Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / drivers / staging / lustre / include / linux / libcfs / libcfs_private.h
diff --git a/kernel/drivers/staging/lustre/include/linux/libcfs/libcfs_private.h b/kernel/drivers/staging/lustre/include/linux/libcfs/libcfs_private.h
new file mode 100644 (file)
index 0000000..fef8825
--- /dev/null
@@ -0,0 +1,556 @@
+/*
+ * GPL HEADER START
+ *
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 only,
+ * as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License version 2 for more details (a copy is included
+ * in the LICENSE file that accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License
+ * version 2 along with this program; If not, see
+ * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
+ *
+ * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
+ * CA 95054 USA or visit www.sun.com if you need additional information or
+ * have any questions.
+ *
+ * GPL HEADER END
+ */
+/*
+ * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Use is subject to license terms.
+ *
+ * Copyright (c) 2011, 2012, Intel Corporation.
+ */
+/*
+ * This file is part of Lustre, http://www.lustre.org/
+ * Lustre is a trademark of Sun Microsystems, Inc.
+ *
+ * libcfs/include/libcfs/libcfs_private.h
+ *
+ * Various defines for libcfs.
+ *
+ */
+
+#ifndef __LIBCFS_PRIVATE_H__
+#define __LIBCFS_PRIVATE_H__
+
+/* XXX this layering violation is for nidstrings */
+#include "../lnet/types.h"
+
+#ifndef DEBUG_SUBSYSTEM
+# define DEBUG_SUBSYSTEM S_UNDEFINED
+#endif
+
+
+/*
+ * When this is on, LASSERT macro includes check for assignment used instead
+ * of equality check, but doesn't have unlikely(). Turn this on from time to
+ * time to make test-builds. This shouldn't be on for production release.
+ */
+#define LASSERT_CHECKED (0)
+
+#define LASSERTF(cond, fmt, ...)                                       \
+do {                                                                   \
+       if (unlikely(!(cond))) {                                        \
+               LIBCFS_DEBUG_MSG_DATA_DECL(__msg_data, D_EMERG, NULL);  \
+               libcfs_debug_msg(&__msg_data,                           \
+                                "ASSERTION( %s ) failed: " fmt, #cond, \
+                                ## __VA_ARGS__);                       \
+               lbug_with_loc(&__msg_data);                             \
+       }                                                               \
+} while (0)
+
+#define LASSERT(cond) LASSERTF(cond, "\n")
+
+#ifdef CONFIG_LUSTRE_DEBUG_EXPENSIVE_CHECK
+/**
+ * This is for more expensive checks that one doesn't want to be enabled all
+ * the time. LINVRNT() has to be explicitly enabled by
+ * CONFIG_LUSTRE_DEBUG_EXPENSIVE_CHECK option.
+ */
+# define LINVRNT(exp) LASSERT(exp)
+#else
+# define LINVRNT(exp) ((void)sizeof !!(exp))
+#endif
+
+#define KLASSERT(e) LASSERT(e)
+
+void lbug_with_loc(struct libcfs_debug_msg_data *)__attribute__((noreturn));
+
+#define LBUG()                                                   \
+do {                                                               \
+       LIBCFS_DEBUG_MSG_DATA_DECL(msgdata, D_EMERG, NULL);          \
+       lbug_with_loc(&msgdata);                                        \
+} while (0)
+
+extern atomic_t libcfs_kmemory;
+/*
+ * Memory
+ */
+
+# define libcfs_kmem_inc(ptr, size)            \
+do {                                           \
+       atomic_add(size, &libcfs_kmemory);      \
+} while (0)
+
+# define libcfs_kmem_dec(ptr, size)            \
+do {                                           \
+       atomic_sub(size, &libcfs_kmemory);      \
+} while (0)
+
+# define libcfs_kmem_read()                    \
+       atomic_read(&libcfs_kmemory)
+
+#ifndef LIBCFS_VMALLOC_SIZE
+#define LIBCFS_VMALLOC_SIZE    (2 << PAGE_CACHE_SHIFT) /* 2 pages */
+#endif
+
+#define LIBCFS_ALLOC_PRE(size, mask)                                       \
+do {                                                                       \
+       LASSERT(!in_interrupt() ||                                          \
+               ((size) <= LIBCFS_VMALLOC_SIZE &&                           \
+                ((mask) & __GFP_WAIT) == 0));                              \
+} while (0)
+
+#define LIBCFS_ALLOC_POST(ptr, size)                                       \
+do {                                                                       \
+       if (unlikely((ptr) == NULL)) {                                      \
