diff options
| -rw-r--r-- | init/Kconfig | 2 | ||||
| -rw-r--r-- | kernel/Makefile | 2 | ||||
| -rw-r--r-- | kernel/bpf/Makefile | 1 | ||||
| -rw-r--r-- | kernel/bpf/inode.c | 116 | ||||
| -rw-r--r-- | kernel/bpf/preload/Kconfig | 23 | ||||
| -rw-r--r-- | kernel/bpf/preload/Makefile | 23 | ||||
| -rw-r--r-- | kernel/bpf/preload/bpf_preload.h | 16 | ||||
| -rw-r--r-- | kernel/bpf/preload/bpf_preload_kern.c | 91 | ||||
| -rw-r--r-- | kernel/bpf/preload/bpf_preload_umd_blob.S | 7 | ||||
| -rw-r--r-- | kernel/bpf/preload/iterators/bpf_preload_common.h | 13 | ||||
| -rw-r--r-- | kernel/bpf/preload/iterators/iterators.c | 94 | ||||
| -rw-r--r-- | net/bpfilter/Kconfig | 1 | ||||
| -rw-r--r-- | tools/lib/bpf/Makefile | 7 |
13 files changed, 390 insertions, 6 deletions
diff --git a/init/Kconfig b/init/Kconfig index d6a0b31b13dc..fc10f7ede5f6 100644 --- a/init/Kconfig +++ b/init/Kconfig @@ -1710,6 +1710,8 @@ config BPF_JIT_DEFAULT_ON def_bool ARCH_WANT_DEFAULT_BPF_JIT || BPF_JIT_ALWAYS_ON depends on HAVE_EBPF_JIT && BPF_JIT +source "kernel/bpf/preload/Kconfig" + config USERFAULTFD bool "Enable userfaultfd() system call" depends on MMU diff --git a/kernel/Makefile b/kernel/Makefile index 9a20016d4900..22b0760660fc 100644 --- a/kernel/Makefile +++ b/kernel/Makefile @@ -12,7 +12,7 @@ obj-y = fork.o exec_domain.o panic.o \ notifier.o ksysfs.o cred.o reboot.o \ async.o range.o smpboot.o ucount.o regset.o -obj-$(CONFIG_BPFILTER) += usermode_driver.o +obj-$(CONFIG_USERMODE_DRIVER) += usermode_driver.o obj-$(CONFIG_MODULES) += kmod.o obj-$(CONFIG_MULTIUSER) += groups.o diff --git a/kernel/bpf/Makefile b/kernel/bpf/Makefile index e6eb9c0402da..19e137aae40e 100644 --- a/kernel/bpf/Makefile +++ b/kernel/bpf/Makefile @@ -29,3 +29,4 @@ ifeq ($(CONFIG_BPF_JIT),y) obj-$(CONFIG_BPF_SYSCALL) += bpf_struct_ops.o obj-${CONFIG_BPF_LSM} += bpf_lsm.o endif +obj-$(CONFIG_BPF_PRELOAD) += preload/ diff --git a/kernel/bpf/inode.c b/kernel/bpf/inode.c index fb878ba3f22f..b48a56f53495 100644 --- a/kernel/bpf/inode.c +++ b/kernel/bpf/inode.c @@ -20,6 +20,7 @@ #include <linux/filter.h> #include <linux/bpf.h> #include <linux/bpf_trace.h> +#include "preload/bpf_preload.h" enum bpf_type { BPF_TYPE_UNSPEC = 0, @@ -369,9 +370,10 @@ static struct dentry * bpf_lookup(struct inode *dir, struct dentry *dentry, unsigned flags) { /* Dots in names (e.g. "/sys/fs/bpf/foo.bar") are reserved for future - * extensions. + * extensions. That allows popoulate_bpffs() create special files. */ - if (strchr(dentry->d_name.name, '.')) + if ((dir->i_mode & S_IALLUGO) && + strchr(dentry->d_name.name, '.')) return ERR_PTR(-EPERM); return simple_lookup(dir, dentry, flags); @@ -409,6 +411,27 @@ static const struct inode_operations bpf_dir_iops = { .unlink = simple_unlink, }; +/* pin iterator link into bpffs */ +static int bpf_iter_link_pin_kernel(struct dentry *parent, + const char *name, struct bpf_link *link) +{ + umode_t mode = S_IFREG | S_IRUSR; + struct dentry *dentry; + int ret; + + inode_lock(parent->d_inode); + dentry = lookup_one_len(name, parent, strlen(name)); + if (IS_ERR(dentry)) { + inode_unlock(parent->d_inode); + return PTR_ERR(dentry); + } + ret = bpf_mkobj_ops(dentry, mode, link, &bpf_link_iops, + &bpf_iter_fops); + dput(dentry); + inode_unlock(parent->d_inode); + return ret; +} + static int bpf_obj_do_pin(const char __user *pathname, void *raw, enum bpf_type type) { @@ -638,6 +661,91 @@ static int bpf_parse_param(struct fs_context *fc, struct fs_parameter *param) return 0; } +struct bpf_preload_ops *bpf_preload_ops; +EXPORT_SYMBOL_GPL(bpf_preload_ops); + +static bool bpf_preload_mod_get(void) +{ + /* If bpf_preload.ko wasn't loaded earlier then load it now. + * When bpf_preload is built into vmlinux the module's __init + * function will populate it. + */ + if (!bpf_preload_ops) { + request_module("bpf_preload"); + if (!bpf_preload_ops) + return false; + } + /* And grab the reference, so the module doesn't disappear while the + * kernel is interacting with the kernel module and its UMD. + */ + if (!try_module_get(bpf_preload_ops->owner)) { + pr_err("bpf_preload module get failed.\n"); + return false; + } + return true; +} + +static void bpf_preload_mod_put(void) +{ + if (bpf_preload_ops) + /* now user can "rmmod bpf_preload" if necessary */ + module_put(bpf_preload_ops->owner); +} + +static DEFINE_MUTEX(bpf_preload_lock); + +static int populate_bpffs(struct dentry *parent) +{ + struct bpf_preload_info objs[BPF_PRELOAD_LINKS] = {}; + struct bpf_link *links[BPF_PRELOAD_LINKS] = {}; + int err = 0, i; + + /* grab the mutex to make sure the kernel interactions with bpf_preload + * UMD are serialized + */ + mutex_lock(&bpf_preload_lock); + + /* if bpf_preload.ko wasn't built into vmlinux then load it */ + if (!bpf_preload_mod_get()) + goto out; + + if (!bpf_preload_ops->info.tgid) { + /* preload() will start UMD that will load BPF iterator programs */ + err = bpf_preload_ops->preload(objs); + if (err) + goto out_put; + for (i = 0; i < BPF_PRELOAD_LINKS; i++) { + links[i] = bpf_link_by_id(objs[i].link_id); + if (IS_ERR(links[i])) { + err = PTR_ERR(links[i]); + goto out_put; + } + } + for (i = 0; i < BPF_PRELOAD_LINKS; i++) { + err = bpf_iter_link_pin_kernel(parent, + objs[i].link_name, links[i]); + if (err) + goto out_put; + /* do not unlink successfully pinned links even + * if later link fails to pin + */ + links[i] = NULL; + } + /* finish() will tell UMD process to exit */ + err = bpf_preload_ops->finish(); + if (err) + goto out_put; + } +out_put: + bpf_preload_mod_put(); +out: + mutex_unlock(&bpf_preload_lock); + for (i = 0; i < BPF_PRELOAD_LINKS && err; i++) + if (!IS_ERR_OR_NULL(links[i])) + bpf_link_put(links[i]); + return err; +} + static int bpf_fill_super(struct super_block *sb, struct fs_context *fc) { static const struct tree_descr bpf_rfiles[] = { { "" } }; @@ -654,8 +762,8 @@ static int bpf_fill_super(struct super_block *sb, struct fs_context *fc) inode = sb->s_root->d_inode; inode->i_op = &bpf_dir_iops; inode->i_mode &= ~S_IALLUGO; + populate_bpffs(sb->s_root); inode->i_mode |= S_ISVTX | opts->mode; - return 0; } @@ -705,6 +813,8 @@ static int __init