net: bpf: make eBPF interpreter images read-only

With eBPF getting more extended and exposure to user space is on it's way,
hardening the memory range the interpreter uses to steer its command flow
seems appropriate.  This patch moves the to be interpreted bytecode to
read-only pages.

In case we execute a corrupted BPF interpreter image for some reason e.g.
caused by an attacker which got past a verifier stage, it would not only
provide arbitrary read/write memory access but arbitrary function calls
as well. After setting up the BPF interpreter image, its contents do not
change until destruction time, thus we can setup the image on immutable
made pages in order to mitigate modifications to that code. The idea
is derived from commit 314beb9bca ("x86: bpf_jit_comp: secure bpf jit
against spraying attacks").

This is possible because bpf_prog is not part of sk_filter anymore.
After setup bpf_prog cannot be altered during its life-time. This prevents
any modifications to the entire bpf_prog structure (incl. function/JIT
image pointer).

Every eBPF program (including classic BPF that are migrated) have to call
bpf_prog_select_runtime() to select either interpreter or a JIT image
as a last setup step, and they all are being freed via bpf_prog_free(),
including non-JIT. Therefore, we can easily integrate this into the
eBPF life-time, plus since we directly allocate a bpf_prog, we have no
performance penalty.

Tested with seccomp and test_bpf testsuite in JIT/non-JIT mode and manual
inspection of kernel_page_tables.  Brad Spengler proposed the same idea
via Twitter during development of this patch.

Joint work with Hannes Frederic Sowa.

Suggested-by: Brad Spengler <spender@grsecurity.net>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Cc: Alexei Starovoitov <ast@plumgrid.com>
Cc: Kees Cook <keescook@chromium.org>
Acked-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

arch/arm/net/bpf_jit_32.c

@@ -930,5 +930,6 @@ void bpf_jit_free(struct bpf_prog *fp)
 {
 	if (fp->jited)
 		module_free(NULL, fp->bpf_func);
-	kfree(fp);
+
+	bpf_prog_unlock_free(fp);
 }

arch/mips/net/bpf_jit.c

@@ -1427,5 +1427,6 @@ void bpf_jit_free(struct bpf_prog *fp)
 {
 	if (fp->jited)
 		module_free(NULL, fp->bpf_func);
-	kfree(fp);
+
+	bpf_prog_unlock_free(fp);
 }

arch/powerpc/net/bpf_jit_comp.c

@@ -697,5 +697,6 @@ void bpf_jit_free(struct bpf_prog *fp)
 {
 	if (fp->jited)
 		module_free(NULL, fp->bpf_func);
-	kfree(fp);
+
+	bpf_prog_unlock_free(fp);
 }

arch/s390/net/bpf_jit_comp.c

@@ -887,5 +887,5 @@ void bpf_jit_free(struct bpf_prog *fp)
 	module_free(NULL, header);
 
 free_filter:
-	kfree(fp);
+	bpf_prog_unlock_free(fp);
 }

arch/sparc/net/bpf_jit_comp.c

@@ -812,5 +812,6 @@ void bpf_jit_free(struct bpf_prog *fp)
 {
 	if (fp->jited)
 		module_free(NULL, fp->bpf_func);
-	kfree(fp);
+
+	bpf_prog_unlock_free(fp);
 }

arch/x86/net/bpf_jit_comp.c

@@ -972,23 +972,17 @@ out:
 	kfree(addrs);
 }
 
-static void bpf_jit_free_deferred(struct work_struct *work)
+void bpf_jit_free(struct bpf_prog *fp)
 {
-	struct bpf_prog *fp = container_of(work, struct bpf_prog, work);
 	unsigned long addr = (unsigned long)fp->bpf_func & PAGE_MASK;
 	struct bpf_binary_header *header = (void *)addr;
 
+	if (!fp->jited)
+		goto free_filter;
+
 	set_memory_rw(addr, header->pages);
 	module_free(NULL, header);
-	kfree(fp);
-}
 
-void bpf_jit_free(struct bpf_prog *fp)
-{
-	if (fp->jited) {
-		INIT_WORK(&fp->work, bpf_jit_free_deferred);
-		schedule_work(&fp->work);
-	} else {
-		kfree(fp);
-	}
+free_filter:
+	bpf_prog_unlock_free(fp);
 }

include/linux/filter.h

@@ -9,6 +9,11 @@
 #include <linux/skbuff.h>
 #include <linux/workqueue.h>
 #include <uapi/linux/filter.h>
+#include <asm/cacheflush.h>
+
+struct sk_buff;
+struct sock;
+struct seccomp_data;
 
 /* Internally used and optimized filter representation with extended
  * instruction set based on top of classic BPF.
@@ -320,20 +325,23 @@ struct sock_fprog_kern {
 	struct sock_filter	*filter;
 };
 