+               CERROR("LNET: out of memory at %s:%d (tried to alloc '"     \
+                      #ptr "' = %d)\n", __FILE__, __LINE__, (int)(size));  \
+               CERROR("LNET: %d total bytes allocated by lnet\n",          \
+                      libcfs_kmem_read());                                 \
+       } else {                                                            \
+               memset((ptr), 0, (size));                                   \
+               libcfs_kmem_inc((ptr), (size));                             \
+               CDEBUG(D_MALLOC, "alloc '" #ptr "': %d at %p (tot %d).\n",  \
+                      (int)(size), (ptr), libcfs_kmem_read());             \
+       }                                                                  \
+} while (0)
+
+/**
+ * allocate memory with GFP flags @mask
+ */
+#define LIBCFS_ALLOC_GFP(ptr, size, mask)                                  \
+do {                                                                       \
+       LIBCFS_ALLOC_PRE((size), (mask));                                   \
+       (ptr) = (size) <= LIBCFS_VMALLOC_SIZE ?                             \
+               kmalloc((size), (mask)) : vmalloc(size);            \
+       LIBCFS_ALLOC_POST((ptr), (size));                                   \
+} while (0)
+
+/**
+ * default allocator
+ */
+#define LIBCFS_ALLOC(ptr, size) \
+       LIBCFS_ALLOC_GFP(ptr, size, GFP_NOFS)
+
+/**
+ * non-sleeping allocator
+ */
+#define LIBCFS_ALLOC_ATOMIC(ptr, size) \
+       LIBCFS_ALLOC_GFP(ptr, size, GFP_ATOMIC)
+
+/**
+ * allocate memory for specified CPU partition
+ *   \a cptab != NULL, \a cpt is CPU partition id of \a cptab
+ *   \a cptab == NULL, \a cpt is HW NUMA node id
+ */
+#define LIBCFS_CPT_ALLOC_GFP(ptr, cptab, cpt, size, mask)                  \
+do {                                                                       \
+       LIBCFS_ALLOC_PRE((size), (mask));                                   \
+       (ptr) = (size) <= LIBCFS_VMALLOC_SIZE ?                             \
+               kmalloc_node((size), (mask), cfs_cpt_spread_node(cptab, cpt)) :\
+               vmalloc_node(size, cfs_cpt_spread_node(cptab, cpt));        \
+       LIBCFS_ALLOC_POST((ptr), (size));                                   \
+} while (0)
+
+/** default numa allocator */
+#define LIBCFS_CPT_ALLOC(ptr, cptab, cpt, size)                                    \
+       LIBCFS_CPT_ALLOC_GFP(ptr, cptab, cpt, size, GFP_NOFS)
+
+#define LIBCFS_FREE(ptr, size)                                   \
+do {                                                               \
+       int s = (size);                                          \
+       if (unlikely((ptr) == NULL)) {                            \
+               CERROR("LIBCFS: free NULL '" #ptr "' (%d bytes) at "    \
+                      "%s:%d\n", s, __FILE__, __LINE__);              \
+               break;                                            \
+       }                                                              \
+       libcfs_kmem_dec((ptr), s);                                    \
+       CDEBUG(D_MALLOC, "kfreed '" #ptr "': %d at %p (tot %d).\n",     \
+              s, (ptr), libcfs_kmem_read());                           \
+       if (unlikely(s > LIBCFS_VMALLOC_SIZE))                    \
+               vfree(ptr);                                 \
+       else                                                        \
+               kfree(ptr);                                       \
+} while (0)
+
+/******************************************************************************/
+
+/* htonl hack - either this, or compile with -O2. Stupid byteorder/generic.h */
+#if defined(__GNUC__) && (__GNUC__ >= 2) && !defined(__OPTIMIZE__)
+#define ___htonl(x) __cpu_to_be32(x)
+#define ___htons(x) __cpu_to_be16(x)
+#define ___ntohl(x) __be32_to_cpu(x)
+#define ___ntohs(x) __be16_to_cpu(x)
+#define htonl(x) ___htonl(x)
+#define ntohl(x) ___ntohl(x)
+#define htons(x) ___htons(x)
+#define ntohs(x) ___ntohs(x)
+#endif
+
+void libcfs_run_upcall(char **argv);
+void libcfs_run_lbug_upcall(struct libcfs_debug_msg_data *);
+void libcfs_debug_dumplog(void);
+int libcfs_debug_init(unsigned long bufsize);
+int libcfs_debug_cleanup(void);
+int libcfs_debug_clear_buffer(void);
+int libcfs_debug_mark_buffer(const char *text);
+
+void libcfs_debug_set_level(unsigned int debug_level);
+
+/*
+ * allocate per-cpu-partition data, returned value is an array of pointers,
+ * variable can be indexed by CPU ID.
