First commit of lustre style internal debug support. These

changes bring over everything lustre had for debugging with
two exceptions.  I dropped by the debug daemon and upcalls
just because it made things a little easier.  They can be
readded easily enough if we feel they are needed.

Everything compiles and seems to work on first inspection
but I suspect there are a handful of issues still lingering
which I'll be sorting out right away.  I just wanted to get
all these changes commited and safe.  I'm getting a little
paranoid about losing them.



git-svn-id: https://outreach.scidac.gov/svn/spl/trunk@75 7e1ea52c-4ff2-0310-8f11-9dd32ca42a1c

1 files changed, 310 insertions, 3 deletions

diff --git a/include/sys/debug.h b/include/sys/debug.h
index 4a6ed5c3d..a670c0374 100644
--- a/include/sys/debug.h
+++ b/include/sys/debug.h
@@ -1,11 +1,318 @@
 #ifndef _SPL_DEBUG_H
 #define _SPL_DEBUG_H
 
+#include <linux/sched.h> /* THREAD_SIZE */
+#include <linux/proc_fs.h>
+
 extern unsigned long spl_debug_mask;
 extern unsigned long spl_debug_subsys;
 
-void __dprintf(const char *file, const char *func, int line, const char *fmt, ...);
-void spl_set_debug_mask(unsigned long mask);
-void spl_set_debug_subsys(unsigned long mask);
+#define S_UNDEFINED   0x00000001
+#define S_ATOMIC      0x00000002
+#define S_KOBJ        0x00000004
+#define S_VNODE       0x00000008
+#define S_TIME        0x00000010
+#define S_RWLOCK      0x00000020
+#define S_THREAD      0x00000040
+#define S_CONDVAR     0x00000080
+#define S_MUTEX       0x00000100
+#define S_RNG         0x00000200
+#define S_TASKQ       0x00000400
+#define S_KMEM        0x00000800
+#define S_DEBUG       0x00001000
+#define S_GENERIC     0x00002000
+#define S_PROC        0x00004000
+
+#define D_TRACE       0x00000001
+#define D_INFO        0x00000002
+#define D_WARNING     0x00000004
+#define D_ERROR       0x00000008
+#define D_EMERG       0x00000010
+#define D_CONSOLE     0x00000020
+#define D_IOCTL       0x00000040
+#define D_DPRINTF     0x00000080
+#define D_OTHER       0x00000100
+
+#define D_CANTMASK    (D_ERROR | D_EMERG | D_WARNING | D_CONSOLE)
+#define DEBUG_SUBSYSTEM S_UNDEFINED
+
+int debug_init(void);
+void debug_fini(void);
+int spl_debug_mask2str(char *str, int size, unsigned long mask, int is_subsys);
+int spl_debug_str2mask(unsigned long *mask, const char *str, int is_subsys);
+
+extern unsigned long spl_debug_subsys;
+extern unsigned long spl_debug_mask;
+extern unsigned long spl_debug_printk;
+extern int spl_debug_mb;
+extern unsigned int spl_debug_binary;
+extern unsigned int spl_debug_catastrophe;
+extern unsigned int spl_debug_panic_on_bug;
+extern char spl_debug_file_path[PATH_MAX];
+extern unsigned int spl_console_ratelimit;
+extern long spl_console_max_delay;
+extern long spl_console_min_delay;
+extern unsigned int spl_console_backoff;
+extern unsigned int spl_debug_stack;
+
+#define TCD_MAX_PAGES			(5 << (20 - PAGE_SHIFT))
+#define TCD_STOCK_PAGES			(TCD_MAX_PAGES)
+#define TRACE_CONSOLE_BUFFER_SIZE	1024
+
+#define SPL_DEFAULT_MAX_DELAY		(600 * HZ)
+#define SPL_DEFAULT_MIN_DELAY		((HZ + 1) / 2)
+#define SPL_DEFAULT_BACKOFF		2
+
+typedef struct {
+        unsigned long cdls_next;
+        int           cdls_count;
+        long          cdls_delay;
+} spl_debug_limit_state_t;
+
+/* Three trace data types */
+typedef enum {
+        TCD_TYPE_PROC,
+        TCD_TYPE_SOFTIRQ,
+        TCD_TYPE_IRQ,
+        TCD_TYPE_MAX
+} tcd_type_t;
+
+union trace_data_union {
+	struct trace_cpu_data {
+		/* pages with trace records not yet processed by tracefiled */
+		struct list_head        tcd_pages;
+		/* number of pages on ->tcd_pages */
+		unsigned long           tcd_cur_pages;
+		/* Max number of pages allowed on ->tcd_pages */
+		unsigned long           tcd_max_pages;
+
+		/*
