perf: Support deferred user unwind

Add support for deferred userspace unwind to perf.

Where perf currently relies on in-place stack unwinding; from NMI
context and all that. This moves the userspace part of the unwind to
right before the return-to-userspace.

This has two distinct benefits, the biggest is that it moves the
unwind to a faultable context. It becomes possible to fault in debug
info (.eh_frame, SFrame etc.) that might not otherwise be readily
available. And secondly, it de-duplicates the user callchain where
multiple samples happen during the same kernel entry.

To facilitate this the perf interface is extended with a new record
type:

PERF_RECORD_CALLCHAIN_DEFERRED

and two new attribute flags:

perf_event_attr::defer_callchain - to request the user unwind be deferred
perf_event_attr::defer_output - to request PERF_RECORD_CALLCHAIN_DEFERRED records

The existing PERF_RECORD_SAMPLE callchain section gets a new
context type:

PERF_CONTEXT_USER_DEFERRED

After which will come a single entry, denoting the 'cookie' of the
deferred callchain that should be attached here, matching the 'cookie'
field of the above mentioned PERF_RECORD_CALLCHAIN_DEFERRED.

The 'defer_callchain' flag is expected on all events with
PERF_SAMPLE_CALLCHAIN. The 'defer_output' flag is expect on the event
responsible for collecting side-band events (like mmap, comm etc.).
Setting 'defer_output' on multiple events will get you duplicated
PERF_RECORD_CALLCHAIN_DEFERRED records.

Based on earlier patches by Josh and Steven.

Signed-off-by: Peter Zijlstra (Intel) <[e-mail]>
Link: https://patch.msgid.link/[e-mail]

include/linux/perf_event.h

···
 extern void perf_callchain_kernel(struct perf_callchain_entry_ctx *entry, struct pt_regs *regs);
 extern struct perf_callchain_entry *
 get_perf_callchain(struct pt_regs *regs, bool kernel, bool user,
-		   u32 max_stack, bool crosstask, bool add_mark);
+		   u32 max_stack, bool crosstask, bool add_mark, u64 defer_cookie);
 extern int get_callchain_buffers(int max_stack);
 extern void put_callchain_buffers(void);
 extern struct perf_callchain_entry *get_callchain_entry(int *rctx);

include/linux/unwind_deferred.h

···
 #include <linux/unwind_user.h>
 #include <linux/unwind_deferred_types.h>
 
-struct unwind_work;
-
-typedef void (*unwind_callback_t)(struct unwind_work *work,
-				  struct unwind_stacktrace *trace,
-				  u64 cookie);
-
-struct unwind_work {
-	struct list_head		list;
-	unwind_callback_t		func;
-	int				bit;
-};
-
 #ifdef CONFIG_UNWIND_USER
 
 enum {

include/linux/unwind_deferred_types.h

···
 	union unwind_task_id	id;
 };
 
+struct unwind_work;
+struct unwind_stacktrace;
+
+typedef void (*unwind_callback_t)(struct unwind_work *work,
+				  struct unwind_stacktrace *trace,
+				  u64 cookie);
+
+struct unwind_work {
+	struct list_head		list;
+	unwind_callback_t		func;
+	int				bit;
+};
+
 #endif /* _LINUX_UNWIND_USER_DEFERRED_TYPES_H */

include/uapi/linux/perf_event.h

···
 				inherit_thread :  1, /* children only inherit if cloned with CLONE_THREAD */
 				remove_on_exec :  1, /* event is removed from task on exec */
 				sigtrap        :  1, /* send synchronous SIGTRAP on event */
-				__reserved_1   : 26;
+				defer_callchain:  1, /* request PERF_RECORD_CALLCHAIN_DEFERRED records */
+				defer_output   :  1, /* output PERF_RECORD_CALLCHAIN_DEFERRED records */
+				__reserved_1   : 24;
 
 	union {
 		__u32		wakeup_events;	  /* wake up every n events */
···
 	 */
 	PERF_RECORD_AUX_OUTPUT_HW_ID		= 21,
 
