Skip to contentSkip to conversation

bpf: Fix issue with bpf preload module taking over stdout/stdin of kernel.

PATCH Superseded Yucong Sun 3 participants 3 messages Feb 25, 2022

Subsystems (3)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

Yucong Sun

author
#0 lore raw

PATCH v2 1/1 bpf: Fix issue with bpf preload module taking over stdout/stdin of kernel.

In a previous commit (1), BPF preload process was switched from user mode process to use in-kernel light skeleton instead. However, in the kernel context the available fd starts from 0, instead of normally 3 for user mode process. and the preload process leaked two FDs, taking over FD 0 and 1. This which later caused issues when kernel trys to setup stdin/stdout/stderr for init process, assuming fd 0,1,2 is available.

As seen here:

Before fix:
ls -lah /proc/1/fd/*

lrwx------1 root root 64 Feb 23 17:20 /proc/1/fd/0 -> /dev/null lrwx------ 1 root root 64 Feb 23 17:20 /proc/1/fd/1 -> /dev/null lrwx------ 1 root root 64 Feb 23 17:20 /proc/1/fd/2 -> /dev/console lrwx------ 1 root root 64 Feb 23 17:20 /proc/1/fd/6 -> /dev/console lrwx------ 1 root root 64 Feb 23 17:20 /proc/1/fd/7 -> /dev/console

After Fix / Normal:

ls -lah /proc/1/fd/*

lrwx------ 1 root root 64 Feb 24 21:23 /proc/1/fd/0 -> /dev/console lrwx------ 1 root root 64 Feb 24 21:23 /proc/1/fd/1 -> /dev/console lrwx------ 1 root root 64 Feb 24 21:23 /proc/1/fd/2 -> /dev/console

In this patch:
- skel_closenz was changed to skel_closegez to correctly handle

    FD=0 case.

- various places detecting FD > 0 was changed to FD >= 0.
- Call iterators_skel__detach() funciton to release FDs after links are obtained.

1: commit cb80ddc67152 ("bpf: Convert bpf_preload.ko to use light skeleton.")

Fixes commit cb80ddc67152 ("bpf: Convert bpf_preload.ko to use light skeleton.")

V2 -> V1: rename skel_closenez to skel_closegez, added comment as requested.

---
 kernel/bpf/preload/bpf_preload_kern.c          |  4 ++++
 kernel/bpf/preload/iterators/iterators.lskel.h | 16 +++++++++-------
 tools/bpf/bpftool/gen.c                        |  9 +++++----
 tools/lib/bpf/skel_internal.h                  |  8 ++++----
 4 files changed, 22 insertions(+), 15 deletions(-)
kernel/bpf/preload/bpf_preload_kern.c +4 -0
index 30207c048d36..3cc8bbfd15b1 100644
@@ -14,6 +14,8 @@ static void free_links_and_skel(void)
 		bpf_link_put(maps_link);
 	if (!IS_ERR_OR_NULL(progs_link))
 		bpf_link_put(progs_link);
+	/* __detach() was already called before this, __destory() will call it again, but
+	  with no effect. */
 	iterators_bpf__destroy(skel);
 }
 
@@ -54,6 +56,8 @@ static int load_skel(void)
 		err = PTR_ERR(progs_link);
 		goto out;
 	}
+	/* Release all FDs */
+	iterators_bpf__detach(skel);
 	return 0;
 out:
 	free_links_and_skel();
kernel/bpf/preload/iterators/iterators.lskel.h +9 -7
index 70f236a82fe1..6a93538fa69f 100644
@@ -28,7 +28,7 @@ iterators_bpf__dump_bpf_map__attach(struct iterators_bpf *skel)
 	int prog_fd = skel->progs.dump_bpf_map.prog_fd;
 	int fd = skel_link_create(prog_fd, 0, BPF_TRACE_ITER);
 
-	if (fd > 0)
+	if (fd >= 0)
 		skel->links.dump_bpf_map_fd = fd;
 	return fd;
 }
@@ -39,7 +39,7 @@ iterators_bpf__dump_bpf_prog__attach(struct iterators_bpf *skel)
 	int prog_fd = skel->progs.dump_bpf_prog.prog_fd;
 	int fd = skel_link_create(prog_fd, 0, BPF_TRACE_ITER);
 
