x86, bts: provide in-kernel branch-trace interface

Impact: cleanup

Move the BTS bits from ptrace.c into ds.c.

Signed-off-by: Markus Metzger <markus.t.metzger@intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

1 files changed, 105 insertions, 311 deletions

diff --git a/arch/x86/kernel/ptrace.c b/arch/x86/kernel/ptrace.c
index b2998fe1166b..45e9855da2d2 100644
--- a/arch/x86/kernel/ptrace.c
+++ b/arch/x86/kernel/ptrace.c
@@ -581,153 +581,73 @@ static int ioperm_get(struct task_struct *target,
 }
 
 #ifdef CONFIG_X86_PTRACE_BTS
-/*
- * The configuration for a particular BTS hardware implementation.
- */
-struct bts_configuration {
-	/* the size of a BTS record in bytes; at most BTS_MAX_RECORD_SIZE */
-	unsigned char  sizeof_bts;
-	/* the size of a field in the BTS record in bytes */
-	unsigned char  sizeof_field;
-	/* a bitmask to enable/disable BTS in DEBUGCTL MSR */
-	unsigned long debugctl_mask;
-};
-static struct bts_configuration bts_cfg;
-
-#define BTS_MAX_RECORD_SIZE (8 * 3)
-
-
-/*
- * Branch Trace Store (BTS) uses the following format. Different
- * architectures vary in the size of those fields.
- * - source linear address
- * - destination linear address
- * - flags
- *
- * Later architectures use 64bit pointers throughout, whereas earlier
- * architectures use 32bit pointers in 32bit mode.
- *
- * We compute the base address for the first 8 fields based on:
- * - the field size stored in the DS configuration
- * - the relative field position
- *
- * In order to store additional information in the BTS buffer, we use
- * a special source address to indicate that the record requires
- * special interpretation.
- *
- * Netburst indicated via a bit in the flags field whether the branch
- * was predicted; this is ignored.
- */
-
-enum bts_field {
-	bts_from = 0,
-	bts_to,
-	bts_flags,
-
-	bts_escape = (unsigned long)-1,
-	bts_qual = bts_to,
-	bts_jiffies = bts_flags
-};
-
-static inline unsigned long bts_get(const char *base, enum bts_field field)
-{
-	base += (bts_cfg.sizeof_field * field);
-	return *(unsigned long *)base;
-}
-
-static inline void bts_set(char *base, enum bts_field field, unsigned long val)
-{
-	base += (bts_cfg.sizeof_field * field);;
-	(*(unsigned long *)base) = val;
-}
-
-/*
- * Translate a BTS record from the raw format into the bts_struct format
- *
- * out (out): bts_struct interpretation
- * raw: raw BTS record
- */
-static void ptrace_bts_translate_record(struct bts_struct *out, const void *raw)
-{
-	memset(out, 0, sizeof(*out));
-	if (bts_get(raw, bts_from) == bts_escape) {
-		out->qualifier       = bts_get(raw, bts_qual);
-		out->variant.jiffies = bts_get(raw, bts_jiffies);
-	} else {
-		out->qualifier = BTS_BRANCH;
-		out->variant.lbr.from_ip = bts_get(raw, bts_from);
-		out->variant.lbr.to_ip   = bts_get(raw, bts_to);
-	}
-}
-
 static int ptrace_bts_read_record(struct task_struct *child, size_t index,
 				  struct bts_struct __user *out)
 {
-	struct bts_struct ret;
-	const void *bts_record;
-	size_t bts_index, bts_end;
+	const struct bts_trace *trace;
+	struct bts_struct bts;
+	const unsigned char *at;
 	int error;
 
-	error = ds_get_bts_end(child->bts, &bts_end);
-	if (error < 0)
-		return error;
-
-	if (bts_end <= index)
-		return -EINVAL;
+	trace = ds_read_bts(child->bts);
+	if (!trace)
+		return -EPERM;
 