+	/*
+	 * This user callchain capture was deferred until shortly before
+	 * returning to user space.  Previous samples would have kernel
+	 * callchains only and they need to be stitched with this to make full
+	 * callchains.
+	 *
+	 * struct {
+	 *	struct perf_event_header	header;
+	 *	u64				cookie;
+	 *	u64				nr;
+	 *	u64				ips[nr];
+	 *	struct sample_id		sample_id;
+	 * };
+	 */
+	PERF_RECORD_CALLCHAIN_DEFERRED		= 22,
+
 	PERF_RECORD_MAX,			/* non-ABI */
 };
 
···
 	PERF_CONTEXT_HV				= (__u64)-32,
 	PERF_CONTEXT_KERNEL			= (__u64)-128,
 	PERF_CONTEXT_USER			= (__u64)-512,
+	PERF_CONTEXT_USER_DEFERRED		= (__u64)-640,
 
 	PERF_CONTEXT_GUEST			= (__u64)-2048,
 	PERF_CONTEXT_GUEST_KERNEL		= (__u64)-2176,

kernel/bpf/stackmap.c

···
 		max_depth = sysctl_perf_event_max_stack;
 
 	trace = get_perf_callchain(regs, kernel, user, max_depth,
-				   false, false);
+				   false, false, 0);
 
 	if (unlikely(!trace))
 		/* couldn't fetch the stack trace */
···
 		trace = get_callchain_entry_for_task(task, max_depth);
 	else
 		trace = get_perf_callchain(regs, kernel, user, max_depth,
-					   crosstask, false);
+					   crosstask, false, 0);
 
 	if (unlikely(!trace) || trace->nr < skip) {
 		if (may_fault)

kernel/events/callchain.c

···
 
 struct perf_callchain_entry *
 get_perf_callchain(struct pt_regs *regs, bool kernel, bool user,
-		   u32 max_stack, bool crosstask, bool add_mark)
+		   u32 max_stack, bool crosstask, bool add_mark, u64 defer_cookie)
 {
 	struct perf_callchain_entry *entry;
 	struct perf_callchain_entry_ctx ctx;
···
 			if (current->flags & (PF_KTHREAD | PF_USER_WORKER))
 				goto exit_put;
 			regs = task_pt_regs(current);
+		}
+
+		if (defer_cookie) {
+			/*
+			 * Foretell the coming of PERF_RECORD_CALLCHAIN_DEFERRED
+			 * which can be stitched to this one, and add
+			 * the cookie after it (it will be cut off when the
+			 * user stack is copied to the callchain).
+			 */
+			perf_callchain_store_context(&ctx, PERF_CONTEXT_USER_DEFERRED);
+			perf_callchain_store_context(&ctx, defer_cookie);
+			goto exit_put;
 		}
 
 		if (add_mark)

kernel/events/core.c

···
 #include <linux/buildid.h>
 #include <linux/task_work.h>
 #include <linux/percpu-rwsem.h>
+#include <linux/unwind_deferred.h>
 
 #include "internal.h"
 
···
 
 static struct perf_callchain_entry __empty_callchain = { .nr = 0, };
 
+static struct unwind_work perf_unwind_work;
+
 struct perf_callchain_entry *
 perf_callchain(struct perf_event *event, struct pt_regs *regs)
 {
···
 		!(current->flags & (PF_KTHREAD | PF_USER_WORKER));
 	/* Disallow cross-task user callchains. */
 	bool crosstask = event->ctx->task && event->ctx->task != current;
+	bool defer_user = IS_ENABLED(CONFIG_UNWIND_USER) && user &&
+			  event->attr.defer_callchain;
 	const u32 max_stack = event->attr.sample_max_stack;
 	struct perf_callchain_entry *callchain;
+	u64 defer_cookie;
 
 	if (!current->mm)
 		user = false;
···
 	if (!kernel && !user)
 		return &__empty_callchain;
 
-	callchain = get_perf_callchain(regs, kernel, user,
-				       max_stack, crosstask, true);
+	if (!(user && defer_user && !crosstask &&
+	      unwind_deferred_request(&perf_unwind_work, &defer_cookie) >= 0))
+		defer_cookie = 0;
+
+	callchain = get_perf_callchain(regs, kernel, user, max_stack,
+				       crosstask, true, defer_cookie);
+
 	return callchain ?: &__empty_callchain;
 }
 