-	if (fd > 0)
+	if (fd >= 0)
 		skel->links.dump_bpf_prog_fd = fd;
 	return fd;
 }
@@ -57,8 +57,10 @@ iterators_bpf__attach(struct iterators_bpf *skel)
 static inline void
 iterators_bpf__detach(struct iterators_bpf *skel)
 {
-	skel_closenz(skel->links.dump_bpf_map_fd);
-	skel_closenz(skel->links.dump_bpf_prog_fd);
+	skel_closegez(skel->links.dump_bpf_map_fd);
+	skel->links.dump_bpf_map_fd = -1;
+	skel_closegez(skel->links.dump_bpf_prog_fd);
+	skel->links.dump_bpf_prog_fd = -1;
 }
 static void
 iterators_bpf__destroy(struct iterators_bpf *skel)
@@ -66,10 +68,10 @@ iterators_bpf__destroy(struct iterators_bpf *skel)
 	if (!skel)
 		return;
 	iterators_bpf__detach(skel);
-	skel_closenz(skel->progs.dump_bpf_map.prog_fd);
-	skel_closenz(skel->progs.dump_bpf_prog.prog_fd);
+	skel_closegez(skel->progs.dump_bpf_map.prog_fd);
+	skel_closegez(skel->progs.dump_bpf_prog.prog_fd);
 	skel_free_map_data(skel->rodata, skel->maps.rodata.initial_value, 4096);
-	skel_closenz(skel->maps.rodata.map_fd);
+	skel_closegez(skel->maps.rodata.map_fd);
 	skel_free(skel);
 }
 static inline struct iterators_bpf *
tools/bpf/bpftool/gen.c +5 -4
index 145734b4fe41..e5e65f507e00 100644
@@ -469,7 +469,7 @@ static void codegen_attach_detach(struct bpf_object *obj, const char *obj_name)
 		codegen("\
 			\n\
 										    \n\
-				if (fd > 0)					    \n\
+				if (fd >= 0)					    \n\
 					skel->links.%1$s_fd = fd;		    \n\
 				return fd;					    \n\
 			}							    \n\
@@ -506,7 +506,8 @@ static void codegen_attach_detach(struct bpf_object *obj, const char *obj_name)
 	bpf_object__for_each_program(prog, obj) {
 		codegen("\
 			\n\
-				skel_closenz(skel->links.%1$s_fd);	    \n\
+				skel_closegez(skel->links.%1$s_fd);	    \n\
+				skel->links.%1$s_fd = -1;	    \n\
 			", bpf_program__name(prog));
 	}
 
@@ -536,7 +537,7 @@ static void codegen_destroy(struct bpf_object *obj, const char *obj_name)
 	bpf_object__for_each_program(prog, obj) {
 		codegen("\
 			\n\
-				skel_closenz(skel->progs.%1$s.prog_fd);	    \n\
+				skel_closegez(skel->progs.%1$s.prog_fd);	    \n\
 			", bpf_program__name(prog));
 	}
 
@@ -549,7 +550,7 @@ static void codegen_destroy(struct bpf_object *obj, const char *obj_name)
 			       ident, bpf_map_mmap_sz(map));
 		codegen("\
 			\n\
-				skel_closenz(skel->maps.%1$s.map_fd);	    \n\
+				skel_closegez(skel->maps.%1$s.map_fd);	    \n\
 			", ident);
 	}
 	codegen("\
tools/lib/bpf/skel_internal.h +4 -4
index bd6f4505e7b1..89c0b8632254 100644
@@ -204,11 +204,11 @@ static inline void *skel_finalize_map_data(__u64 *init_val, size_t mmap_sz, int
 }
 #endif
 
-static inline int skel_closenz(int fd)
+static inline int skel_closegez(int fd)
 {
-	if (fd > 0)
-		return close(fd);
-	return -EINVAL;
+	if (fd < 0)
+		return -EINVAL;
+	return close(fd);
 }
 
 #ifndef offsetofend
Signature · 1 line
2.30.2
Acked-by on v2 1/1
#1 lore raw
Quoted 1 line · Yucong Sun

In a previous commit (1), BPF preload process was switched from user

For commit, you can just say in:
In commit cb80ddc67152 ("bpf: Convert bpf_preload.ko to use light skeleton."), BPF preload process ...