-struct sk_buff;
-struct sock;
-struct seccomp_data;
+struct bpf_work_struct {
+	struct bpf_prog *prog;
+	struct work_struct work;
+};
 
 struct bpf_prog {
+	u32			pages;		/* Number of allocated pages */
 	u32			jited:1,	/* Is our filter JIT'ed? */
 				len:31;		/* Number of filter blocks */
 	struct sock_fprog_kern	*orig_prog;	/* Original BPF program */
+	struct bpf_work_struct	*work;		/* Deferred free work struct */
 	unsigned int		(*bpf_func)(const struct sk_buff *skb,
 					    const struct bpf_insn *filter);
+	/* Instructions for interpreter */
 	union {
 		struct sock_filter	insns[0];
 		struct bpf_insn		insnsi[0];
-		struct work_struct	work;
 	};
 };
 
@@ -353,6 +361,26 @@ static inline unsigned int bpf_prog_size(unsigned int proglen)
 
 #define bpf_classic_proglen(fprog) (fprog->len * sizeof(fprog->filter[0]))
 
+#ifdef CONFIG_DEBUG_SET_MODULE_RONX
+static inline void bpf_prog_lock_ro(struct bpf_prog *fp)
+{
+	set_memory_ro((unsigned long)fp, fp->pages);
+}
+
+static inline void bpf_prog_unlock_ro(struct bpf_prog *fp)
+{
+	set_memory_rw((unsigned long)fp, fp->pages);
+}
+#else
+static inline void bpf_prog_lock_ro(struct bpf_prog *fp)
+{
+}
+
+static inline void bpf_prog_unlock_ro(struct bpf_prog *fp)
+{
+}
+#endif /* CONFIG_DEBUG_SET_MODULE_RONX */
+
 int sk_filter(struct sock *sk, struct sk_buff *skb);
 
 void bpf_prog_select_runtime(struct bpf_prog *fp);
@@ -361,6 +389,17 @@ void bpf_prog_free(struct bpf_prog *fp);
 int bpf_convert_filter(struct sock_filter *prog, int len,
 		       struct bpf_insn *new_prog, int *new_len);
 
+struct bpf_prog *bpf_prog_alloc(unsigned int size, gfp_t gfp_extra_flags);
+struct bpf_prog *bpf_prog_realloc(struct bpf_prog *fp_old, unsigned int size,
+				  gfp_t gfp_extra_flags);
+void __bpf_prog_free(struct bpf_prog *fp);
+
+static inline void bpf_prog_unlock_free(struct bpf_prog *fp)
+{
+	bpf_prog_unlock_ro(fp);
+	__bpf_prog_free(fp);
+}
+
 int bpf_prog_create(struct bpf_prog **pfp, struct sock_fprog_kern *fprog);
 void bpf_prog_destroy(struct bpf_prog *fp);
 
@@ -450,7 +489,7 @@ static inline void bpf_jit_compile(struct bpf_prog *fp)
 
 static inline void bpf_jit_free(struct bpf_prog *fp)
 {
-	kfree(fp);
+	bpf_prog_unlock_free(fp);
 }
 #endif /* CONFIG_BPF_JIT */
 

kernel/bpf/core.c

@@ -22,6 +22,7 @@
  */
 #include <linux/filter.h>
 #include <linux/skbuff.h>
+#include <linux/vmalloc.h>
 #include <asm/unaligned.h>
 
 /* Registers */
@@ -63,6 +64,67 @@ void *bpf_internal_load_pointer_neg_helper(const struct sk_buff *skb, int k, uns
 	return NULL;
 }
 
+struct bpf_prog *bpf_prog_alloc(unsigned int size, gfp_t gfp_extra_flags)
+{
+	gfp_t gfp_flags = GFP_KERNEL | __GFP_HIGHMEM | __GFP_ZERO |
+			  gfp_extra_flags;
+	struct bpf_work_struct *ws;
+	struct bpf_prog *fp;
+
+	size = round_up(size, PAGE_SIZE);
+	fp = __vmalloc(size, gfp_flags, PAGE_KERNEL);
+	if (fp == NULL)
+		return NULL;
+
+	ws = kmalloc(sizeof(*ws), GFP_KERNEL | gfp_extra_flags);
+	if (ws == NULL) {
+		vfree(fp);
+		return NULL;
+	}
+
+	fp->pages = size / PAGE_SIZE;
+	fp->work = ws;
+
+	return fp;
+}
+EXPORT_SYMBOL_GPL(bpf_prog_alloc);
+
+struct bpf_prog *bpf_prog_realloc(struct bpf_prog *fp_old, unsigned int size,
+				  gfp_t gfp_extra_flags)
+{
+	gfp_t gfp_flags = GFP_KERNEL | __GFP_HIGHMEM | __GFP_ZERO |
+			  gfp_extra_flags;
+	struct bpf_prog *fp;
+
+	BUG_ON(fp_old == NULL);
+
+	size = round_up(size, PAGE_SIZE);
+	if (size <= fp_old->pages * PAGE_SIZE)
+		return fp_old;
+
+	fp = __vmalloc(size, gfp_flags, PAGE_KERNEL);
+	if (fp != NULL) {
+		memcpy(fp, fp_old, fp_old->pages * PAGE_SIZE);
+		fp->pages = size / PAGE_SIZE;
+
+		/* We keep fp->work from fp_old around in the new
+		 * reallocated structure.
+		 */
+		fp_old->work = NULL;
+		__bpf_prog_free(fp_old);
+	}
+
+	return fp;
+}
+EXPORT_SYMBOL_GPL(bpf_prog_realloc);
+
+void __bpf_prog_free(struct bpf_prog *fp)
+{
+	kfree(fp->work);
+	vfree(fp);
+}
+EXPORT_SYMBOL_GPL(__bpf_prog_free);
+
 /* Base function for offset calculation. Needs to go into .text section,
  * therefore keeping it non-static as well; will also be used by JITs
  * anyway later on, so do not let the compiler omit it.
@@ -523,12 +585,26 @@ void bpf_prog_select_runtime(struct bpf_prog *fp)
 