-	error = ds_get_bts_index(child->bts, &bts_index);
-	if (error < 0)
-		return error;
+	at = trace->ds.top - ((index + 1) * trace->ds.size);
+	if ((void *)at < trace->ds.begin)
+		at += (trace->ds.n * trace->ds.size);
 
-	/* translate the ptrace bts index into the ds bts index */
-	bts_index += bts_end - (index + 1);
-	if (bts_end <= bts_index)
-		bts_index -= bts_end;
+	if (!trace->read)
+		return -EOPNOTSUPP;
 
-	error = ds_access_bts(child->bts, bts_index, &bts_record);
+	error = trace->read(child->bts, at, &bts);
 	if (error < 0)
 		return error;
 
-	ptrace_bts_translate_record(&ret, bts_record);
-
-	if (copy_to_user(out, &ret, sizeof(ret)))
+	if (copy_to_user(out, &bts, sizeof(bts)))
 		return -EFAULT;
 
-	return sizeof(ret);
+	return sizeof(bts);
 }
 
 static int ptrace_bts_drain(struct task_struct *child,
 			    long size,
 			    struct bts_struct __user *out)
 {
-	struct bts_struct ret;
-	const unsigned char *raw;
-	size_t end, i;
-	int error;
+	const struct bts_trace *trace;
+	const unsigned char *at;
+	int error, drained = 0;
 
-	error = ds_get_bts_index(child->bts, &end);
-	if (error < 0)
-		return error;
+	trace = ds_read_bts(child->bts);
+	if (!trace)
+		return -EPERM;
 
-	if (size < (end * sizeof(struct bts_struct)))
+	if (!trace->read)
+		return -EOPNOTSUPP;
+
+	if (size < (trace->ds.top - trace->ds.begin))
 		return -EIO;
 
-	error = ds_access_bts(child->bts, 0, (const void **)&raw);
-	if (error < 0)
-		return error;
+	for (at = trace->ds.begin; (void *)at < trace->ds.top;
+	     out++, drained++, at += trace->ds.size) {
+		struct bts_struct bts;
+		int error;
 
-	for (i = 0; i < end; i++, out++, raw += bts_cfg.sizeof_bts) {
-		ptrace_bts_translate_record(&ret, raw);
+		error = trace->read(child->bts, at, &bts);
+		if (error < 0)
+			return error;
 
-		if (copy_to_user(out, &ret, sizeof(ret)))
+		if (copy_to_user(out, &bts, sizeof(bts)))
 			return -EFAULT;
 	}
 
-	error = ds_clear_bts(child->bts);
+	memset(trace->ds.begin, 0, trace->ds.n * trace->ds.size);
+
+	error = ds_reset_bts(child->bts);
 	if (error < 0)
 		return error;
 
-	return end;
+	return drained;
 }
 
 static int ptrace_bts_config(struct task_struct *child,
@@ -735,136 +655,89 @@ static int ptrace_bts_config(struct task_struct *child,
 			     const struct ptrace_bts_config __user *ucfg)
 {
 	struct ptrace_bts_config cfg;
-	int error = 0;
-
-	error = -EOPNOTSUPP;
-	if (!bts_cfg.sizeof_bts)
-		goto errout;
+	unsigned int flags = 0;
 
-	error = -EIO;
 	if (cfg_size < sizeof(cfg))
-		goto errout;
+		return -EIO;
 
-	error = -EFAULT;
 	if (copy_from_user(&cfg, ucfg, sizeof(cfg)))
-		goto errout;
-
-	error = -EINVAL;
-	if ((cfg.flags & PTRACE_BTS_O_SIGNAL) &&
-	    !(cfg.flags & PTRACE_BTS_O_ALLOC))
-		goto errout;
-
-	if (cfg.flags & PTRACE_BTS_O_ALLOC) {
-		bts_ovfl_callback_t ovfl = NULL;
-		unsigned int sig = 0;
-
-		error = -EINVAL;
-		if (cfg.size < (10 * bts_cfg.sizeof_bts))
-			goto errout;
+		return -EFAULT;
 