People uses reference ([1]) is for web links.

Quoted 25 lines

mode process to use in-kernel light skeleton instead. However, in the kernel context the available fd starts from 0, instead of normally 3 for user mode process. and the preload process leaked two FDs, taking over FD 0 and 1. This which later caused issues when kernel trys to setup stdin/stdout/stderr for init process, assuming fd 0,1,2 is available.

As seen here:

Before fix:
ls -lah /proc/1/fd/*

lrwx------1 root root 64 Feb 23 17:20 /proc/1/fd/0 -> /dev/null lrwx------ 1 root root 64 Feb 23 17:20 /proc/1/fd/1 -> /dev/null lrwx------ 1 root root 64 Feb 23 17:20 /proc/1/fd/2 -> /dev/console lrwx------ 1 root root 64 Feb 23 17:20 /proc/1/fd/6 -> /dev/console lrwx------ 1 root root 64 Feb 23 17:20 /proc/1/fd/7 -> /dev/console

After Fix / Normal:

ls -lah /proc/1/fd/*

lrwx------ 1 root root 64 Feb 24 21:23 /proc/1/fd/0 -> /dev/console lrwx------ 1 root root 64 Feb 24 21:23 /proc/1/fd/1 -> /dev/console lrwx------ 1 root root 64 Feb 24 21:23 /proc/1/fd/2 -> /dev/console

Quoted 8 lines

In this patch:
- skel_closenz was changed to skel_closegez to correctly handle

     FD=0 case.

- various places detecting FD > 0 was changed to FD >= 0.
- Call iterators_skel__detach() funciton to release FDs after links are obtained.

1: commit cb80ddc67152 ("bpf: Convert bpf_preload.ko to use light skeleton.")

You don't need the above line.

Quoted 3 lines
Fixes commit cb80ddc67152 ("bpf: Convert bpf_preload.ko to use light skeleton.")

LGTM. One comment below.

Quoted 15 lines

V2 -> V1: rename skel_closenez to skel_closegez, added comment as requested.

---
  kernel/bpf/preload/bpf_preload_kern.c          |  4 ++++
  kernel/bpf/preload/iterators/iterators.lskel.h | 16 +++++++++-------
  tools/bpf/bpftool/gen.c                        |  9 +++++----
  tools/lib/bpf/skel_internal.h                  |  8 ++++----
  4 files changed, 22 insertions(+), 15 deletions(-)
kernel/bpf/preload/bpf_preload_kern.c +0 -0
index 30207c048d36..3cc8bbfd15b1 100644

kernel/bpf/preload/bpf_preload_kern.c static void free_links_and_skel(void) · line 19

Quoted 6 lines
kernel/bpf/preload/bpf_preload_kern.c +2 -0
@@ -14,6 +14,8 @@ static void free_links_and_skel(void)
  		bpf_link_put(maps_link);
  	if (!IS_ERR_OR_NULL(progs_link))
  		bpf_link_put(progs_link);
+	/* __detach() was already called before this, __destory() will call it again, but
+	  with no effect. */
  	iterators_bpf__destroy(skel);

This is not the right place to put the comment as free_links_and_skel() is also called in load_skel() in failure path.

kernel/bpf/preload/bpf_preload_kern.c static int load_skel(void) · line 60

Quoted 8 lines
kernel/bpf/preload/bpf_preload_kern.c +2 -0
  }
 
@@ -54,6 +56,8 @@ static int load_skel(void)
  		err = PTR_ERR(progs_link);
  		goto out;
  	}
+	/* Release all FDs */
+	iterators_bpf__detach(skel);

How about we put the comments in free_links_and_skel() here. The comments can be something like:

	/* Release all FDs to avoid impacting stdin/stdout/stderr setup
	 * in init process. Later call of this function in 

iterators_bpf__destroy() will be a noop. */

Quoted 10 lines
kernel/bpf/preload/bpf_preload_kern.c +0 -0
  	return 0;
  out:
  	free_links_and_skel();
kernel/bpf/preload/iterators/iterators.lskel.h +0 -0
index 70f236a82fe1..6a93538fa69f 100644
@@ -28,7 +28,7 @@ iterators_bpf__dump_bpf_map__attach(struct iterators_bpf *skel)
  	int prog_fd = skel->progs.dump_bpf_map.prog_fd;
  	int fd = skel_link_create(prog_fd, 0, BPF_TRACE_ITER);

[...]

