Skip to contentSkip to conversation

tcp: fix slab-out-of-bounds in tcp_zerocopy_receive()

PATCH In review Eric Dumazet 4 participants 4 messages Oct 3, 2019

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 tcp: fix slab-out-of-bounds in tcp_zerocopy_receive()

Apparently a refactoring patch brought a bug, that was caught by syzbot [1]

Original code was correct, do not try to be smarter than the compiler :/

[1]

BUG: KASAN: slab-out-of-bounds in tcp_zerocopy_receive net/ipv4/tcp.c:1807 [inline]
BUG: KASAN: slab-out-of-bounds in do_tcp_getsockopt.isra.0+0x2c6c/0x3120 net/ipv4/tcp.c:3654
Read of size 4 at addr ffff8880943cf188 by task syz-executor.2/17508

CPU: 0 PID: 17508 Comm: syz-executor.2 Not tainted 5.3.0-rc7+ #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Call Trace:
 __dump_stack lib/dump_stack.c:77 [inline]
 dump_stack+0x172/0x1f0 lib/dump_stack.c:113
 print_address_description.cold+0xd4/0x306 mm/kasan/report.c:351
 __kasan_report.cold+0x1b/0x36 mm/kasan/report.c:482
 kasan_report+0x12/0x17 mm/kasan/common.c:618
 __asan_report_load4_noabort+0x14/0x20 mm/kasan/generic_report.c:131
 tcp_zerocopy_receive net/ipv4/tcp.c:1807 [inline]
 do_tcp_getsockopt.isra.0+0x2c6c/0x3120 net/ipv4/tcp.c:3654
 tcp_getsockopt+0xbf/0xe0 net/ipv4/tcp.c:3680
 sock_common_getsockopt+0x94/0xd0 net/core/sock.c:3098
 __sys_getsockopt+0x16d/0x310 net/socket.c:2129
 __do_sys_getsockopt net/socket.c:2144 [inline]
 __se_sys_getsockopt net/socket.c:2141 [inline]
 __x64_sys_getsockopt+0xbe/0x150 net/socket.c:2141
 do_syscall_64+0xfd/0x6a0 arch/x86/entry/common.c:296
Fixes d8e18a516f8f ("net: Use skb accessors in network core")
Cc Matthew Wilcox (Oracle) <[email protected]>
---
 net/ipv4/tcp.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
net/ipv4/tcp.c +2 -4
index 79c325a07ba5dc7cfad0a846d1f03bf1787f840b..f98a1882e537dca0102e829cb349be50302d83ab 100644
@@ -1798,13 +1798,11 @@ static int tcp_zerocopy_receive(struct sock *sk,
 		}
 		if (skb_frag_size(frags) != PAGE_SIZE || skb_frag_off(frags)) {
 			int remaining = zc->recv_skip_hint;
-			int size = skb_frag_size(frags);
 
-			while (remaining && (size != PAGE_SIZE ||
+			while (remaining && (skb_frag_size(frags) != PAGE_SIZE ||
 					     skb_frag_off(frags))) {
-				remaining -= size;
+				remaining -= skb_frag_size(frags);
 				frags++;
-				size = skb_frag_size(frags);
 			}
 			zc->recv_skip_hint -= remaining;
 			break;
Signature · 1 line
2.23.0.581.g78d2f28ef7-goog
#1 lore raw
Quoted 2 lines · Eric Dumazet

Apparently a refactoring patch brought a bug, that was caught by syzbot [1]

That wasn't refactoring. As you know (because we talked about it at LSFMM), this is an enabling patch for supporting hch's work to fix get_user_pages().

Quoted 2 lines

Original code was correct, do not try to be smarter than the compiler :/

That wasn't an attempt to be smarter than the compiler. I was trying to keep the line length below 80 columns. Which you probably now see that you haven't done.

I must have a blind spot here. I can't see the difference between the two versions.

Quoted 20 lines

+++ b/net/ipv4/tcp.c

net/ipv4/tcp.c +2 -4
@@ -1798,13 +1798,11 @@ static int tcp_zerocopy_receive(struct sock *sk,
 		}
 		if (skb_frag_size(frags) != PAGE_SIZE || skb_frag_off(frags)) {
 			int remaining = zc->recv_skip_hint;
-			int size = skb_frag_size(frags);
 
-			while (remaining && (size != PAGE_SIZE ||
+			while (remaining && (skb_frag_size(frags) != PAGE_SIZE ||
 					     skb_frag_off(frags))) {
-				remaining -= size;
+				remaining -= skb_frag_size(frags);
 				frags++;
-				size = skb_frag_size(frags);
 			}
 			zc->recv_skip_hint -= remaining;
 			break;
Signature · 1 line
2.23.0.581.g78d2f28ef7-goog
#2 lore raw
Quoted 7 lines · Matthew Wilcox
Quoted 2 lines · Eric Dumazet

Apparently a refactoring patch brought a bug, that was caught by syzbot [1]

That wasn't refactoring. As you know (because we talked about it at LSFMM), this is an enabling patch for supporting hch's work to fix get_user_pages().

Changing frags->size by skb_frag_size(frags) is what I call refactoring.

This was claimed in the changelog of your patch :

<quote>
net: Use skb accessors in network core

In preparation for unifying the skb_frag and bio_vec, use the fine accessors which already exist and use skb_frag_t instead of struct skb_frag_struct.
</quote>

I guess David trusted you enough and did not checked that you made other changes.

Quoted 7 lines
Quoted 3 lines

Original code was correct, do not try to be smarter than the compiler :/

That wasn't an attempt to be smarter than the compiler. I was trying to keep the line length below 80 columns. Which you probably now see that you haven't done.

Seriously we do not care of the 80 column rule in this particular function, we care about code correctness first.

We do not mix fixes and cleanups in the same patch.

Quoted 3 lines

I must have a blind spot here. I can't see the difference between the two versions.

No worries.

The major difference is that with your code, when @remaining reaches zero, it fetches an extra "size = skb_frag_size(frags);" at line 1807 as the syzbot report correctly points.

MAX_SKB_FRAGS is 17 on x86, meaning struct skb_shared_info frags[] array has 17 elements.

Accessing the 18th element can trigger a page fault, if struct skb_shared_info sits exactly at the end of a page.

while (remaining && (size != PAGE_SIZE ||
skb_frag_off(frags))) {

remaining -= size; // when this reaches 0

frags++; // this points to the next fragment in the array, not initialized.

size = skb_frag_size(frags); // Crash here if crossing a page and next page is unmapped.
}

Original code from Soheil was correct :

while (remaining && (frags->size != PAGE_SIZE ||

		     frags->page_offset)) {
	remaining -= frags->size;
	frags++;
}

My patch simply restore this code, keeping the accessors changes which were perfectly fine, thank you.

maintainer
#3 lore raw
In reply to Eric Dumazet

From: Eric Dumazet <[email protected]>
Date: Wed, 2 Oct 2019 20:19:59 -0700

Quoted 7 lines

Apparently a refactoring patch brought a bug, that was caught by syzbot [1]

Original code was correct, do not try to be smarter than the compiler :/

[1]

...

Quoted 5 lines
Fixes d8e18a516f8f ("net: Use skb accessors in network core")
Cc Matthew Wilcox (Oracle) <[email protected]>

Applied and queued up for -stable.