4.17-stable review patch. If anyone has any objections, please let me know.
------------------
From: Daniel Borkmann daniel@iogearbox.net
commit 9facc336876f7ecf9edba4c67b90426fde4ec898 upstream.
We currently lock any JITed image as read-only via bpf_jit_binary_lock_ro() as well as the BPF image as read-only through bpf_prog_lock_ro(). In the case any of these would fail we throw a WARN_ON_ONCE() in order to yell loudly to the log. Perhaps, to some extend, this may be comparable to an allocation where __GFP_NOWARN is explicitly not set.
Added via 65869a47f348 ("bpf: improve read-only handling"), this behavior is slightly different compared to any of the other in-kernel set_memory_ro() users who do not check the return code of set_memory_ro() and friends /at all/ (e.g. in the case of module_enable_ro() / module_disable_ro()). Given in BPF this is mandatory hardening step, we want to know whether there are any issues that would leave both BPF data writable. So it happens that syzkaller enabled fault injection and it triggered memory allocation failure deep inside x86's change_page_attr_set_clr() which was triggered from set_memory_ro().
Now, there are two options: i) leaving everything as is, and ii) reworking the image locking code in order to have a final checkpoint out of the central bpf_prog_select_runtime() which probes whether any of the calls during prog setup weren't successful, and then bailing out with an error. Option ii) is a better approach since this additional paranoia avoids altogether leaving any potential W+X pages from BPF side in the system. Therefore, lets be strict about it, and reject programs in such unlikely occasion. While testing I noticed also that one bpf_prog_lock_ro() call was missing on the outer dummy prog in case of calls, e.g. in the destructor we call bpf_prog_free_deferred() on the main prog where we try to bpf_prog_unlock_free() the program, and since we go via bpf_prog_select_runtime() do that as well.
Reported-by: syzbot+3b889862e65a98317058@syzkaller.appspotmail.com Reported-by: syzbot+9e762b52dd17e616a7a5@syzkaller.appspotmail.com Signed-off-by: Daniel Borkmann daniel@iogearbox.net Acked-by: Martin KaFai Lau kafai@fb.com Signed-off-by: Alexei Starovoitov ast@kernel.org Signed-off-by: Greg Kroah-Hartman gregkh@linuxfoundation.org
--- include/linux/filter.h | 60 +++++++++++++++++++++++++++++++------------------ kernel/bpf/core.c | 53 ++++++++++++++++++++++++++++++++++++++----- kernel/bpf/syscall.c | 4 --- 3 files changed, 86 insertions(+), 31 deletions(-)
--- a/include/linux/filter.h +++ b/include/linux/filter.h @@ -453,7 +453,8 @@ struct sock_fprog_kern { };
struct bpf_binary_header { - unsigned int pages; + u16 pages; + u16 locked:1; u8 image[]; };
@@ -644,15 +645,18 @@ bpf_ctx_narrow_access_ok(u32 off, u32 si
#define bpf_classic_proglen(fprog) (fprog->len * sizeof(fprog->filter[0]))
-#ifdef CONFIG_ARCH_HAS_SET_MEMORY static inline void bpf_prog_lock_ro(struct bpf_prog *fp) { +#ifdef CONFIG_ARCH_HAS_SET_MEMORY fp->locked = 1; - WARN_ON_ONCE(set_memory_ro((unsigned long)fp, fp->pages)); + if (set_memory_ro((unsigned long)fp, fp->pages)) + fp->locked = 0; +#endif }
static inline void bpf_prog_unlock_ro(struct bpf_prog *fp) { +#ifdef CONFIG_ARCH_HAS_SET_MEMORY if (fp->locked) { WARN_ON_ONCE(set_memory_rw((unsigned long)fp, fp->pages)); /* In case set_memory_rw() fails, we want to be the first @@ -660,34 +664,30 @@ static inline void bpf_prog_unlock_ro(st */ fp->locked = 0; } +#endif }
static inline void bpf_jit_binary_lock_ro(struct bpf_binary_header *hdr) { - WARN_ON_ONCE(set_memory_ro((unsigned long)hdr, hdr->pages)); -} - -static inline void bpf_jit_binary_unlock_ro(struct bpf_binary_header *hdr) -{ - WARN_ON_ONCE(set_memory_rw((unsigned long)hdr, hdr->pages)); -} -#else -static inline void bpf_prog_lock_ro(struct bpf_prog *fp) -{ -} - -static inline void bpf_prog_unlock_ro(struct bpf_prog *fp) -{ -} - -static inline void bpf_jit_binary_lock_ro(struct bpf_binary_header *hdr) -{ +#ifdef CONFIG_ARCH_HAS_SET_MEMORY + hdr->locked = 1; + if (set_memory_ro((unsigned long)hdr, hdr->pages)) + hdr->locked = 0; +#endif }
