bpf.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: John Fastabend <john.fastabend@gmail.com>
To: bpf@vger.kernel.org, netdev@vger.kernel.org
Cc: john.fastabend@gmail.com, daniel@iogearbox.net,
	joamaki@gmail.com, xiyou.wangcong@gmail.com
Subject: [PATCH bpf 4/4] bpf, sockmap: sk_skb data_end access incorrect when src_reg = dst_reg
Date: Mon, 11 Oct 2021 12:16:47 -0700	[thread overview]
Message-ID: <20211011191647.418704-5-john.fastabend@gmail.com> (raw)
In-Reply-To: <20211011191647.418704-1-john.fastabend@gmail.com>

From: Jussi Maki <joamaki@gmail.com>

The current conversion of skb->data_end reads like this,

  ; data_end = (void*)(long)skb->data_end;
   559: (79) r1 = *(u64 *)(r2 +200)   ; r1  = skb->data
   560: (61) r11 = *(u32 *)(r2 +112)  ; r11 = skb->len
   561: (0f) r1 += r11
   562: (61) r11 = *(u32 *)(r2 +116)
   563: (1f) r1 -= r11

But similar to the case

 ("bpf: sock_ops sk access may stomp registers when dst_reg = src_reg"),

the code will read an incorrect skb->len when src == dst. In this case we
end up generating this xlated code.

  ; data_end = (void*)(long)skb->data_end;
   559: (79) r1 = *(u64 *)(r1 +200)   ; r1  = skb->data
   560: (61) r11 = *(u32 *)(r1 +112)  ; r11 = (skb->data)->len
   561: (0f) r1 += r11
   562: (61) r11 = *(u32 *)(r1 +116)
   563: (1f) r1 -= r11

where line 560 is the reading 4B of (skb->data + 112) instead of the
intended skb->len Here the skb pointer in r1 gets set to skb->data and
the later deref for skb->len ends up following skb->data instead of skb.

This fixes the issue similarly to the patch mentioned above by creating
an additional temporary variable and using to store the register when
dst_reg = src_reg. We name the variable bpf_temp_reg and place it in the
cb context for sk_skb. Then we restore from the temp to ensure nothing
is lost.

Fixes: 16137b09a66f2 ("bpf: Compute data_end dynamically with JIT code")
Signed-off-by: Jussi Maki <joamaki@gmail.com>
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
---
 include/net/strparser.h |  4 ++++
 net/core/filter.c       | 36 ++++++++++++++++++++++++++++++------
 2 files changed, 34 insertions(+), 6 deletions(-)

diff --git a/include/net/strparser.h b/include/net/strparser.h
index bec1439bd3be..732b7097d78e 100644
--- a/include/net/strparser.h
+++ b/include/net/strparser.h
@@ -66,6 +66,10 @@ struct sk_skb_cb {
 #define SK_SKB_CB_PRIV_LEN 20
 	unsigned char data[SK_SKB_CB_PRIV_LEN];
 	struct _strp_msg strp;
+	/* temp_reg is a temporary register used for bpf_convert_data_end_access
+	 * when dst_reg == src_reg.
+	 */
+	u64 temp_reg;
 };
 