···
 
 	memcpy(bpf_event.event_id.tag, prog->tag, BPF_TAG_SIZE);
 	perf_iterate_sb(perf_event_bpf_output, &bpf_event, NULL);
+}
+
+struct perf_callchain_deferred_event {
+	struct unwind_stacktrace *trace;
+	struct {
+		struct perf_event_header	header;
+		u64				cookie;
+		u64				nr;
+		u64				ips[];
+	} event;
+};
+
+static void perf_callchain_deferred_output(struct perf_event *event, void *data)
+{
+	struct perf_callchain_deferred_event *deferred_event = data;
+	struct perf_output_handle handle;
+	struct perf_sample_data sample;
+	int ret, size = deferred_event->event.header.size;
+
+	if (!event->attr.defer_output)
+		return;
+
+	/* XXX do we really need sample_id_all for this ??? */
+	perf_event_header__init_id(&deferred_event->event.header, &sample, event);
+
+	ret = perf_output_begin(&handle, &sample, event,
+				deferred_event->event.header.size);
+	if (ret)
+		goto out;
+
+	perf_output_put(&handle, deferred_event->event);
+	for (int i = 0; i < deferred_event->trace->nr; i++) {
+		u64 entry = deferred_event->trace->entries[i];
+		perf_output_put(&handle, entry);
+	}
+	perf_event__output_id_sample(event, &handle, &sample);
+
+	perf_output_end(&handle);
+out:
+	deferred_event->event.header.size = size;
+}
+
+static void perf_unwind_deferred_callback(struct unwind_work *work,
+					 struct unwind_stacktrace *trace, u64 cookie)
+{
+	struct perf_callchain_deferred_event deferred_event = {
+		.trace = trace,
+		.event = {
+			.header = {
+				.type = PERF_RECORD_CALLCHAIN_DEFERRED,
+				.misc = PERF_RECORD_MISC_USER,
+				.size = sizeof(deferred_event.event) +
+					(trace->nr * sizeof(u64)),
+			},
+			.cookie = cookie,
+			.nr = trace->nr,
+		},
+	};
+
+	perf_iterate_sb(perf_callchain_deferred_output, &deferred_event, NULL);
 }
 
 struct perf_text_poke_event {
···
 	int ret;
 
 	idr_init(&pmu_idr);
+
+	unwind_deferred_init(&perf_unwind_work,
+			     perf_unwind_deferred_callback);
 
 	perf_event_init_all_cpus();
 	init_srcu_struct(&pmus_srcu);

tools/include/uapi/linux/perf_event.h

···
 				inherit_thread :  1, /* children only inherit if cloned with CLONE_THREAD */
 				remove_on_exec :  1, /* event is removed from task on exec */
 				sigtrap        :  1, /* send synchronous SIGTRAP on event */
-				__reserved_1   : 26;
+				defer_callchain:  1, /* request PERF_RECORD_CALLCHAIN_DEFERRED records */
+				defer_output   :  1, /* output PERF_RECORD_CALLCHAIN_DEFERRED records */
+				__reserved_1   : 24;
 
 	union {
 		__u32		wakeup_events;	  /* wake up every n events */
···
 	 */
 	PERF_RECORD_AUX_OUTPUT_HW_ID		= 21,
 
+	/*
+	 * This user callchain capture was deferred until shortly before
+	 * returning to user space.  Previous samples would have kernel
+	 * callchains only and they need to be stitched with this to make full
+	 * callchains.
+	 *
+	 * struct {
+	 *	struct perf_event_header	header;
+	 *	u64				cookie;
+	 *	u64				nr;
+	 *	u64				ips[nr];
+	 *	struct sample_id		sample_id;
+	 * };
+	 */
+	PERF_RECORD_CALLCHAIN_DEFERRED		= 22,
+
 	PERF_RECORD_MAX,			/* non-ABI */
 };
 
···
 	PERF_CONTEXT_HV				= (__u64)-32,
 	PERF_CONTEXT_KERNEL			= (__u64)-128,
 	PERF_CONTEXT_USER			= (__u64)-512,
+	PERF_CONTEXT_USER_DEFERRED		= (__u64)-640,
 
 	PERF_CONTEXT_GUEST			= (__u64)-2048,
 	PERF_CONTEXT_GUEST_KERNEL		= (__u64)-2176,