+		 * preallocated pages to write trace records into. Pages from
+		 * ->tcd_stock_pages are moved to ->tcd_pages by spl_debug_msg().
+		 *
+		 * This list is necessary, because on some platforms it's
+		 * impossible to perform efficient atomic page allocation in a
+		 * non-blockable context.
+		 *
+		 * Such platforms fill ->tcd_stock_pages "on occasion", when
+		 * tracing code is entered in blockable context.
+		 *
+		 * trace_get_tage_try() tries to get a page from
+		 * ->tcd_stock_pages first and resorts to atomic page
+		 * allocation only if this queue is empty. ->tcd_stock_pages
+		 * is replenished when tracing code is entered in blocking
+		 * context (darwin-tracefile.c:trace_get_tcd()). We try to
+		 * maintain TCD_STOCK_PAGES (40 by default) pages in this
+		 * queue. Atomic allocation is only required if more than
+		 * TCD_STOCK_PAGES pagesful are consumed by trace records all
+		 * emitted in non-blocking contexts. Which is quite unlikely.
+		 */
+		struct list_head        tcd_stock_pages;
+		/* number of pages on ->tcd_stock_pages */
+		unsigned long           tcd_cur_stock_pages;
+
+		unsigned short          tcd_shutting_down;
+		unsigned short          tcd_cpu;
+		unsigned short          tcd_type;
+		/* The factors to share debug memory. */
+		unsigned short          tcd_pages_factor;
+	} tcd;
+	char __pad[L1_CACHE_ALIGN(sizeof(struct trace_cpu_data))];
+};
+
+extern union trace_data_union (*trace_data[TCD_TYPE_MAX])[NR_CPUS];
+
+#define tcd_for_each(tcd, i, j)                                       \
+    for (i = 0; trace_data[i] != NULL; i++)                           \
+        for (j = 0, ((tcd) = &(*trace_data[i])[j].tcd);               \
+             j < num_possible_cpus(); j++, (tcd) = &(*trace_data[i])[j].tcd)
+
+#define tcd_for_each_type_lock(tcd, i)                                \
+    for (i = 0; trace_data[i] &&                                      \
+         (tcd = &(*trace_data[i])[smp_processor_id()].tcd) &&         \
+         trace_lock_tcd(tcd); trace_unlock_tcd(tcd), i++)
+
+struct trace_page {
+	struct page *    page;    /* page itself */
+	struct list_head linkage;  /* Used by lists in trace_data_union */
+	unsigned int     used;     /* number of bytes used within this page */
+	unsigned short   cpu;      /* cpu that owns this page */
+	unsigned short   type;     /* type(context) of this page */
+};
+
+struct page_collection {
+        struct list_head  pc_pages;
+        spinlock_t        pc_lock;
+        int               pc_want_daemon_pages;
+};
+
+/* ASSERTION that is safe to use within the debug system */
+#define __ASSERT(cond)							\
+do {									\
+	if (unlikely(!(cond))) {					\
+                printk(KERN_ERR "ASSERTION("#cond") failed");           \
+		BUG();                                                  \
+	}								\
+} while (0)
+
+#define __ASSERT_TAGE_INVARIANT(tage)					\
+do {									\
+        __ASSERT(tage != NULL);						\
+        __ASSERT(tage->page != NULL);					\
+        __ASSERT(tage->used <= PAGE_SIZE);				\
+        __ASSERT(page_count(tage->page) > 0);				\
+} while(0)
+
+/* ASSERTION that will debug log used outside the debug sysytem */
+#define ASSERT(cond)                                                    \
+do {                                                                    \
+        if (unlikely(!(cond))) {                                        \
+                spl_debug_msg(NULL, DEBUG_SUBSYSTEM, D_EMERG,           \
+                              __FILE__, __FUNCTION__, __LINE__,         \
+                              "ASSERTION(" #cond ") failed\n");         \
+                spl_debug_bug(__FILE__, __FUNCTION__, __LINE__);        \
+	}                                                               \
+} while (0)
+
+#define ASSERTF(cond, fmt, a...)                                        \