 static inline struct strp_msg *strp_msg(struct sk_buff *skb)
diff --git a/net/core/filter.c b/net/core/filter.c
index 23a9bf92b5bb..f4a63af45f00 100644
--- a/net/core/filter.c
+++ b/net/core/filter.c
@@ -9735,22 +9735,46 @@ static u32 sock_ops_convert_ctx_access(enum bpf_access_type type,
 static struct bpf_insn *bpf_convert_data_end_access(const struct bpf_insn *si,
 						    struct bpf_insn *insn)
 {
-	/* si->dst_reg = skb->data */
+	int reg;
+	int temp_reg_off = offsetof(struct sk_buff, cb) +
+			   offsetof(struct sk_skb_cb, temp_reg);
+
+	if (si->src_reg == si->dst_reg) {
+		/* We need an extra register, choose and save a register. */
+		reg = BPF_REG_9;
+		if (si->src_reg == reg || si->dst_reg == reg)
+			reg--;
+		if (si->src_reg == reg || si->dst_reg == reg)
+			reg--;
+		*insn++ = BPF_STX_MEM(BPF_DW, si->src_reg, reg, temp_reg_off);
+	} else {
+		reg = si->dst_reg;
+	}
+
+	/* reg = skb->data */
 	*insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF(struct sk_buff, data),
-			      si->dst_reg, si->src_reg,
+			      reg, si->src_reg,
 			      offsetof(struct sk_buff, data));
 	/* AX = skb->len */
 	*insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF(struct sk_buff, len),
 			      BPF_REG_AX, si->src_reg,
 			      offsetof(struct sk_buff, len));
-	/* si->dst_reg = skb->data + skb->len */
-	*insn++ = BPF_ALU64_REG(BPF_ADD, si->dst_reg, BPF_REG_AX);
+	/* reg = skb->data + skb->len */
+	*insn++ = BPF_ALU64_REG(BPF_ADD, reg, BPF_REG_AX);
 	/* AX = skb->data_len */
 	*insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF(struct sk_buff, data_len),
 			      BPF_REG_AX, si->src_reg,
 			      offsetof(struct sk_buff, data_len));
-	/* si->dst_reg = skb->data + skb->len - skb->data_len */
-	*insn++ = BPF_ALU64_REG(BPF_SUB, si->dst_reg, BPF_REG_AX);
+
+	/* reg = skb->data + skb->len - skb->data_len */
+	*insn++ = BPF_ALU64_REG(BPF_SUB, reg, BPF_REG_AX);
+
+	if (si->src_reg == si->dst_reg) {
+		/* Restore the saved register */
+		*insn++ = BPF_MOV64_REG(BPF_REG_AX, si->src_reg);
+		*insn++ = BPF_MOV64_REG(si->dst_reg, reg);
+		*insn++ = BPF_LDX_MEM(BPF_DW, reg, BPF_REG_AX, temp_reg_off);
+	}
 
 	return insn;
 }
-- 
2.33.0


  parent reply	other threads:[~2021-10-11 19:17 UTC|newest]

Thread overview: 14+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-10-11 19:16 [PATCH bpf 0/4] bpf, sockmap: fixes stress testing and regression John Fastabend
2021-10-11 19:16 ` [PATCH bpf 1/4] bpf, sockmap: Remove unhash handler for BPF sockmap usage John Fastabend
2021-10-19  7:17   ` Jakub Sitnicki
2021-10-20  5:28     ` John Fastabend
2021-10-20 15:11       ` Jakub Sitnicki
2021-10-20 15:51         ` John Fastabend
2021-10-20 16:35           ` Jakub Sitnicki
2021-10-21 19:24             ` John Fastabend
2021-10-11 19:16 ` [PATCH bpf 2/4] bpf, sockmap: Fix race in ingress receive verdict with redirect to self John Fastabend
2021-10-19  9:16   ` Jakub Sitnicki
2021-10-11 19:16 ` [PATCH bpf 3/4] bpf: sockmap, strparser, and tls are reusing qdisc_skb_cb and colliding John Fastabend
2021-10-19 15:39   ` Jakub Sitnicki
2021-10-11 19:16 ` John Fastabend [this message]
2021-10-23 13:05   ` [PATCH bpf 4/4] bpf, sockmap: sk_skb data_end access incorrect when src_reg = dst_reg Jakub Sitnicki

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20211011191647.418704-5-john.fastabend@gmail.com \
    --to=john.fastabend@gmail.com \
    --cc=bpf@vger.kernel.org \
    --cc=daniel@iogearbox.net \
    --cc=joamaki@gmail.com \
    --cc=netdev@vger.kernel.org \
    --cc=xiyou.wangcong@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).