bpf: Add kernel module with user mode driver that populates bpffs.

Add kernel module with user mode driver that populates bpffs with
BPF iterators.

$ mount bpffs /my/bpffs/ -t bpf
$ ls -la /my/bpffs/
total 4
drwxrwxrwt  2 root root    0 Jul  2 00:27 .
drwxr-xr-x 19 root root 4096 Jul  2 00:09 ..
-rw-------  1 root root    0 Jul  2 00:27 maps.debug
-rw-------  1 root root    0 Jul  2 00:27 progs.debug

The user mode driver will load BPF Type Formats, create BPF maps, populate BPF
maps, load two BPF programs, attach them to BPF iterators, and finally send two
bpf_link IDs back to the kernel.
The kernel will pin two bpf_links into newly mounted bpffs instance under
names "progs.debug" and "maps.debug". These two files become human readable.

$ cat /my/bpffs/progs.debug
  id name            attached
  11 dump_bpf_map    bpf_iter_bpf_map
  12 dump_bpf_prog   bpf_iter_bpf_prog
  27 test_pkt_access
  32 test_main       test_pkt_access test_pkt_access
  33 test_subprog1   test_pkt_access_subprog1 test_pkt_access
  34 test_subprog2   test_pkt_access_subprog2 test_pkt_access
  35 test_subprog3   test_pkt_access_subprog3 test_pkt_access
  36 new_get_skb_len get_skb_len test_pkt_access
  37 new_get_skb_ifindex get_skb_ifindex test_pkt_access
  38 new_get_constant get_constant test_pkt_access

The BPF program dump_bpf_prog() in iterators.bpf.c is printing this data about
all BPF programs currently loaded in the system. This information is unstable
and will change from kernel to kernel as ".debug" suffix conveys.

Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20200819042759.51280-4-alexei.starovoitov@gmail.com

kernel/bpf/preload/Kconfig

@@ -0,0 +1,23 @@
+# SPDX-License-Identifier: GPL-2.0-only
+config USERMODE_DRIVER
+	bool
+	default n
+
+menuconfig BPF_PRELOAD
+	bool "Preload BPF file system with kernel specific program and map iterators"
+	depends on BPF
+	select USERMODE_DRIVER
+	help
+	  This builds kernel module with several embedded BPF programs that are
+	  pinned into BPF FS mount point as human readable files that are
+	  useful in debugging and introspection of BPF programs and maps.
+
+if BPF_PRELOAD
+config BPF_PRELOAD_UMD
+	tristate "bpf_preload kernel module with user mode driver"
+	depends on CC_CAN_LINK
+	depends on m || CC_CAN_LINK_STATIC
+	default m
+	help
+	  This builds bpf_preload kernel module with embedded user mode driver.
+endif

kernel/bpf/preload/Makefile

@@ -0,0 +1,23 @@
+# SPDX-License-Identifier: GPL-2.0
+
+LIBBPF_SRCS = $(srctree)/tools/lib/bpf/
+LIBBPF_A = $(obj)/libbpf.a
+LIBBPF_OUT = $(abspath $(obj))
+
+$(LIBBPF_A):
+	$(Q)$(MAKE) -C $(LIBBPF_SRCS) OUTPUT=$(LIBBPF_OUT)/ $(LIBBPF_OUT)/libbpf.a
+
+userccflags += -I $(srctree)/tools/include/ -I $(srctree)/tools/include/uapi \
+	-I $(srctree)/tools/lib/ -Wno-unused-result
+
+userprogs := bpf_preload_umd
+
+bpf_preload_umd-objs := iterators/iterators.o
+bpf_preload_umd-userldlibs := $(LIBBPF_A) -lelf -lz
+
+$(obj)/bpf_preload_umd: $(LIBBPF_A)
+
+$(obj)/bpf_preload_umd_blob.o: $(obj)/bpf_preload_umd
+
+obj-$(CONFIG_BPF_PRELOAD_UMD) += bpf_preload.o
+bpf_preload-objs += bpf_preload_kern.o bpf_preload_umd_blob.o

kernel/bpf/preload/bpf_preload.h

@@ -0,0 +1,16 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _BPF_PRELOAD_H
+#define _BPF_PRELOAD_H
+
+#include <linux/usermode_driver.h>
+#include "iterators/bpf_preload_common.h"
+
+struct bpf_preload_ops {
+        struct umd_info info;
+	int (*preload)(struct bpf_preload_info *);
+	int (*finish)(void);
+	struct module *owner;
+};
+extern struct bpf_preload_ops *bpf_preload_ops;
+#define BPF_PRELOAD_LINKS 2
+#endif