+ *     cptable != NULL: size of array is number of CPU partitions
+ *     cptable == NULL: size of array is number of HW cores
+ */
+void *cfs_percpt_alloc(struct cfs_cpt_table *cptab, unsigned int size);
+/*
+ * destroy per-cpu-partition variable
+ */
+void  cfs_percpt_free(void *vars);
+int   cfs_percpt_number(void *vars);
+void *cfs_percpt_current(void *vars);
+void *cfs_percpt_index(void *vars, int idx);
+
+#define cfs_percpt_for_each(var, i, vars)              \
+       for (i = 0; i < cfs_percpt_number(vars) &&      \
+                   ((var) = (vars)[i]) != NULL; i++)
+
+/*
+ * allocate a variable array, returned value is an array of pointers.
+ * Caller can specify length of array by count.
+ */
+void *cfs_array_alloc(int count, unsigned int size);
+void  cfs_array_free(void *vars);
+
+#define LASSERT_ATOMIC_ENABLED   (1)
+
+#if LASSERT_ATOMIC_ENABLED
+
+/** assert value of @a is equal to @v */
+#define LASSERT_ATOMIC_EQ(a, v)                                 \
+do {                                                       \
+       LASSERTF(atomic_read(a) == v,                  \
+                "value: %d\n", atomic_read((a)));        \
+} while (0)
+
+/** assert value of @a is unequal to @v */
+#define LASSERT_ATOMIC_NE(a, v)                                 \
+do {                                                       \
+       LASSERTF(atomic_read(a) != v,                  \
+                "value: %d\n", atomic_read((a)));        \
+} while (0)
+
+/** assert value of @a is little than @v */
+#define LASSERT_ATOMIC_LT(a, v)                                 \
+do {                                                       \
+       LASSERTF(atomic_read(a) < v,                    \
+                "value: %d\n", atomic_read((a)));        \
+} while (0)
+
+/** assert value of @a is little/equal to @v */
+#define LASSERT_ATOMIC_LE(a, v)                                 \
+do {                                                       \
+       LASSERTF(atomic_read(a) <= v,                  \
+                "value: %d\n", atomic_read((a)));        \
+} while (0)
+
+/** assert value of @a is great than @v */
+#define LASSERT_ATOMIC_GT(a, v)                                 \
+do {                                                       \
+       LASSERTF(atomic_read(a) > v,                    \
+                "value: %d\n", atomic_read((a)));        \
+} while (0)
+
+/** assert value of @a is great/equal to @v */
+#define LASSERT_ATOMIC_GE(a, v)                                 \
+do {                                                       \
+       LASSERTF(atomic_read(a) >= v,                  \
+                "value: %d\n", atomic_read((a)));        \
+} while (0)
+
+/** assert value of @a is great than @v1 and little than @v2 */
+#define LASSERT_ATOMIC_GT_LT(a, v1, v2)                         \
+do {                                                       \
+       int __v = atomic_read(a);                          \
+       LASSERTF(__v > v1 && __v < v2, "value: %d\n", __v);     \
+} while (0)
+
+/** assert value of @a is great than @v1 and little/equal to @v2 */
+#define LASSERT_ATOMIC_GT_LE(a, v1, v2)                         \
+do {                                                       \
+       int __v = atomic_read(a);                          \
+       LASSERTF(__v > v1 && __v <= v2, "value: %d\n", __v);    \
+} while (0)
+
+/** assert value of @a is great/equal to @v1 and little than @v2 */
+#define LASSERT_ATOMIC_GE_LT(a, v1, v2)                         \
+do {                                                       \
+       int __v = atomic_read(a);                          \
+       LASSERTF(__v >= v1 && __v < v2, "value: %d\n", __v);    \
+} while (0)
+
+/** assert value of @a is great/equal to @v1 and little/equal to @v2 */
+#define LASSERT_ATOMIC_GE_LE(a, v1, v2)                         \
+do {                                                       \
+       int __v = atomic_read(a);                          \
+       LASSERTF(__v >= v1 && __v <= v2, "value: %d\n", __v);   \
+} while (0)
+
+#else /* !LASSERT_ATOMIC_ENABLED */
+
+#define LASSERT_ATOMIC_EQ(a, v)                 do {} while (0)
+#define LASSERT_ATOMIC_NE(a, v)                 do {} while (0)
+#define LASSERT_ATOMIC_LT(a, v)                 do {} while (0)
+#define LASSERT_ATOMIC_LE(a, v)                 do {} while (0)
+#define LASSERT_ATOMIC_GT(a, v)                 do {} while (0)
+#define LASSERT_ATOMIC_GE(a, v)                 do {} while (0)
+#define LASSERT_ATOMIC_GT_LT(a, v1, v2)         do {} while (0)
+#define LASSERT_ATOMIC_GT_LE(a, v1, v2)         do {} while (0)
+#define LASSERT_ATOMIC_GE_LT(a, v1, v2)         do {} while (0)
+#define LASSERT_ATOMIC_GE_LE(a, v1, v2)         do {} while (0)
+
+#endif /* LASSERT_ATOMIC_ENABLED */
+
+#define LASSERT_ATOMIC_ZERO(a)           LASSERT_ATOMIC_EQ(a, 0)
+#define LASSERT_ATOMIC_POS(a)             LASSERT_ATOMIC_GT(a, 0)
+
+#define CFS_ALLOC_PTR(ptr)      LIBCFS_ALLOC(ptr, sizeof(*(ptr)))
+#define CFS_FREE_PTR(ptr)       LIBCFS_FREE(ptr, sizeof(*(ptr)))
+
+/*
+ * percpu partition lock
+ *
+ * There are some use-cases like this in Lustre:
+ * . each CPU partition has it's own private data which is frequently changed,
+ *   and mostly by the local CPU partition.