static inline void bpf_jit_binary_unlock_ro(struct bpf_binary_header *hdr) { +#ifdef CONFIG_ARCH_HAS_SET_MEMORY + if (hdr->locked) { + WARN_ON_ONCE(set_memory_rw((unsigned long)hdr, hdr->pages)); + /* In case set_memory_rw() fails, we want to be the first + * to crash here instead of some random place later on. + */ + hdr->locked = 0; + } +#endif } -#endif /* CONFIG_ARCH_HAS_SET_MEMORY */
static inline struct bpf_binary_header * bpf_jit_binary_hdr(const struct bpf_prog *fp) @@ -698,6 +698,22 @@ bpf_jit_binary_hdr(const struct bpf_prog return (void *)addr; }
+#ifdef CONFIG_ARCH_HAS_SET_MEMORY +static inline int bpf_prog_check_pages_ro_single(const struct bpf_prog *fp) +{ + if (!fp->locked) + return -ENOLCK; + if (fp->jited) { + const struct bpf_binary_header *hdr = bpf_jit_binary_hdr(fp); + + if (!hdr->locked) + return -ENOLCK; + } + + return 0; +} +#endif + int sk_filter_trim_cap(struct sock *sk, struct sk_buff *skb, unsigned int cap); static inline int sk_filter(struct sock *sk, struct sk_buff *skb) { --- a/kernel/bpf/core.c +++ b/kernel/bpf/core.c @@ -583,6 +583,8 @@ bpf_jit_binary_alloc(unsigned int progle bpf_fill_ill_insns(hdr, size);
hdr->pages = size / PAGE_SIZE; + hdr->locked = 0; + hole = min_t(unsigned int, size - (proglen + sizeof(*hdr)), PAGE_SIZE - sizeof(*hdr)); start = (get_random_int() % hole) & ~(alignment - 1); @@ -1513,6 +1515,33 @@ static int bpf_check_tail_call(const str return 0; }
+static int bpf_prog_check_pages_ro_locked(const struct bpf_prog *fp) +{ +#ifdef CONFIG_ARCH_HAS_SET_MEMORY + int i, err; + + for (i = 0; i < fp->aux->func_cnt; i++) { + err = bpf_prog_check_pages_ro_single(fp->aux->func[i]); + if (err) + return err; + } + + return bpf_prog_check_pages_ro_single(fp); +#endif + return 0; +} + +static void bpf_prog_select_func(struct bpf_prog *fp) +{ +#ifndef CONFIG_BPF_JIT_ALWAYS_ON + u32 stack_depth = max_t(u32, fp->aux->stack_depth, 1); + + fp->bpf_func = interpreters[(round_up(stack_depth, 32) / 32) - 1]; +#else + fp->bpf_func = __bpf_prog_ret0_warn; +#endif +} + /** * bpf_prog_select_runtime - select exec runtime for BPF program * @fp: bpf_prog populated with internal BPF program @@ -1523,13 +1552,13 @@ static int bpf_check_tail_call(const str */ struct bpf_prog *bpf_prog_select_runtime(struct bpf_prog *fp, int *err) { -#ifndef CONFIG_BPF_JIT_ALWAYS_ON - u32 stack_depth = max_t(u32, fp->aux->stack_depth, 1); + /* In case of BPF to BPF calls, verifier did all the prep + * work with regards to JITing, etc. + */ + if (fp->bpf_func) + goto finalize;
- fp->bpf_func = interpreters[(round_up(stack_depth, 32) / 32) - 1]; -#else - fp->bpf_func = __bpf_prog_ret0_warn; -#endif + bpf_prog_select_func(fp);
/* eBPF JITs can rewrite the program in case constant * blinding is active. However, in case of error during @@ -1550,6 +1579,8 @@ struct bpf_prog *bpf_prog_select_runtime if (*err) return fp; } + +finalize: bpf_prog_lock_ro(fp);
/* The tail call compatibility check can only be done at @@ -1558,7 +1589,17 @@ struct bpf_prog *bpf_prog_select_runtime * all eBPF JITs might immediately support all features. */ *err = bpf_check_tail_call(fp); + if (*err) + return fp;
+ /* Checkpoint: at this point onwards any cBPF -> eBPF or + * native eBPF program is read-only. If we failed to change + * the page attributes (e.g. allocation failure from + * splitting large pages), then reject the whole program + * in order to guarantee not ending up with any W+X pages + * from BPF side in kernel. + */ + *err = bpf_prog_check_pages_ro_locked(fp); return fp; } EXPORT_SYMBOL_GPL(bpf_prog_select_runtime); --- a/kernel/bpf/syscall.c +++ b/kernel/bpf/syscall.c @@ -1328,9 +1328,7 @@ static int bpf_prog_load(union bpf_attr if (err < 0) goto free_used_maps;
- /* eBPF program is ready to be JITed */ - if (!prog->bpf_func) - prog = bpf_prog_select_runtime(prog, &err); + prog = bpf_prog_select_runtime(prog, &err); if (err < 0) goto free_used_maps;