ipv4: update fib_info_cnt under spinlock protection
Subsystems (3)
PATCH v2 1/1 ipv4: update fib_info_cnt under spinlock protection
From: Eric Dumazet <[email protected]>
In the past, free_fib_info() was supposed to be called under RTNL protection.
This eventually was no longer the case.
Instead of enforcing RTNL it seems we simply can move fib_info_cnt changes to occur when fib_info_lock is held.
v2: David Laight suggested to update fib_info_cnt only when an entry is added/deleted to/from the hash table, as fib_info_cnt is used to make sure hash table size is optimal.
BUG: KCSAN: data-race in fib_create_info / free_fib_info
write to 0xffffffff86e243a0 of 4 bytes by task 26429 on cpu 0:
fib_create_info+0xe78/0x3440 net/ipv4/fib_semantics.c:1428 fib_table_insert+0x148/0x10c0 net/ipv4/fib_trie.c:1224 fib_magic+0x195/0x1e0 net/ipv4/fib_frontend.c:1087 fib_add_ifaddr+0xd0/0x2e0 net/ipv4/fib_frontend.c:1109 fib_netdev_event+0x178/0x510 net/ipv4/fib_frontend.c:1466 notifier_call_chain kernel/notifier.c:83 [inline] raw_notifier_call_chain+0x53/0xb0 kernel/notifier.c:391 __dev_notify_flags+0x1d3/0x3b0 dev_change_flags+0xa2/0xc0 net/core/dev.c:8872 do_setlink+0x810/0x2410 net/core/rtnetlink.c:2719 rtnl_group_changelink net/core/rtnetlink.c:3242 [inline] __rtnl_newlink net/core/rtnetlink.c:3396 [inline] rtnl_newlink+0xb10/0x13b0 net/core/rtnetlink.c:3506 rtnetlink_rcv_msg+0x745/0x7e0 net/core/rtnetlink.c:5571 netlink_rcv_skb+0x14e/0x250 net/netlink/af_netlink.c:2496 rtnetlink_rcv+0x18/0x20 net/core/rtnetlink.c:5589 netlink_unicast_kernel net/netlink/af_netlink.c:1319 [inline] netlink_unicast+0x5fc/0x6c0 net/netlink/af_netlink.c:1345 netlink_sendmsg+0x726/0x840 net/netlink/af_netlink.c:1921 sock_sendmsg_nosec net/socket.c:704 [inline] sock_sendmsg net/socket.c:724 [inline] ____sys_sendmsg+0x39a/0x510 net/socket.c:2409 ___sys_sendmsg net/socket.c:2463 [inline] __sys_sendmsg+0x195/0x230 net/socket.c:2492 __do_sys_sendmsg net/socket.c:2501 [inline] __se_sys_sendmsg net/socket.c:2499 [inline] __x64_sys_sendmsg+0x42/0x50 net/socket.c:2499 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x44/0xd0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x44/0xae
read to 0xffffffff86e243a0 of 4 bytes by task 31505 on cpu 1:
free_fib_info+0x35/0x80 net/ipv4/fib_semantics.c:252 fib_info_put include/net/ip_fib.h:575 [inline] nsim_fib4_rt_destroy drivers/net/netdevsim/fib.c:294 [inline] nsim_fib4_rt_replace drivers/net/netdevsim/fib.c:403 [inline] nsim_fib4_rt_insert drivers/net/netdevsim/fib.c:431 [inline] nsim_fib4_event drivers/net/netdevsim/fib.c:461 [inline] nsim_fib_event drivers/net/netdevsim/fib.c:881 [inline] nsim_fib_event_work+0x15ca/0x2cf0 drivers/net/netdevsim/fib.c:1477 process_one_work+0x3fc/0x980 kernel/workqueue.c:2298 process_scheduled_works kernel/workqueue.c:2361 [inline] worker_thread+0x7df/0xa70 kernel/workqueue.c:2447 kthread+0x2c7/0x2e0 kernel/kthread.c:327 ret_from_fork+0x1f/0x30
value changed: 0x00000d2d -> 0x00000d2e
Reported by Kernel Concurrency Sanitizer on:
CPU: 1 PID: 31505 Comm: kworker/1:21 Not tainted 5.16.0-rc6-syzkaller #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 Workqueue: events nsim_fib_event_work
--- net/ipv4/fib_semantics.c | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-)
net/ipv4/fib_semantics.c +8 -3
