Skip to contentSkip to conversation

net: bpf: arm64: address randomize and write protect JIT code

PATCH Superseded Daniel Borkmann 2 participants 3 messages Sep 15, 2014

Subsystems (3)

v3 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

PATCH v3 1/1 net: bpf: arm64: address randomize and write protect JIT code

This is the ARM64 variant for 314beb9bcab ("x86: bpf_jit_comp: secure bpf jit against spraying attacks").

Thanks to commit 11d91a770f1f ("arm64: Add CONFIG_DEBUG_SET_MODULE_RONX support") which added necessary infrastructure, we can now implement RO marking of eBPF generated JIT image pages and randomize start offset for the JIT code, so that it does not reside directly on a page boundary anymore. Likewise, the holes are filled with illegal instructions: here we use BRK #0x100 (opcode 0xd4202000) to trigger a fault in the kernel (unallocated BRKs would trigger a fault through do_debug_exception). This seems more reliable as we don't have a guaranteed undefined instruction space on ARM64.

This is basically the ARM64 variant of what we already have in ARM via commit 55309dd3d4cd ("net: bpf: arm: address randomize and write protect JIT code"). Moreover, this commit also presents a merge resolution due to conflicts with commit 60a3b2253c41 ("net: bpf: make eBPF interpreter images read-only") as we don't use kfree() in bpf_jit_free() anymore to release the locked bpf_prog structure, but instead bpf_prog_unlock_free() through a different allocator.

JIT tested on aarch64 with BPF test suite.

Reference: http://mainisusuallyafunction.blogspot.com/2012/11/attacking-hardened-linux-systems-with.html

Cc Zi Shen Lim <[email protected]>
Cc Will Deacon <[email protected]>
Cc Catalin Marinas <[email protected]>
Cc David S. Miller <[email protected]>
Cc Alexei Starovoitov <[email protected]>

---
v2->v3:
- Use cpu_to_le32() as suggested by Zi/Will
v1->v2:
- Use brk insn as suggested by Catalin
Note:
- This patch depends on net-next being merged to mainline due to the mentioned merge conflict.

 arch/arm64/net/bpf_jit_comp.c | 38 +++++++++++++++++++++++++++++---------
 1 file changed, 29 insertions(+), 9 deletions(-)
arch/arm64/net/bpf_jit_comp.c +29 -9
index 7ae3354..4b71779 100644
@@ -19,7 +19,6 @@
 #define pr_fmt(fmt) "bpf_jit: " fmt
 
 #include <linux/filter.h>
-#include <linux/moduleloader.h>
 #include <linux/printk.h>
 #include <linux/skbuff.h>
 #include <linux/slab.h>
@@ -119,6 +118,15 @@ static inline int bpf2a64_offset(int bpf_to, int bpf_from,
 	return to - from;
 }
 
+static void jit_fill_hole(void *area, unsigned int size)
+{
+	/* We use brk #0x100 to trigger a fault. */
+	u32 *ptr;
+	/* We are guaranteed to have aligned memory. */
+	for (ptr = area; size >= sizeof(u32); size -= sizeof(u32))
+		*ptr++ = cpu_to_le32(0xd4202000);
+}
+
 static inline int epilogue_offset(const struct jit_ctx *ctx)
 {
 	int to = ctx->offset[ctx->prog->len - 1];
@@ -613,8 +621,10 @@ void bpf_jit_compile(struct bpf_prog *prog)
 
 void bpf_int_jit_compile(struct bpf_prog *prog)
 {
+	struct bpf_binary_header *header;
 	struct jit_ctx ctx;
 	int image_size;
+	u8 *image_ptr;
 
 	if (!bpf_jit_enable)
 		return;
@@ -636,23 +646,25 @@ void bpf_int_jit_compile(struct bpf_prog *prog)
 		goto out;
 
 	build_prologue(&ctx);
-
 	build_epilogue(&ctx);
 
 	/* Now we know the actual image size. */
 	image_size = sizeof(u32) * ctx.idx;
-	ctx.image = module_alloc(image_size);
-	if (unlikely(ctx.image == NULL))
+	header = bpf_jit_binary_alloc(image_size, &image_ptr,
+				      sizeof(u32), jit_fill_hole);
+	if (header == NULL)
 		goto out;
 