#2 lore raw
In reply to Yucong Sun
Quoted 168 lines · Yucong Sun

In a previous commit (1), BPF preload process was switched from user mode process to use in-kernel light skeleton instead. However, in the kernel context the available fd starts from 0, instead of normally 3 for user mode process. and the preload process leaked two FDs, taking over FD 0 and 1. This which later caused issues when kernel trys to setup stdin/stdout/stderr for init process, assuming fd 0,1,2 is available.

As seen here:

Before fix:
ls -lah /proc/1/fd/*

lrwx------1 root root 64 Feb 23 17:20 /proc/1/fd/0 -> /dev/null lrwx------ 1 root root 64 Feb 23 17:20 /proc/1/fd/1 -> /dev/null lrwx------ 1 root root 64 Feb 23 17:20 /proc/1/fd/2 -> /dev/console lrwx------ 1 root root 64 Feb 23 17:20 /proc/1/fd/6 -> /dev/console lrwx------ 1 root root 64 Feb 23 17:20 /proc/1/fd/7 -> /dev/console

After Fix / Normal:

ls -lah /proc/1/fd/*

lrwx------ 1 root root 64 Feb 24 21:23 /proc/1/fd/0 -> /dev/console lrwx------ 1 root root 64 Feb 24 21:23 /proc/1/fd/1 -> /dev/console lrwx------ 1 root root 64 Feb 24 21:23 /proc/1/fd/2 -> /dev/console

In this patch:
- skel_closenz was changed to skel_closegez to correctly handle

    FD=0 case.

- various places detecting FD > 0 was changed to FD >= 0.
- Call iterators_skel__detach() funciton to release FDs after links are obtained.

1: commit cb80ddc67152 ("bpf: Convert bpf_preload.ko to use light skeleton.")

Fixes commit cb80ddc67152 ("bpf: Convert bpf_preload.ko to use light skeleton.")

V2 -> V1: rename skel_closenez to skel_closegez, added comment as requested.

---
 kernel/bpf/preload/bpf_preload_kern.c          |  4 ++++
 kernel/bpf/preload/iterators/iterators.lskel.h | 16 +++++++++-------
 tools/bpf/bpftool/gen.c                        |  9 +++++----
 tools/lib/bpf/skel_internal.h                  |  8 ++++----
 4 files changed, 22 insertions(+), 15 deletions(-)
kernel/bpf/preload/bpf_preload_kern.c +4 -0
index 30207c048d36..3cc8bbfd15b1 100644
@@ -14,6 +14,8 @@ static void free_links_and_skel(void)
                bpf_link_put(maps_link);
        if (!IS_ERR_OR_NULL(progs_link))
                bpf_link_put(progs_link);
+       /* __detach() was already called before this, __destory() will call it again, but
+         with no effect. */
        iterators_bpf__destroy(skel);
 }
 
@@ -54,6 +56,8 @@ static int load_skel(void)
                err = PTR_ERR(progs_link);
                goto out;
        }
+       /* Release all FDs */
+       iterators_bpf__detach(skel);
        return 0;
 out:
        free_links_and_skel();
kernel/bpf/preload/iterators/iterators.lskel.h +9 -7
index 70f236a82fe1..6a93538fa69f 100644
@@ -28,7 +28,7 @@ iterators_bpf__dump_bpf_map__attach(struct iterators_bpf *skel)
        int prog_fd = skel->progs.dump_bpf_map.prog_fd;
        int fd = skel_link_create(prog_fd, 0, BPF_TRACE_ITER);
 
-       if (fd > 0)
+       if (fd >= 0)
                skel->links.dump_bpf_map_fd = fd;
        return fd;
 }
@@ -39,7 +39,7 @@ iterators_bpf__dump_bpf_prog__attach(struct iterators_bpf *skel)
        int prog_fd = skel->progs.dump_bpf_prog.prog_fd;
        int fd = skel_link_create(prog_fd, 0, BPF_TRACE_ITER);
 