-		if (cfg.flags & PTRACE_BTS_O_SIGNAL) {
-			if (!cfg.signal)
-				goto errout;
+	if (child->bts) {
+		ds_release_bts(child->bts);
+		child->bts = NULL;
+	}
 
-			error = -EOPNOTSUPP;
-			goto errout;
+	if (cfg.flags & PTRACE_BTS_O_SIGNAL) {
+		if (!cfg.signal)
+			return -EINVAL;
 
-			sig  = cfg.signal;
-		}
+		return -EOPNOTSUPP;
 
-		if (child->bts) {
-			(void)ds_release_bts(child->bts);
-			kfree(child->bts_buffer);
+		child->thread.bts_ovfl_signal = cfg.signal;
+	}
 
-			child->bts = NULL;
-			child->bts_buffer = NULL;
-		}
+	if ((cfg.flags & PTRACE_BTS_O_ALLOC) &&
+	    (cfg.size != child->bts_size)) {
+		kfree(child->bts_buffer);
 
-		error = -ENOMEM;
+		child->bts_size = cfg.size;
 		child->bts_buffer = kzalloc(cfg.size, GFP_KERNEL);
-		if (!child->bts_buffer)
-			goto errout;
-
-		child->bts = ds_request_bts(child, child->bts_buffer, cfg.size,
-					    ovfl, /* th = */ (size_t)-1);
-		if (IS_ERR(child->bts)) {
-			error = PTR_ERR(child->bts);
-			kfree(child->bts_buffer);
-			child->bts = NULL;
-			child->bts_buffer = NULL;
-			goto errout;
+		if (!child->bts_buffer) {
+			child->bts_size = 0;
+			return -ENOMEM;
 		}
-
-		child->thread.bts_ovfl_signal = sig;
 	}
 
-	error = -EINVAL;
-	if (!child->thread.ds_ctx && cfg.flags)
-		goto errout;
-
 	if (cfg.flags & PTRACE_BTS_O_TRACE)
-		child->thread.debugctlmsr |= bts_cfg.debugctl_mask;
-	else
-		child->thread.debugctlmsr &= ~bts_cfg.debugctl_mask;
+		flags |= BTS_USER;
 
 	if (cfg.flags & PTRACE_BTS_O_SCHED)
-		set_tsk_thread_flag(child, TIF_BTS_TRACE_TS);
-	else
-		clear_tsk_thread_flag(child, TIF_BTS_TRACE_TS);
+		flags |= BTS_TIMESTAMPS;
 
-	error = sizeof(cfg);
+	child->bts = ds_request_bts(child, child->bts_buffer, child->bts_size,
+				    /* ovfl = */ NULL, /* th = */ (size_t)-1,
+				    flags);
+	if (IS_ERR(child->bts)) {
+		int error = PTR_ERR(child->bts);
 
-out:
-	if (child->thread.debugctlmsr)
-		set_tsk_thread_flag(child, TIF_DEBUGCTLMSR);
-	else
-		clear_tsk_thread_flag(child, TIF_DEBUGCTLMSR);
+		kfree(child->bts_buffer);
+		child->bts = NULL;
+		child->bts_buffer = NULL;
+		child->bts_size = 0;
 
-	return error;
+		return error;
+	}
 
-errout:
-	child->thread.debugctlmsr &= ~bts_cfg.debugctl_mask;
-	clear_tsk_thread_flag(child, TIF_BTS_TRACE_TS);
-	goto out;
+	return sizeof(cfg);
 }
 
 static int ptrace_bts_status(struct task_struct *child,
 			     long cfg_size,
 			     struct ptrace_bts_config __user *ucfg)
 {
+	const struct bts_trace *trace;
 	struct ptrace_bts_config cfg;
-	size_t end;
-	const void *base, *max;
-	int error;
 
 	if (cfg_size < sizeof(cfg))
 		return -EIO;
 