 	/* 2. Now, the actual pass. */
 
+	ctx.image = (u32 *)image_ptr;
 	ctx.idx = 0;
+
 	build_prologue(&ctx);
 
 	ctx.body_offset = ctx.idx;
 	if (build_body(&ctx)) {
-		module_free(NULL, ctx.image);
+		bpf_jit_binary_free(header);
 		goto out;
 	}
 
@@ -663,17 +675,25 @@ void bpf_int_jit_compile(struct bpf_prog *prog)
 		bpf_jit_dump(prog->len, image_size, 2, ctx.image);
 
 	bpf_flush_icache(ctx.image, ctx.image + ctx.idx);
+
+	set_memory_ro((unsigned long)header, header->pages);
 	prog->bpf_func = (void *)ctx.image;
 	prog->jited = 1;
-
 out:
 	kfree(ctx.offset);
 }
 
 void bpf_jit_free(struct bpf_prog *prog)
 {
-	if (prog->jited)
-		module_free(NULL, prog->bpf_func);
+	unsigned long addr = (unsigned long)prog->bpf_func & PAGE_MASK;
+	struct bpf_binary_header *header = (void *)addr;
+
+	if (!prog->jited)
+		goto free_filter;
+
+	set_memory_rw(addr, header->pages);
+	bpf_jit_binary_free(header);
 
-	kfree(prog);
+free_filter:
+	bpf_prog_unlock_free(prog);
 }
Signature · 1 line
1.9.3
#1 lore raw

Hi Daniel,

One small comment than I forgot to mention before (see below).

Quoted 56 lines · Daniel Borkmann

This is the ARM64 variant for 314beb9bcab ("x86: bpf_jit_comp: secure bpf jit against spraying attacks").

Thanks to commit 11d91a770f1f ("arm64: Add CONFIG_DEBUG_SET_MODULE_RONX support") which added necessary infrastructure, we can now implement RO marking of eBPF generated JIT image pages and randomize start offset for the JIT code, so that it does not reside directly on a page boundary anymore. Likewise, the holes are filled with illegal instructions: here we use BRK #0x100 (opcode 0xd4202000) to trigger a fault in the kernel (unallocated BRKs would trigger a fault through do_debug_exception). This seems more reliable as we don't have a guaranteed undefined instruction space on ARM64.

This is basically the ARM64 variant of what we already have in ARM via commit 55309dd3d4cd ("net: bpf: arm: address randomize and write protect JIT code"). Moreover, this commit also presents a merge resolution due to conflicts with commit 60a3b2253c41 ("net: bpf: make eBPF interpreter images read-only") as we don't use kfree() in bpf_jit_free() anymore to release the locked bpf_prog structure, but instead bpf_prog_unlock_free() through a different allocator.

JIT tested on aarch64 with BPF test suite.

Reference: http://mainisusuallyafunction.blogspot.com/2012/11/attacking-hardened-linux-systems-with.html

Cc Zi Shen Lim <[email protected]>
Cc Will Deacon <[email protected]>
Cc Catalin Marinas <[email protected]>
Cc David S. Miller <[email protected]>
Cc Alexei Starovoitov <[email protected]>

---
v2->v3:
- Use cpu_to_le32() as suggested by Zi/Will
v1->v2:
- Use brk insn as suggested by Catalin
Note:
- This patch depends on net-next being merged to mainline due to the mentioned merge conflict.

 arch/arm64/net/bpf_jit_comp.c | 38 +++++++++++++++++++++++++++++---------
 1 file changed, 29 insertions(+), 9 deletions(-)
arch/arm64/net/bpf_jit_comp.c +0 -1
index 7ae3354..4b71779 100644
@@ -19,7 +19,6 @@
 #define pr_fmt(fmt) "bpf_jit: " fmt
 
 #include <linux/filter.h>
-#include <linux/moduleloader.h>
 #include <linux/printk.h>
 #include <linux/skbuff.h>
 #include <linux/slab.h>
@@ -119,6 +118,15 @@ static inline int bpf2a64_offset(int bpf_to, int bpf_from,
 	return to - from;

arch/arm64/net/bpf_jit_comp.c static inline int bpf2a64_offset(int bpf_to, int bpf_from, · line 128

Quoted 10 lines
arch/arm64/net/bpf_jit_comp.c +8 -0
 }
 
+static void jit_fill_hole(void *area, unsigned int size)
+{
+	/* We use brk #0x100 to trigger a fault. */
+	u32 *ptr;
+	/* We are guaranteed to have aligned memory. */
+	for (ptr = area; size >= sizeof(u32); size -= sizeof(u32))
+		*ptr++ = cpu_to_le32(0xd4202000);
+}

Please can you use the existing macros in debug-monitors.h for constructing this value? We have AARCH64_BREAK_MON to get the BRK encoding, then you can add an immediate to that file to reserve something for bpf. That way, we make sure that other parts of the kernel don't accidentally repurpose your fault code.