-       if (fd > 0)
+       if (fd >= 0)
                skel->links.dump_bpf_prog_fd = fd;
        return fd;
 }
@@ -57,8 +57,10 @@ iterators_bpf__attach(struct iterators_bpf *skel)
 static inline void
 iterators_bpf__detach(struct iterators_bpf *skel)
 {
-       skel_closenz(skel->links.dump_bpf_map_fd);
-       skel_closenz(skel->links.dump_bpf_prog_fd);
+       skel_closegez(skel->links.dump_bpf_map_fd);
+       skel->links.dump_bpf_map_fd = -1;
+       skel_closegez(skel->links.dump_bpf_prog_fd);
+       skel->links.dump_bpf_prog_fd = -1;
 }
 static void
 iterators_bpf__destroy(struct iterators_bpf *skel)
@@ -66,10 +68,10 @@ iterators_bpf__destroy(struct iterators_bpf *skel)
        if (!skel)
                return;
        iterators_bpf__detach(skel);
-       skel_closenz(skel->progs.dump_bpf_map.prog_fd);
-       skel_closenz(skel->progs.dump_bpf_prog.prog_fd);
+       skel_closegez(skel->progs.dump_bpf_map.prog_fd);
+       skel_closegez(skel->progs.dump_bpf_prog.prog_fd);
        skel_free_map_data(skel->rodata, skel->maps.rodata.initial_value, 4096);
-       skel_closenz(skel->maps.rodata.map_fd);
+       skel_closegez(skel->maps.rodata.map_fd);
        skel_free(skel);
 }
 static inline struct iterators_bpf *
tools/bpf/bpftool/gen.c +5 -4
index 145734b4fe41..e5e65f507e00 100644
@@ -469,7 +469,7 @@ static void codegen_attach_detach(struct bpf_object *obj, const char *obj_name)
                codegen("\
                        \n\
                                                                                    \n\
-                               if (fd > 0)                                         \n\
+                               if (fd >= 0)                                        \n\
                                        skel->links.%1$s_fd = fd;                   \n\
                                return fd;                                          \n\
                        }                                                           \n\
@@ -506,7 +506,8 @@ static void codegen_attach_detach(struct bpf_object *obj, const char *obj_name)
        bpf_object__for_each_program(prog, obj) {
                codegen("\
                        \n\
-                               skel_closenz(skel->links.%1$s_fd);          \n\
+                               skel_closegez(skel->links.%1$s_fd);         \n\
+                               skel->links.%1$s_fd = -1;           \n\
                        ", bpf_program__name(prog));
        }
 
@@ -536,7 +537,7 @@ static void codegen_destroy(struct bpf_object *obj, const char *obj_name)
        bpf_object__for_each_program(prog, obj) {
                codegen("\
                        \n\
-                               skel_closenz(skel->progs.%1$s.prog_fd);     \n\
+                               skel_closegez(skel->progs.%1$s.prog_fd);            \n\
                        ", bpf_program__name(prog));
        }
 
@@ -549,7 +550,7 @@ static void codegen_destroy(struct bpf_object *obj, const char *obj_name)
                               ident, bpf_map_mmap_sz(map));
                codegen("\
                        \n\
-                               skel_closenz(skel->maps.%1$s.map_fd);       \n\
+                               skel_closegez(skel->maps.%1$s.map_fd);      \n\
                        ", ident);
        }
        codegen("\
tools/lib/bpf/skel_internal.h +0 -0
index bd6f4505e7b1..89c0b8632254 100644
@@ -204,11 +204,11 @@ static inline void *skel_finalize_map_data(__u64 *init_val, size_t mmap_sz, int
 }
 #endif
 

tools/lib/bpf/skel_internal.h static inline void *skel_finalize_map_data(__u64 *init_val, size_t mmap_sz, int · line 212

Quoted 10 lines
tools/lib/bpf/skel_internal.h +4 -4
-static inline int skel_closenz(int fd)
+static inline int skel_closegez(int fd)
 {
-       if (fd > 0)
-               return close(fd);
-       return -EINVAL;
+       if (fd < 0)
+               return -EINVAL;
+       return close(fd);
 }

Unfortunately this won't work. Many places in gen_loader.c rely on fd == 0 being a signal that fd wasn't allocated. The global data, stack, loader_ctx, etc. All are zero initialized. Thankfully no need to do any of these changes.
Just closing two link_fd in load_skel() is enough.