+ * . all CPU partitions share some global data, these data are rarely changed.
+ *
+ * LNet is typical example.
+ * CPU partition lock is designed for this kind of use-cases:
+ * . each CPU partition has it's own private lock
+ * . change on private data just needs to take the private lock
+ * . read on shared data just needs to take _any_ of private locks
+ * . change on shared data needs to take _all_ private locks,
+ *   which is slow and should be really rare.
+ */
+
+enum {
+       CFS_PERCPT_LOCK_EX      = -1, /* negative */
+};
+
+struct cfs_percpt_lock {
+       /* cpu-partition-table for this lock */
+       struct cfs_cpt_table    *pcl_cptab;
+       /* exclusively locked */
+       unsigned int            pcl_locked;
+       /* private lock table */
+       spinlock_t              **pcl_locks;
+};
+
+/* return number of private locks */
+static inline int
+cfs_percpt_lock_num(struct cfs_percpt_lock *pcl)
+{
+       return cfs_cpt_number(pcl->pcl_cptab);
+}
+
+/*
+ * create a cpu-partition lock based on CPU partition table \a cptab,
+ * each private lock has extra \a psize bytes padding data
+ */
+struct cfs_percpt_lock *cfs_percpt_lock_alloc(struct cfs_cpt_table *cptab);
+/* destroy a cpu-partition lock */
+void cfs_percpt_lock_free(struct cfs_percpt_lock *pcl);
+
+/* lock private lock \a index of \a pcl */
+void cfs_percpt_lock(struct cfs_percpt_lock *pcl, int index);
+/* unlock private lock \a index of \a pcl */
+void cfs_percpt_unlock(struct cfs_percpt_lock *pcl, int index);
+/* create percpt (atomic) refcount based on @cptab */
+atomic_t **cfs_percpt_atomic_alloc(struct cfs_cpt_table *cptab, int val);
+/* destroy percpt refcount */
+void cfs_percpt_atomic_free(atomic_t **refs);
+/* return sum of all percpu refs */
+int cfs_percpt_atomic_summary(atomic_t **refs);
+
+/** Compile-time assertion.
+
+ * Check an invariant described by a constant expression at compile time by
+ * forcing a compiler error if it does not hold.  \a cond must be a constant
+ * expression as defined by the ISO C Standard:
+ *
+ *       6.8.4.2  The switch statement
+ *       ....
+ *       [#3] The expression of each case label shall be  an  integer
+ *       constant   expression  and  no  two  of  the  case  constant
+ *       expressions in the same switch statement shall have the same
+ *       value  after  conversion...