@@ -249,7 +249,6 @@ void free_fib_info(struct fib_info *fi)pr_warn ("Freeing alive fib_info %p\n" , fi);return ; } - fib_info_cnt--;call_rcu (&fi->rcu, free_fib_info_rcu); }
@@ -260,6 +259,10 @@ void fib_release_info(struct fib_info *fi)spin_lock_bh (&fib_info_lock);if (fi &&refcount_dec_and_test (&fi->fib_treeref)) {hlist_del (&fi->fib_hash); + +/* Paired with READ_ONCE() in fib_create_info(). */ +WRITE_ONCE (fib_info_cnt, fib_info_cnt -1 ); +if (fi->fib_prefsrc)hlist_del (&fi->fib_lhash);if (fi->nh) {
@@ -1430,7 +1433,9 @@ struct fib_info *fib_create_info(struct fib_config *cfg,#endif err = -ENOBUFS ; -if (fib_info_cnt >= fib_info_hash_size) { + +/* Paired with WRITE_ONCE() in fib_release_info() */ +if (READ_ONCE (fib_info_cnt) >= fib_info_hash_size) {unsigned int new_size = fib_info_hash_size <<1 ;struct hlist_head *new_info_hash;struct hlist_head *new_laddrhash;
@@ -1462,7 +1467,6 @@ struct fib_info *fib_create_info(struct fib_config *cfg,return ERR_PTR (err); } - fib_info_cnt++; fi->fib_net = net; fi->fib_protocol = cfg->fc_protocol; fi->fib_scope = cfg->fc_scope;
@@ -1591,6 +1595,7 @@ struct fib_info *fib_create_info(struct fib_config *cfg,refcount_set (&fi->fib_treeref,1 );refcount_set (&fi->fib_clntref,1 );spin_lock_bh (&fib_info_lock); + fib_info_cnt++;hlist_add_head (&fi->fib_hash, &fib_info_hash[fib_info_hashfn (fi)]);if (fi->fib_prefsrc) {
Signature · 1 line
2.34.1.703.g22d0c6ccf7-goog
Quoted 15 lines · Eric Dumazet
From: Eric Dumazet <[email protected]>
In the past,
free_fib_info()was supposed to be called under RTNL protection.This eventually was no longer the case.
Instead of enforcing RTNL it seems we simply can move
fib_info_cntchanges to occur whenfib_info_lockis held.v2: David Laight suggested to update
fib_info_cntonly when an entry is added/deleted to/from the hash table, asfib_info_cntis used to make sure hash table size is optimal.
[...]
Quoted 6 lines
Fixes 48bb9eb47b27 ("netdevsim: fib: Add dummy implementation for FIB offload")Signed-off-by Eric Dumazet <[email protected]>Reported-by syzbot <[email protected]>Cc David Laight <[email protected]>Cc Ido Schimmel <[email protected]>Cc Jiri Pirko <[email protected]>
This patch was applied to netdev/net.git (master) by David S. Miller <[email protected]>:
Hello:
This patch was applied to netdev/net.git (master)
by David S. Miller <[email protected]>:
Quoted 8 lines · On Sun, 16 Jan 2022 01:02:20 -0800 you wrote:
From: Eric Dumazet <[email protected]>
In the past,
free_fib_info()was supposed to be called under RTNL protection.This eventually was no longer the case.
[...]
Here is the summary with links:
- [v2,net] ipv4: update fib_info_cnt under spinlock protection
https://git.kernel.org/netdev/net/c/0a6e6b3c7db6
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html
From: Eric Dumazet
Quoted 7 lines
Sent: 16 January 2022 09:02
In the past,
free_fib_info()was supposed to be called under RTNL protection.This eventually was no longer the case.
Quoted 8 lines
Instead of enforcing RTNL it seems we simply can move
fib_info_cntchanges to occur whenfib_info_lockis held.v2: David Laight suggested to update
fib_info_cntonly when an entry is added/deleted to/from the hash table, asfib_info_cntis used to make sure hash table size is optimal.
Already applied, but
acked-by: David Laight
...