+do {                                                                    \
+        if (unlikely(!(cond))) {                                        \
+                spl_debug_msg(NULL, DEBUG_SUBSYSTEM, D_EMERG,           \
+                              __FILE__, __FUNCTION__, __LINE__,         \
+                              "ASSERTION(" #cond ") failed:" fmt,       \
+                                 ## a);                                 \
+                spl_debug_bug(__FILE__, __FUNCTION__, __LINE__)         \
+        }                                                               \
+} while (0)
+
+#define VERIFY3_IMPL(LEFT, OP, RIGHT, TYPE, FMT, CAST)                  \
+do {                                                                    \
+        const TYPE __left = (TYPE)(LEFT);                               \
+        const TYPE __right = (TYPE)(RIGHT);                             \
+        if (!(__left OP __right)) {                                     \
+                spl_debug_msg(NULL, DEBUG_SUBSYSTEM, D_EMERG,           \
+                              __FILE__, __FUNCTION__, __LINE__,         \
+                              "VERIFY3(" FMT " " #OP " " FMT ")\n",     \
+                              CAST __left,  CAST __right);              \
+                spl_debug_bug(__FILE__, __FUNCTION__, __LINE__)         \
+        }                                                               \
+} while (0)
+
+#define VERIFY3S(x,y,z) VERIFY3_IMPL(x, y, z, int64_t, "%ld", (long))
+#define VERIFY3U(x,y,z) VERIFY3_IMPL(x, y, z, uint64_t, "%lu", (unsigned long))
+#define VERIFY3P(x,y,z) VERIFY3_IMPL(x, y, z, uintptr_t, "%p", (void *))
+
+#define ASSERT3S(x,y,z) VERIFY3S(x, y, z)
+#define ASSERT3U(x,y,z) VERIFY3U(x, y, z)
+#define ASSERT3P(x,y,z) VERIFY3P(x, y, z)
+
+#define VERIFY(x)	ASSERT(x)
+#define SBUG()		spl_debug_bug(__FILE__, __FUNCTION__, __LINE__);
+
+#define spl_debug_msg(cdls, subsys, mask, file, fn, line, format, a...) \
+        spl_debug_vmsg(cdls, subsys, mask, file, fn,                    \
+                       line, NULL, NULL, format, ##a)
+
+#ifdef  __ia64__
+#define CDEBUG_STACK() (THREAD_SIZE -                                   \
+                       ((unsigned long)__builtin_dwarf_cfa() &          \
+                       (THREAD_SIZE - 1)))
+#else
+#define CDEBUG_STACK() (THREAD_SIZE -                                   \
+                       ((unsigned long)__builtin_frame_address(0) &     \
+                        (THREAD_SIZE - 1)))
+# endif /* __ia64__ */
+
+#define __CHECK_STACK(file, func, line)                                 \
+do {                                                                    \
+        unsigned long _stack = CDEBUG_STACK();                          \
+                                                                        \
+        if (_stack > (3*THREAD_SIZE/4) && _stack > spl_debug_stack) {   \
+                spl_debug_stack = _stack;                               \
+                spl_debug_msg(NULL, D_TRACE, D_WARNING,                 \
+                              file, func, line,                         \
+                              "Exceeded maximum safe stack "            \
+                              "%lu/%lu\n", _stack, THREAD_SIZE);        \
+                __ASSERT(0);                                            \
+        }                                                               \
+} while (0)
+
+#define CHECK_STACK()__CHECK_STACK(__FILE__, __func__, __LINE__)
+
+#define __CDEBUG(cdls, subsys, mask, format, a...)                      \
+do {                                                                    \
+        CHECK_STACK();                                                  \
+                                                                        \