-	error = ds_get_bts_end(child->bts, &end);
-	if (error < 0)
-		return error;
-
-	error = ds_access_bts(child->bts, /* index = */ 0, &base);
-	if (error < 0)
-		return error;
-
-	error = ds_access_bts(child->bts, /* index = */ end, &max);
-	if (error < 0)
-		return error;
+	trace = ds_read_bts(child->bts);
+	if (!trace)
+		return -EPERM;
 
 	memset(&cfg, 0, sizeof(cfg));
-	cfg.size = (max - base);
+	cfg.size = trace->ds.end - trace->ds.begin;
 	cfg.signal = child->thread.bts_ovfl_signal;
 	cfg.bts_size = sizeof(struct bts_struct);
 
 	if (cfg.signal)
 		cfg.flags |= PTRACE_BTS_O_SIGNAL;
 
-	if (test_tsk_thread_flag(child, TIF_DEBUGCTLMSR) &&
-	    child->thread.debugctlmsr & bts_cfg.debugctl_mask)
+	if (trace->ds.flags & BTS_USER)
 		cfg.flags |= PTRACE_BTS_O_TRACE;
 
-	if (test_tsk_thread_flag(child, TIF_BTS_TRACE_TS))
+	if (trace->ds.flags & BTS_TIMESTAMPS)
 		cfg.flags |= PTRACE_BTS_O_SCHED;
 
 	if (copy_to_user(ucfg, &cfg, sizeof(cfg)))
@@ -873,105 +746,28 @@ static int ptrace_bts_status(struct task_struct *child,
 	return sizeof(cfg);
 }
 
-static int ptrace_bts_write_record(struct task_struct *child,
-				   const struct bts_struct *in)
+static int ptrace_bts_clear(struct task_struct *child)
 {
-	unsigned char bts_record[BTS_MAX_RECORD_SIZE];
+	const struct bts_trace *trace;
 
-	if (BTS_MAX_RECORD_SIZE < bts_cfg.sizeof_bts)
-		return -EOVERFLOW;
+	trace = ds_read_bts(child->bts);
+	if (!trace)
+		return -EPERM;
 
-	memset(bts_record, 0, bts_cfg.sizeof_bts);
-	switch (in->qualifier) {
-	case BTS_INVALID:
-		break;
+	memset(trace->ds.begin, 0, trace->ds.n * trace->ds.size);
 
-	case BTS_BRANCH:
-		bts_set(bts_record, bts_from, in->variant.lbr.from_ip);
-		bts_set(bts_record, bts_to,   in->variant.lbr.to_ip);
-		break;
-
-	case BTS_TASK_ARRIVES:
-	case BTS_TASK_DEPARTS:
-		bts_set(bts_record, bts_from,    bts_escape);
-		bts_set(bts_record, bts_qual,    in->qualifier);
-		bts_set(bts_record, bts_jiffies, in->variant.jiffies);
-		break;
-
-	default:
-		return -EINVAL;
-	}
-
-	return ds_write_bts(child->bts, bts_record, bts_cfg.sizeof_bts);
+	return ds_reset_bts(child->bts);
 }
 
-void ptrace_bts_take_timestamp(struct task_struct *tsk,
-			       enum bts_qualifier qualifier)
+static int ptrace_bts_size(struct task_struct *child)
 {
-	struct bts_struct rec = {
-		.qualifier = qualifier,
-		.variant.jiffies = jiffies_64
-	};
-
-	ptrace_bts_write_record(tsk, &rec);
-}
-
-static const struct bts_configuration bts_cfg_netburst = {
-	.sizeof_bts    = sizeof(long) * 3,
-	.sizeof_field  = sizeof(long),
-	.debugctl_mask = (1<<2)|(1<<3)|(1<<5)
-};
+	const struct bts_trace *trace;
 
-static const struct bts_configuration bts_cfg_pentium_m = {
-	.sizeof_bts    = sizeof(long) * 3,
-	.sizeof_field  = sizeof(long),
-	.debugctl_mask = (1<<6)|(1<<7)
-};
+	trace = ds_read_bts(child->bts);
+	if (!trace)
+		return -EPERM;
 