If you are going to add READ_ONCE() markers then one on 'fib_info_hash_size' would be much more appropriate since this value is used twice.
net/ipv4/fib_semantics.c struct fib_info *fib_create_info(struct fib_config *cfg,
Quoted 9 lines
net/ipv4/fib_semantics.c +3
-1err = -ENOBUFS ; -if (fib_info_cnt >= fib_info_hash_size) { + +/* Paired with WRITE_ONCE() in fib_release_info() */ +if (READ_ONCE (fib_info_cnt) >= fib_info_hash_size) {unsigned int new_size = fib_info_hash_size <<1 ;struct hlist_head *new_info_hash;struct hlist_head *new_laddrhash;@@ -1462,7 +1467,6 @@ struct fib_info *fib_create_info(struct fib_config *cfg,
If is also possible for two (or many) threads to decide to increase the hash table size at the same time.
The code that moves the items to the new hash tables should probably discard the new tables is they aren't larger than the existing ones.
The copy does look safe - just a waste of time.
It is also technically possible (but very unlikely) that the table will get shrunk!
It will grow again on the next allocate.
But this is a different bug.
I also though Linus said that the WRITE_ONCE() weren't needed here because the kernel basically assumes the compiler isn't stupid enough to do 'write tearing' on word sized items (or just write zero before every write).
David
-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)
Quoted 44 lines · David Laight
From: Eric Dumazet
Quoted 7 lines
Sent: 16 January 2022 09:02
In the past,
free_fib_info()was supposed to be called under RTNL protection.This eventually was no longer the case.
Quoted 8 lines
Instead of enforcing RTNL it seems we simply can move
fib_info_cntchanges to occur whenfib_info_lockis held.v2: David Laight suggested to update
fib_info_cntonly when an entry is added/deleted to/from the hash table, asfib_info_cntis used to make sure hash table size is optimal.Already applied, but
acked-by: David Laight...
If you are going to addREAD_ONCE()markers then one on 'fib_info_hash_size' would be much more appropriate since this value is used twice.net/ipv4/fib_semantics.c struct fib_info *fib_create_info(struct fib_config *cfg,
Quoted 9 lines
err = -ENOBUFS;net/ipv4/fib_semantics.c +3
-1-if (fib_info_cnt >= fib_info_hash_size) { + +/* Paired with WRITE_ONCE() in fib_release_info() */ +if (READ_ONCE (fib_info_cnt) >= fib_info_hash_size) {unsigned int new_size = fib_info_hash_size <<1 ;struct hlist_head *new_info_hash;struct hlist_head *new_laddrhash;@@ -1462,7 +1467,6 @@ struct fib_info *fib_create_info(struct fib_config *cfg,If is also possible for two (or many) threads to decide to increase the hash table size at the same time.
The code that moves the items to the new hash tables should probably discard the new tables is they aren't larger than the existing ones.
The copy does look safe - just a waste of time.
Quoted 6 lines
It is also technically possible (but very unlikely) that the table will get shrunk!
It will grow again on the next allocate.But this is a different bug.
There is no bug.
fib_create_info() is called with RTNL held.
Quoted 5 lines
I also though Linus said that the
WRITE_ONCE()weren't needed here because the kernel basically assumes the compiler isn't stupid enough to do 'write tearing' on word sized items (or just write zero before every write).
That is not true. READ_ONCE()/WRITE_ONCE() have their own purpose, we can not assume a compiler will follow arbitrary rules about word sizes.
From: Eric Dumazet
Quoted 4 lines
Sent: 16 January 2022 09:02
From: Eric Dumazet <[email protected]>
Quoted 8 lines
In the past,
free_fib_info()was supposed to be called under RTNL protection.This eventually was no longer the case.
Instead of enforcing RTNL it seems we simply can move
fib_info_cntchanges to occur whenfib_info_lockis held.
This probably ought to be a stable candidate.
If an increment is lost due to the unlocked inc/dec it is possible for the count to wrap to -1 if all fib are deleted. That will cause the table size to be doubled on every allocate (until the count goes +ve again).
Losing a decrement is less of a problem. You'd need to lose a lot of them for any ill effect.
David
-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)
Quoted 10 lines · David Laight
This probably ought to be a stable candidate.
If an increment is lost due to the unlocked inc/dec it is possible for the count to wrap to -1 if all fib are deleted. That will cause the table size to be doubled on every allocate (until the count goes +ve again).
Losing a decrement is less of a problem. You'd need to lose a lot of them for any ill effect.
You are rephrasing the syzbot report, and the Fixes: tag I added in the changelog.
hint: netdev maintainers handle the stable backports based on the
Fixes: tag, and their
own judgement.