bpf_init(void) { int ret; + mutex_init(&bpf_preload_lock); + ret = sysfs_create_mount_point(fs_kobj, "bpf"); if (ret) return ret; diff --git a/kernel/bpf/preload/Kconfig b/kernel/bpf/preload/Kconfig new file mode 100644 index 000000000000..7144e2d01ee4 --- /dev/null +++ b/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 diff --git a/kernel/bpf/preload/Makefile b/kernel/bpf/preload/Makefile new file mode 100644 index 000000000000..12c7b62b9b6e --- /dev/null +++ b/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 diff --git a/kernel/bpf/preload/bpf_preload.h b/kernel/bpf/preload/bpf_preload.h new file mode 100644 index 000000000000..2f9932276f2e --- /dev/null +++ b/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 diff --git a/kernel/bpf/preload/bpf_preload_kern.c b/kernel/bpf/preload/bpf_preload_kern.c new file mode 100644 index 000000000000..79c5772465f1 --- /dev/null +++ b/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"); diff --git a/kernel/bpf/preload/bpf_preload_umd_blob.S b/kernel/bpf/preload/bpf_preload_umd_blob.S new file mode 100644 index 000000000000..f1f40223b5c3 --- /dev/null +++ b/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: diff --git a/kernel/bpf/preload/iterators/bpf_preload_common.h b/kernel/bpf/preload/iterators/bpf_preload_common.h new file mode 100644 index 000000000000..8464d1a48c05 --- /dev/null +++ b/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 diff --git a/kernel/bpf/preload/iterators/iterators.c b/kernel/bpf/preload/iterators/iterators.c new file mode 100644 index 000000000000..b7ff87939172 --- /dev/null +++ b/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; +} diff --git a/net/bpfilter/Kconfig b/net/bpfilter/Kconfig index 73d0b12789f1..8ad0233ce497 100644 --- a/net/bpfilter/Kconfig +++ b/net/bpfilter/Kconfig @@ -2,6 +2,7 @@ menuconfig BPFILTER bool "BPF based packet filtering framework (BPFILTER)" depends on NET && BPF && INET + select USERMODE_DRIVER help This builds experimental bpfilter framework that is aiming to provide netfilter compatible functionality via BPF diff --git a/tools/lib/bpf/Makefile b/tools/lib/bpf/Makefile index c5dbfafdf889..66b2cfadf262 100644 --- a/tools/lib/bpf/Makefile +++ b/tools/lib/bpf/Makefile @@ -1,6 +1,9 @@ # SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause) # Most of this file is copied from tools/lib/traceevent/Makefile +RM ?= rm +srctree = $(abs_srctree) + LIBBPF_VERSION := $(shell \ grep -oE '^LIBBPF_([0-9.]+)' libbpf.map | \ sort -rV | head -n1 | cut -d'_' -f2) @@ -188,7 +191,7 @@ $(OUTPUT)libbpf.so.$(LIBBPF_VERSION): $(BPF_IN_SHARED) @ln -sf $(@F) $(OUTPUT)libbpf.so.$(LIBBPF_MAJOR_VERSION) $(OUTPUT)libbpf.a: $(BPF_IN_STATIC) - $(QUIET_LINK)$(RM) $@; $(AR) rcs $@ $^ + $(QUIET_LINK)$(RM) -f $@; $(AR) rcs $@ $^ $(OUTPUT)libbpf.pc: $(QUIET_GEN)sed -e "s|@PREFIX@|$(prefix)|" \ @@ -291,7 +294,7 @@ cscope: cscope -b -q -I $(srctree)/include -f cscope.out tags: - rm -f TAGS tags + $(RM) -f TAGS tags ls *.c *.h | xargs $(TAGS_PROG) -a # Declare the contents of the .PHONY variable as phony. We keep that |