Cheers,

Will

#2 lore raw
Quoted 63 lines · Will Deacon

Hi Daniel,

One small comment than I forgot to mention before (see below).

Quoted 58 lines · Daniel Borkmann

This is the ARM64 variant for 314beb9bcab ("x86: bpf_jit_comp: secure bpf jit against spraying attacks").

Thanks to commit 11d91a770f1f ("arm64: Add CONFIG_DEBUG_SET_MODULE_RONX support") which added necessary infrastructure, we can now implement RO marking of eBPF generated JIT image pages and randomize start offset for the JIT code, so that it does not reside directly on a page boundary anymore. Likewise, the holes are filled with illegal instructions: here we use BRK #0x100 (opcode 0xd4202000) to trigger a fault in the kernel (unallocated BRKs would trigger a fault through do_debug_exception). This seems more reliable as we don't have a guaranteed undefined instruction space on ARM64.

This is basically the ARM64 variant of what we already have in ARM via commit 55309dd3d4cd ("net: bpf: arm: address randomize and write protect JIT code"). Moreover, this commit also presents a merge resolution due to conflicts with commit 60a3b2253c41 ("net: bpf: make eBPF interpreter images read-only") as we don't use kfree() in bpf_jit_free() anymore to release the locked bpf_prog structure, but instead bpf_prog_unlock_free() through a different allocator.

JIT tested on aarch64 with BPF test suite.

Reference: http://mainisusuallyafunction.blogspot.com/2012/11/attacking-hardened-linux-systems-with.html

Cc Zi Shen Lim <[email protected]>
Cc Will Deacon <[email protected]>
Cc Catalin Marinas <[email protected]>
Cc David S. Miller <[email protected]>
Cc Alexei Starovoitov <[email protected]>

---
v2->v3:
- Use cpu_to_le32() as suggested by Zi/Will
v1->v2:
- Use brk insn as suggested by Catalin
Note:
- This patch depends on net-next being merged to mainline due to the mentioned merge conflict.

  arch/arm64/net/bpf_jit_comp.c | 38 +++++++++++++++++++++++++++++---------
  1 file changed, 29 insertions(+), 9 deletions(-)
arch/arm64/net/bpf_jit_comp.c +0 -1
index 7ae3354..4b71779 100644
@@ -19,7 +19,6 @@
  #define pr_fmt(fmt) "bpf_jit: " fmt
 
  #include <linux/filter.h>
-#include <linux/moduleloader.h>
  #include <linux/printk.h>
  #include <linux/skbuff.h>
  #include <linux/slab.h>
@@ -119,6 +118,15 @@ static inline int bpf2a64_offset(int bpf_to, int bpf_from,
  	return to - from;
  }
 
Quoted 14 lines

arch/arm64/net/bpf_jit_comp.c static inline int bpf2a64_offset(int bpf_to, int bpf_from, · line 128

Quoted 8 lines
arch/arm64/net/bpf_jit_comp.c +8 -0
+static void jit_fill_hole(void *area, unsigned int size)
+{
+	/* We use brk #0x100 to trigger a fault. */
+	u32 *ptr;
+	/* We are guaranteed to have aligned memory. */
+	for (ptr = area; size >= sizeof(u32); size -= sizeof(u32))
+		*ptr++ = cpu_to_le32(0xd4202000);
+}

Please can you use the existing macros in debug-monitors.h for constructing this value? We have AARCH64_BREAK_MON to get the BRK encoding, then you can add an immediate to that file to reserve something for bpf. That way, we make sure that other parts of the kernel don't accidentally repurpose your fault code.

Okay, no problem, I'll do that and respin tomorrow.

Thanks,

Daniel