct xpmem_thread_group *tg) { DBUG_ON(ihk_atomic_read(&tg->refcnt) <= 0); ihk_atomic_inc(&tg->refcnt); XPMEM_DEBUG("return: tg->refcnt=%d", tg->refcnt); } static inline void xpmem_seg_ref( struct xpmem_segment *seg) { DBUG_ON(ihk_atomic_read(&seg->refcnt) <= 0); ihk_atomic_inc(&seg->refcnt); XPMEM_DEBUG("return: seg->refcnt=%d", seg->refcnt); } static inline void xpmem_ap_ref( struct xpmem_access_permit *ap) { DBUG_ON(ihk_atomic_read(&ap->refcnt) <= 0); ihk_atomic_inc(&ap->refcnt); XPMEM_DEBUG("return: ap->refcnt=%d", ap->refcnt); } static inline void xpmem_att_ref( struct xpmem_attachment *att) { DBUG_ON(ihk_atomic_read(&att->refcnt) <= 0); ihk_atomic_inc(&att->refcnt); XPMEM_DEBUG("return: att->refcnt=%d", att->refcnt); } static inline int xpmem_is_private_data( struct vm_range *vmr) { return (vmr->private_data != NULL); } #endif /* _XPMEM_PRIVATE_H */