]> Pileus Git - ~andy/linux/commitdiff
s390/bpf,jit: address randomize and write protect jit code
authorHeiko Carstens <heiko.carstens@de.ibm.com>
Tue, 16 Jul 2013 11:25:49 +0000 (13:25 +0200)
committerMartin Schwidefsky <schwidefsky@de.ibm.com>
Thu, 18 Jul 2013 10:44:37 +0000 (12:44 +0200)
This is the s390 variant of 314beb9b "x86: bpf_jit_comp: secure bpf
jit against spraying attacks".
With this change the whole jit code and literal pool will be write
protected after creation. In addition the start address of the jit
code won't be always on a page boundary anymore.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
arch/s390/net/bpf_jit_comp.c

index 80828bfee2ec01cf9e232975c5ea54b4812b6af1..788e22395acde967e288f4cc82a20a1d89103a72 100644 (file)
@@ -9,6 +9,7 @@
 #include <linux/netdevice.h>
 #include <linux/if_vlan.h>
 #include <linux/filter.h>
+#include <linux/random.h>
 #include <asm/cacheflush.h>
 #include <asm/processor.h>
 #include <asm/facility.h>
@@ -738,8 +739,41 @@ out:
        return -1;
 }
 
+/*
+ * Note: for security reasons, bpf code will follow a randomly
+ *      sized amount of illegal instructions.
+ */
+struct bpf_binary_header {
+       unsigned int pages;
+       u8 image[];
+};
+
+static struct bpf_binary_header *bpf_alloc_binary(unsigned int bpfsize,
+                                                 u8 **image_ptr)
+{
+       struct bpf_binary_header *header;
+       unsigned int sz, hole;
+
+       /* Most BPF filters are really small, but if some of them fill a page,
+        * allow at least 128 extra bytes for illegal instructions.
+        */
+       sz = round_up(bpfsize + sizeof(*header) + 128, PAGE_SIZE);
+       header = module_alloc(sz);
+       if (!header)
+               return NULL;
+       memset(header, 0, sz);
+       header->pages = sz / PAGE_SIZE;
+       hole = sz - bpfsize + sizeof(*header);
+       /* Insert random number of illegal instructions before BPF code
+        * and make sure the first instruction starts at an even address.
+        */
+       *image_ptr = &header->image[(prandom_u32() % hole) & -2];
+       return header;
+}
+
 void bpf_jit_compile(struct sk_filter *fp)
 {
+       struct bpf_binary_header *header = NULL;
        unsigned long size, prg_len, lit_len;
        struct bpf_jit jit, cjit;
        unsigned int *addrs;
@@ -775,8 +809,8 @@ void bpf_jit_compile(struct sk_filter *fp)
                        size = prg_len + lit_len;
                        if (size >= BPF_SIZE_MAX)
                                goto out;
-                       jit.start = module_alloc(size);
-                       if (!jit.start)
+                       header = bpf_alloc_binary(size, &jit.start);
+                       if (!header)
                                goto out;
                        jit.prg = jit.mid = jit.start + prg_len;
                        jit.lit = jit.end = jit.start + prg_len + lit_len;
@@ -791,14 +825,21 @@ void bpf_jit_compile(struct sk_filter *fp)
                if (jit.start)
                        print_fn_code(jit.start, jit.mid - jit.start);
        }
-       if (jit.start)
+       if (jit.start) {
+               set_memory_ro((unsigned long)header, header->pages);
                fp->bpf_func = (void *) jit.start;
+       }
 out:
        kfree(addrs);
 }
 
 void bpf_jit_free(struct sk_filter *fp)
 {
-       if (fp->bpf_func != sk_run_filter)
-               module_free(NULL, fp->bpf_func);
+       unsigned long addr = (unsigned long)fp->bpf_func & PAGE_MASK;
+       struct bpf_binary_header *header = (void *)addr;
+
+       if (fp->bpf_func == sk_run_filter)
+               return;
+       set_memory_rw(addr, header->pages);
+       module_free(NULL, header);
 }