+ *
+ */
+#define CLASSERT(cond) do {switch (42) {case (cond): case 0: break; } } while (0)
+
+/* support decl needed both by kernel and liblustre */
+int     libcfs_isknown_lnd(int type);
+char       *libcfs_lnd2modname(int type);
+char       *libcfs_lnd2str(int type);
+int     libcfs_str2lnd(const char *str);
+char       *libcfs_net2str(__u32 net);
+char       *libcfs_nid2str(lnet_nid_t nid);
+__u32       libcfs_str2net(const char *str);
+lnet_nid_t  libcfs_str2nid(const char *str);
+int     libcfs_str2anynid(lnet_nid_t *nid, const char *str);
+char       *libcfs_id2str(lnet_process_id_t id);
+void   cfs_free_nidlist(struct list_head *list);
+int     cfs_parse_nidlist(char *str, int len, struct list_head *list);
+int     cfs_match_nid(lnet_nid_t nid, struct list_head *list);
+
+/** \addtogroup lnet_addr
+ * @{ */
+/* how an LNET NID encodes net:address */
+/** extract the address part of an lnet_nid_t */
+#define LNET_NIDADDR(nid)      ((__u32)((nid) & 0xffffffff))
+/** extract the network part of an lnet_nid_t */
+#define LNET_NIDNET(nid)       ((__u32)(((nid) >> 32)) & 0xffffffff)
+/** make an lnet_nid_t from a network part and an address part */
+#define LNET_MKNID(net, addr)   ((((__u64)(net))<<32)|((__u64)(addr)))
+/* how net encodes type:number */
+#define LNET_NETNUM(net)       ((net) & 0xffff)
+#define LNET_NETTYP(net)       (((net) >> 16) & 0xffff)
+#define LNET_MKNET(typ, num)    ((((__u32)(typ))<<16)|((__u32)(num)))
+/** @} lnet_addr */
+
+/* max value for numeric network address */
+#define MAX_NUMERIC_VALUE 0xffffffff
+
+/* implication */
+#define ergo(a, b) (!(a) || (b))
+/* logical equivalence */
+#define equi(a, b) (!!(a) == !!(b))
+
+/* --------------------------------------------------------------------
+ * Light-weight trace
+ * Support for temporary event tracing with minimal Heisenberg effect.
+ * -------------------------------------------------------------------- */
+
+struct libcfs_device_userstate {
+       int        ldu_memhog_pages;
+       struct page   *ldu_memhog_root_page;
+};
+
+#define MKSTR(ptr) ((ptr)) ? (ptr) : ""
+
+static inline int cfs_size_round4(int val)
+{
+       return (val + 3) & (~0x3);
+}
+
+#ifndef HAVE_CFS_SIZE_ROUND
+static inline int cfs_size_round(int val)
+{
+       return (val + 7) & (~0x7);
+}
+
+#define HAVE_CFS_SIZE_ROUND
+#endif
+
+static inline int cfs_size_round16(int val)
+{
+       return (val + 0xf) & (~0xf);
+}
+
+static inline int cfs_size_round32(int val)
+{
+       return (val + 0x1f) & (~0x1f);
+}
+
+static inline int cfs_size_round0(int val)
+{
+       if (!val)
+               return 0;
+       return (val + 1 + 7) & (~0x7);
+}
+
+static inline size_t cfs_round_strlen(char *fset)
+{
+       return (size_t)cfs_size_round((int)strlen(fset) + 1);
+}
+
+/* roundup \a val to power2 */
+static inline unsigned int cfs_power2_roundup(unsigned int val)
+{
+       if (val != LOWEST_BIT_SET(val)) { /* not a power of 2 already */
+               do {
+                       val &= ~LOWEST_BIT_SET(val);
+               } while (val != LOWEST_BIT_SET(val));
+               /* ...and round up */
+               val <<= 1;
+       }
+       return val;
+}
+
+#define LOGL(var, len, ptr)                                   \
+do {                                                       \
+       if (var)                                                \
+               memcpy((char *)ptr, (const char *)var, len);    \
+       ptr += cfs_size_round(len);                          \
+} while (0)
+
+#define LOGU(var, len, ptr)                                   \
+do {                                                       \
+       if (var)                                                \
+               memcpy((char *)var, (const char *)ptr, len);    \
+       ptr += cfs_size_round(len);                          \
+} while (0)
+
+#define LOGL0(var, len, ptr)                         \
+do {                                               \
+       if (!len)                                      \
+               break;                            \
+       memcpy((char *)ptr, (const char *)var, len);    \
+       *((char *)(ptr) + len) = 0;                  \
+       ptr += cfs_size_round(len + 1);          \
+} while (0)
+
+/**
+ *  Lustre Network Driver types.
+ */
+enum {
+       /* Only add to these values (i.e. don't ever change or redefine them):
+        * network addresses depend on them... */
+       QSWLND    = 1,
+       SOCKLND   = 2,
+       GMLND     = 3, /* obsolete, keep it so that libcfs_nid2str works */
+       PTLLND    = 4,
+       O2IBLND   = 5,
+       CIBLND    = 6,
+       OPENIBLND = 7,
+       IIBLND    = 8,
+       LOLND     = 9,
+       RALND     = 10,
+       VIBLND    = 11,
+       MXLND     = 12,
+       GNILND    = 13,
+};
+
+#endif