 	/* Probe if internal BPF can be JITed */
 	bpf_int_jit_compile(fp);
+	/* Lock whole bpf_prog as read-only */
+	bpf_prog_lock_ro(fp);
 }
 EXPORT_SYMBOL_GPL(bpf_prog_select_runtime);
 
-/* free internal BPF program */
+static void bpf_prog_free_deferred(struct work_struct *work)
+{
+	struct bpf_work_struct *ws;
+
+	ws = container_of(work, struct bpf_work_struct, work);
+	bpf_jit_free(ws->prog);
+}
+
+/* Free internal BPF program */
 void bpf_prog_free(struct bpf_prog *fp)
 {
-	bpf_jit_free(fp);
+	struct bpf_work_struct *ws = fp->work;
+
+	INIT_WORK(&ws->work, bpf_prog_free_deferred);
+	ws->prog = fp;
+	schedule_work(&ws->work);
 }
 EXPORT_SYMBOL_GPL(bpf_prog_free);

kernel/seccomp.c

@@ -395,16 +395,15 @@ static struct seccomp_filter *seccomp_prepare_filter(struct sock_fprog *fprog)
 	if (!filter)
 		goto free_prog;
 
-	filter->prog = kzalloc(bpf_prog_size(new_len),
-			       GFP_KERNEL|__GFP_NOWARN);
+	filter->prog = bpf_prog_alloc(bpf_prog_size(new_len), __GFP_NOWARN);
 	if (!filter->prog)
 		goto free_filter;
 
 	ret = bpf_convert_filter(fp, fprog->len, filter->prog->insnsi, &new_len);
 	if (ret)
 		goto free_filter_prog;
-	kfree(fp);
 
+	kfree(fp);
 	atomic_set(&filter->usage, 1);
 	filter->prog->len = new_len;
 
@@ -413,7 +412,7 @@ static struct seccomp_filter *seccomp_prepare_filter(struct sock_fprog *fprog)
 	return filter;
 
 free_filter_prog:
-	kfree(filter->prog);
+	__bpf_prog_free(filter->prog);
 free_filter:
 	kfree(filter);
 free_prog:

lib/test_bpf.c

@@ -1836,7 +1836,7 @@ static struct bpf_prog *generate_filter(int which, int *err)
 		break;
 
 	case INTERNAL:
-		fp = kzalloc(bpf_prog_size(flen), GFP_KERNEL);
+		fp = bpf_prog_alloc(bpf_prog_size(flen), 0);
 		if (fp == NULL) {
 			pr_cont("UNEXPECTED_FAIL no memory left\n");
 			*err = -ENOMEM;

net/core/filter.c

@@ -933,7 +933,7 @@ static struct bpf_prog *bpf_migrate_filter(struct bpf_prog *fp)
 
 	/* Expand fp for appending the new filter representation. */
 	old_fp = fp;
-	fp = krealloc(old_fp, bpf_prog_size(new_len), GFP_KERNEL);
+	fp = bpf_prog_realloc(old_fp, bpf_prog_size(new_len), 0);
 	if (!fp) {
 		/* The old_fp is still around in case we couldn't
 		 * allocate new memory, so uncharge on that one.
@@ -1013,7 +1013,7 @@ int bpf_prog_create(struct bpf_prog **pfp, struct sock_fprog_kern *fprog)
 	if (fprog->filter == NULL)
 		return -EINVAL;
 
-	fp = kmalloc(bpf_prog_size(fprog->len), GFP_KERNEL);
+	fp = bpf_prog_alloc(bpf_prog_size(fprog->len), 0);
 	if (!fp)
 		return -ENOMEM;
 
@@ -1069,7 +1069,7 @@ int sk_attach_filter(struct sock_fprog *fprog, struct sock *sk)
 	if (fprog->filter == NULL)
 		return -EINVAL;
 
-	prog = kmalloc(bpf_fsize, GFP_KERNEL);
+	prog = bpf_prog_alloc(bpf_fsize, 0);
 	if (!prog)
 		return -ENOMEM;