+        if (((mask) & D_CANTMASK) != 0 ||                               \
+            ((spl_debug_mask & (mask)) != 0 &&                          \
+             (spl_debug_subsys & (subsys)) != 0))                       \
+                spl_debug_msg(cdls, subsys, mask,                       \
+                              __FILE__, __FUNCTION__, __LINE__,         \
+                              format, ## a);                            \
+} while (0)
+
+#define CDEBUG(mask, format, a...)                                      \
+        __CDEBUG(NULL, DEBUG_SUBSYSTEM, mask, format, ## a)
+
+#define __CDEBUG_LIMIT(subsys, mask, format, a...)                      \
+do {                                                                    \
+        static spl_debug_limit_state_t cdls;                            \
+                                                                        \
+        __CDEBUG(&cdls, subsys, mask, format, ## a);                    \
+} while (0)
+
+#define CDEBUG_LIMIT(mask, format, a...)                                \
+        __CDEBUG_LIMIT(DEBUG_SUBSYSTEM, mask, format, ## a)
+
+#define dprintf(fmt, a...)             CDEBUG_LIMIT(D_INFO, fmt, ## a)
+#define CWARN(fmt, a...)               CDEBUG_LIMIT(D_WARNING, fmt, ## a)
+#define CERROR(fmt, a...)              CDEBUG_LIMIT(D_ERROR, fmt, ## a)
+#define CEMERG(fmt, a...)              CDEBUG_LIMIT(D_EMERG, fmt, ## a)
+#define CONSOLE(mask, fmt, a...)       CDEBUG(D_CONSOLE | (mask), fmt, ## a)
+
+#define GOTO(label, rc)                                                 \
+do {                                                                    \
+        long GOTO__ret = (long)(rc);                                    \
+        CDEBUG(D_TRACE,"Process leaving via %s (rc=%lu : %ld : %lx)\n", \
+               #label, (unsigned long)GOTO__ret, (signed long)GOTO__ret,\
+               (signed long)GOTO__ret);                                 \
+        goto label;                                                     \
+} while (0)
+
+#define RETURN(rc)                                                      \
+do {                                                                    \
+        typeof(rc) RETURN__ret = (rc);                                  \
+        CDEBUG(D_TRACE, "Process leaving (rc=%lu : %ld : %lx)\n",       \
+               (long)RETURN__ret, (long)RETURN__ret, (long)RETURN__ret);\
+        return RETURN__ret;                                             \
+} while (0)
+
+#define ENTRY                                                           \
+do {                                                                    \
+        CDEBUG(D_TRACE, "Process entered\n");                           \
+} while (0)
+
+#define EXIT                                                            \
+do {                                                                    \
+        CDEBUG(D_TRACE, "Process leaving\n");                           \
+} while(0)
+
+extern int spl_debug_vmsg(spl_debug_limit_state_t *cdls, int subsys, int mask,
+	                  const char *file, const char *fn, const int line,
+			  const char *format1, va_list args, const char *format2, ...);
+
+extern unsigned long spl_debug_set_mask(unsigned long mask);
+extern unsigned long spl_debug_get_mask(void);
+extern unsigned long spl_debug_set_subsys(unsigned long mask);
+extern unsigned long spl_debug_get_subsys(void);
+extern int spl_debug_set_mb(int mb);
+extern int spl_debug_get_mb(void);
+
+extern int spl_debug_dumplog(void);
+extern void spl_debug_dumpstack(struct task_struct *tsk);
+extern void spl_debug_bug(char *file, const char *func, const int line);
+
+extern int spl_debug_clear_buffer(void);
+extern int spl_debug_mark_buffer(char *text);
 
 #endif /* SPL_DEBUG_H */