kernel/bpf/preload/bpf_preload_kern.c

@@ -0,0 +1,91 @@
+// SPDX-License-Identifier: GPL-2.0
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/pid.h>
+#include <linux/fs.h>
+#include <linux/sched/signal.h>
+#include "bpf_preload.h"
+
+extern char bpf_preload_umd_start;
+extern char bpf_preload_umd_end;
+
+static int preload(struct bpf_preload_info *obj);
+static int finish(void);
+
+static struct bpf_preload_ops umd_ops = {
+	.info.driver_name = "bpf_preload",
+	.preload = preload,
+	.finish = finish,
+	.owner = THIS_MODULE,
+};
+
+static int preload(struct bpf_preload_info *obj)
+{
+	int magic = BPF_PRELOAD_START;
+	loff_t pos = 0;
+	int i, err;
+	ssize_t n;
+
+	err = fork_usermode_driver(&umd_ops.info);
+	if (err)
+		return err;
+
+	/* send the start magic to let UMD proceed with loading BPF progs */
+	n = kernel_write(umd_ops.info.pipe_to_umh,
+			 &magic, sizeof(magic), &pos);
+	if (n != sizeof(magic))
+		return -EPIPE;
+
+	/* receive bpf_link IDs and names from UMD */
+	pos = 0;
+	for (i = 0; i < BPF_PRELOAD_LINKS; i++) {
+		n = kernel_read(umd_ops.info.pipe_from_umh,
+				&obj[i], sizeof(*obj), &pos);
+		if (n != sizeof(*obj))
+			return -EPIPE;
+	}
+	return 0;
+}
+
+static int finish(void)
+{
+	int magic = BPF_PRELOAD_END;
+	struct pid *tgid;
+	loff_t pos = 0;
+	ssize_t n;
+
+	/* send the last magic to UMD. It will do a normal exit. */
+	n = kernel_write(umd_ops.info.pipe_to_umh,
+			 &magic, sizeof(magic), &pos);
+	if (n != sizeof(magic))
+		return -EPIPE;
+	tgid = umd_ops.info.tgid;
+	wait_event(tgid->wait_pidfd, thread_group_exited(tgid));
+	umd_ops.info.tgid = NULL;
+	return 0;
+}
+
+static int __init load_umd(void)
+{
+	int err;
+
+	err = umd_load_blob(&umd_ops.info, &bpf_preload_umd_start,
+			    &bpf_preload_umd_end - &bpf_preload_umd_start);
+	if (err)
+		return err;
+	bpf_preload_ops = &umd_ops;
+	return err;
+}
+
+static void __exit fini_umd(void)
+{
+	bpf_preload_ops = NULL;
+	/* kill UMD in case it's still there due to earlier error */
+	kill_pid(umd_ops.info.tgid, SIGKILL, 1);
+	umd_ops.info.tgid = NULL;
+	umd_unload_blob(&umd_ops.info);
+}
+late_initcall(load_umd);
+module_exit(fini_umd);
+MODULE_LICENSE("GPL");

kernel/bpf/preload/bpf_preload_umd_blob.S

@@ -0,0 +1,7 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+	.section .init.rodata, "a"
+	.global bpf_preload_umd_start
+bpf_preload_umd_start:
+	.incbin "kernel/bpf/preload/bpf_preload_umd"
+	.global bpf_preload_umd_end
+bpf_preload_umd_end:

kernel/bpf/preload/iterators/bpf_preload_common.h

@@ -0,0 +1,13 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _BPF_PRELOAD_COMMON_H
+#define _BPF_PRELOAD_COMMON_H
+
+#define BPF_PRELOAD_START 0x5555
+#define BPF_PRELOAD_END 0xAAAA
+
+struct bpf_preload_info {
+	char link_name[16];
+	int link_id;
+};
+
+#endif

kernel/bpf/preload/iterators/iterators.c

@@ -0,0 +1,94 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2020 Facebook */
+#include <argp.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <sys/resource.h>
+#include <bpf/libbpf.h>
+#include <bpf/bpf.h>
+#include <sys/mount.h>
+#include "iterators.skel.h"
+#include "bpf_preload_common.h"
+
+int to_kernel = -1;
+int from_kernel = 0;
+
+static int send_link_to_kernel(struct bpf_link *link, const char *link_name)
+{
+	struct bpf_preload_info obj = {};
+	struct bpf_link_info info = {};
+	__u32 info_len = sizeof(info);
+	int err;
+
+	err = bpf_obj_get_info_by_fd(bpf_link__fd(link), &info, &info_len);
+	if (err)
+		return err;
+	obj.link_id = info.id;
+	if (strlen(link_name) >= sizeof(obj.link_name))
+		return -E2BIG;
+	strcpy(obj.link_name, link_name);
+	if (write(to_kernel, &obj, sizeof(obj)) != sizeof(obj))
+		return -EPIPE;
+	return 0;
+}
+
+int main(int argc, char **argv)
+{
+	struct rlimit rlim = { RLIM_INFINITY, RLIM_INFINITY };
+	struct iterators_bpf *skel;
+	int err, magic;
+	int debug_fd;
+
+	debug_fd = open("/dev/console", O_WRONLY | O_NOCTTY | O_CLOEXEC);
+	if (debug_fd < 0)
+		return 1;
+	to_kernel = dup(1);
+	close(1);
+	dup(debug_fd);
+	/* now stdin and stderr point to /dev/console */
+
+	read(from_kernel, &magic, sizeof(magic));
+	if (magic != BPF_PRELOAD_START) {
+		printf("bad start magic %d\n", magic);
+		return 1;
+	}
+	setrlimit(RLIMIT_MEMLOCK, &rlim);
+	/* libbpf opens BPF object and loads it into the kernel */
+	skel = iterators_bpf__open_and_load();
+	if (!skel) {
+		/* iterators.skel.h is little endian.
+		 * libbpf doesn't support automatic little->big conversion
+		 * of BPF bytecode yet.
+		 * The program load will fail in such case.
+		 */
+		printf("Failed load could be due to wrong endianness\n");
+		return 1;
+	}
+	err = iterators_bpf__attach(skel);
+	if (err)
+		goto cleanup;
+
+	/* send two bpf_link IDs with names to the kernel */
+	err = send_link_to_kernel(skel->links.dump_bpf_map, "maps.debug");
+	if (err)
+		goto cleanup;
+	err = send_link_to_kernel(skel->links.dump_bpf_prog, "progs.debug");
+	if (err)
+		goto cleanup;
+
+	/* The kernel will proceed with pinnging the links in bpffs.
+	 * UMD will wait on read from pipe.
+	 */
+	read(from_kernel, &magic, sizeof(magic));
+	if (magic != BPF_PRELOAD_END) {
+		printf("bad final magic %d\n", magic);
+		err = -EINVAL;
+	}
+cleanup:
+	iterators_bpf__destroy(skel);
+
+	return err != 0;
+}