-static const struct bts_configuration bts_cfg_core2 = {
-	.sizeof_bts    = 8 * 3,
-	.sizeof_field  = 8,
-	.debugctl_mask = (1<<6)|(1<<7)|(1<<9)
-};
-
-static inline void bts_configure(const struct bts_configuration *cfg)
-{
-	bts_cfg = *cfg;
-}
-
-void __cpuinit ptrace_bts_init_intel(struct cpuinfo_x86 *c)
-{
-	switch (c->x86) {
-	case 0x6:
-		switch (c->x86_model) {
-		case 0 ... 0xC:
-			/* sorry, don't know about them */
-			break;
-		case 0xD:
-		case 0xE: /* Pentium M */
-			bts_configure(&bts_cfg_pentium_m);
-			break;
-		default: /* Core2, Atom, ... */
-			bts_configure(&bts_cfg_core2);
-			break;
-		}
-		break;
-	case 0xF:
-		switch (c->x86_model) {
-		case 0x0:
-		case 0x1:
-		case 0x2: /* Netburst */
-			bts_configure(&bts_cfg_netburst);
-			break;
-		default:
-			/* sorry, don't know about them */
-			break;
-		}
-		break;
-	default:
-		/* sorry, don't know about them */
-		break;
-	}
+	return (trace->ds.top - trace->ds.begin) / trace->ds.size;
 }
 #endif /* CONFIG_X86_PTRACE_BTS */
 
@@ -988,15 +784,12 @@ void ptrace_disable(struct task_struct *child)
 #endif
 #ifdef CONFIG_X86_PTRACE_BTS
 	if (child->bts) {
-		(void)ds_release_bts(child->bts);
+		ds_release_bts(child->bts);
+		child->bts = NULL;
+
 		kfree(child->bts_buffer);
 		child->bts_buffer = NULL;
-
-		child->thread.debugctlmsr &= ~bts_cfg.debugctl_mask;
-		if (!child->thread.debugctlmsr)
-			clear_tsk_thread_flag(child, TIF_DEBUGCTLMSR);
-
-		clear_tsk_thread_flag(child, TIF_BTS_TRACE_TS);
+		child->bts_size = 0;
 	}
 #endif /* CONFIG_X86_PTRACE_BTS */
 }
@@ -1129,16 +922,9 @@ long arch_ptrace(struct task_struct *child, long request, long addr, long data)
 			(child, data, (struct ptrace_bts_config __user *)addr);
 		break;
 
-	case PTRACE_BTS_SIZE: {
-		size_t size;
-
-		ret = ds_get_bts_index(child->bts, &size);
-		if (ret == 0) {
-			WARN_ON_ONCE(size != (int) size);
-			ret = (int) size;
-		}
+	case PTRACE_BTS_SIZE:
+		ret = ptrace_bts_size(child);
 		break;
-	}
 
 	case PTRACE_BTS_GET:
 		ret = ptrace_bts_read_record
@@ -1146,7 +932,7 @@ long arch_ptrace(struct task_struct *child, long request, long addr, long data)
 		break;
 
 	case PTRACE_BTS_CLEAR:
-		ret = ds_clear_bts(child->bts);
+		ret = ptrace_bts_clear(child);
 		break;
 
 	case PTRACE_BTS_DRAIN:
@@ -1409,6 +1195,14 @@ long compat_arch_ptrace(struct task_struct *child, compat_long_t request,
 
 	case PTRACE_GET_THREAD_AREA:
 	case PTRACE_SET_THREAD_AREA:
+#ifdef CONFIG_X86_PTRACE_BTS
+	case PTRACE_BTS_CONFIG:
+	case PTRACE_BTS_STATUS:
+	case PTRACE_BTS_SIZE:
+	case PTRACE_BTS_GET:
+	case PTRACE_BTS_CLEAR:
+	case PTRACE_BTS_DRAIN:
+#endif /* CONFIG_X86_PTRACE_BTS */
 		return arch_ptrace(child, request, addr, data);
 
 	default: