From 340c3d337119ea177a98338be2e3bc62ee87ac80 Mon Sep 17 00:00:00 2001 From: Kuniyuki Iwashima Date: Tue, 21 Jun 2022 10:19:08 -0700 Subject: af_unix: Clean up some sock_net() uses. Some functions define a net pointer only for one-shot use. Others call sock_net() redundantly even when a net pointer is available. Let's fix these and make the code simpler. Signed-off-by: Kuniyuki Iwashima Signed-off-by: David S. Miller --- net/unix/diag.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'net/unix/diag.c') diff --git a/net/unix/diag.c b/net/unix/diag.c index bb0b5ea1655f..4e3dc8179fa4 100644 --- a/net/unix/diag.c +++ b/net/unix/diag.c @@ -308,7 +308,6 @@ out_nosk: static int unix_diag_handler_dump(struct sk_buff *skb, struct nlmsghdr *h) { int hdrlen = sizeof(struct unix_diag_req); - struct net *net = sock_net(skb->sk); if (nlmsg_len(h) < hdrlen) return -EINVAL; @@ -317,7 +316,7 @@ static int unix_diag_handler_dump(struct sk_buff *skb, struct nlmsghdr *h) struct netlink_dump_control c = { .dump = unix_diag_dump, }; - return netlink_dump_start(net->diag_nlsk, skb, h, &c); + return netlink_dump_start(sock_net(skb->sk)->diag_nlsk, skb, h, &c); } else return unix_diag_get_exact(skb, h, nlmsg_data(h)); } -- cgit From f302d180c6d430ea99643b9b2b3407aedaa36703 Mon Sep 17 00:00:00 2001 From: Kuniyuki Iwashima Date: Tue, 21 Jun 2022 10:19:09 -0700 Subject: af_unix: Include the whole hash table size in UNIX_HASH_SIZE. Currently, the size of AF_UNIX hash table is UNIX_HASH_SIZE * 2, the first half for bind()ed sockets and the second half for unbound ones. UNIX_HASH_SIZE * 2 is used to define the table and iterate over it. In some places, we use ARRAY_SIZE(unix_socket_table) instead of UNIX_HASH_SIZE * 2. However, we cannot use it anymore because we will allocate the hash table dynamically. Then, we would have to add UNIX_HASH_SIZE * 2 in many places, which would be troublesome. This patch adapts the UNIX_HASH_SIZE definition to include bound and unbound sockets and defines a new UNIX_HASH_MOD macro to ease calculations. Signed-off-by: Kuniyuki Iwashima Signed-off-by: David S. Miller --- net/unix/diag.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) (limited to 'net/unix/diag.c') diff --git a/net/unix/diag.c b/net/unix/diag.c index 4e3dc8179fa4..c5d1cca72aa5 100644 --- a/net/unix/diag.c +++ b/net/unix/diag.c @@ -204,9 +204,7 @@ static int unix_diag_dump(struct sk_buff *skb, struct netlink_callback *cb) s_slot = cb->args[0]; num = s_num = cb->args[1]; - for (slot = s_slot; - slot < ARRAY_SIZE(unix_socket_table); - s_num = 0, slot++) { + for (slot = s_slot; slot < UNIX_HASH_SIZE; s_num = 0, slot++) { struct sock *sk; num = 0; @@ -242,7 +240,7 @@ static struct sock *unix_lookup_by_ino(unsigned int ino) struct sock *sk; int i; - for (i = 0; i < ARRAY_SIZE(unix_socket_table); i++) { + for (i = 0; i < UNIX_HASH_SIZE; i++) { spin_lock(&unix_table_locks[i]); sk_for_each(sk, &unix_socket_table[i]) if (ino == sock_i_ino(sk)) { -- cgit From 79b05beaa5c340e1649dc7462e056ae33b916131 Mon Sep 17 00:00:00 2001 From: Kuniyuki Iwashima Date: Tue, 21 Jun 2022 10:19:11 -0700 Subject: af_unix: Acquire/Release per-netns hash table's locks. This commit adds extra spin_lock/spin_unlock() for a per-netns hash table inside the existing ones for unix_table_locks. As of this commit, sockets are still linked in the global hash table. After putting sockets in a per-netns hash table and removing the old one in the next patch, we remove the global locks in the last patch. Signed-off-by: Kuniyuki Iwashima Signed-off-by: David S. Miller --- net/unix/diag.c | 23 +++++++++++++++-------- 1 file changed, 15 insertions(+), 8 deletions(-) (limited to 'net/unix/diag.c') diff --git a/net/unix/diag.c b/net/unix/diag.c index c5d1cca72aa5..7fc377435114 100644 --- a/net/unix/diag.c +++ b/net/unix/diag.c @@ -195,9 +195,9 @@ static int sk_diag_dump(struct sock *sk, struct sk_buff *skb, struct unix_diag_r static int unix_diag_dump(struct sk_buff *skb, struct netlink_callback *cb) { - struct unix_diag_req *req; - int num, s_num, slot, s_slot; struct net *net = sock_net(skb->sk); + int num, s_num, slot, s_slot; + struct unix_diag_req *req; req = nlmsg_data(cb->nlh); @@ -209,6 +209,7 @@ static int unix_diag_dump(struct sk_buff *skb, struct netlink_callback *cb) num = 0; spin_lock(&unix_table_locks[slot]); + spin_lock(&net->unx.table.locks[slot]); sk_for_each(sk, &unix_socket_table[slot]) { if (!net_eq(sock_net(sk), net)) continue; @@ -220,12 +221,14 @@ static int unix_diag_dump(struct sk_buff *skb, struct netlink_callback *cb) NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq, NLM_F_MULTI) < 0) { + spin_unlock(&net->unx.table.locks[slot]); spin_unlock(&unix_table_locks[slot]); goto done; } next: num++; } + spin_unlock(&net->unx.table.locks[slot]); spin_unlock(&unix_table_locks[slot]); } done: @@ -235,19 +238,22 @@ done: return skb->len; } -static struct sock *unix_lookup_by_ino(unsigned int ino) +static struct sock *unix_lookup_by_ino(struct net *net, unsigned int ino) { struct sock *sk; int i; for (i = 0; i < UNIX_HASH_SIZE; i++) { spin_lock(&unix_table_locks[i]); + spin_lock(&net->unx.table.locks[i]); sk_for_each(sk, &unix_socket_table[i]) if (ino == sock_i_ino(sk)) { sock_hold(sk); + spin_unlock(&net->unx.table.locks[i]); spin_unlock(&unix_table_locks[i]); return sk; } + spin_unlock(&net->unx.table.locks[i]); spin_unlock(&unix_table_locks[i]); } return NULL; @@ -257,16 +263,17 @@ static int unix_diag_get_exact(struct sk_buff *in_skb, const struct nlmsghdr *nlh, struct unix_diag_req *req) { - int err = -EINVAL; - struct sock *sk; - struct sk_buff *rep; - unsigned int extra_len; struct net *net = sock_net(in_skb->sk); + unsigned int extra_len; + struct sk_buff *rep; + struct sock *sk; + int err; + err = -EINVAL; if (req->udiag_ino == 0) goto out_nosk; - sk = unix_lookup_by_ino(req->udiag_ino); + sk = unix_lookup_by_ino(net, req->udiag_ino); err = -ENOENT; if (sk == NULL) goto out_nosk; -- cgit From cf2f225e2653734e66e91c09e1cbe004bfd3d4a7 Mon Sep 17 00:00:00 2001 From: Kuniyuki Iwashima Date: Tue, 21 Jun 2022 10:19:12 -0700 Subject: af_unix: Put a socket into a per-netns hash table. This commit replaces the global hash table with a per-netns one and removes the global one. We now link a socket in each netns's hash table so we can save some netns comparisons when iterating through a hash bucket. Signed-off-by: Kuniyuki Iwashima Signed-off-by: David S. Miller --- net/unix/diag.c | 9 +++------ 1 file changed, 3 insertions(+), 6 deletions(-) (limited to 'net/unix/diag.c') diff --git a/net/unix/diag.c b/net/unix/diag.c index 7fc377435114..4d0f0ca6a1eb 100644 --- a/net/unix/diag.c +++ b/net/unix/diag.c @@ -210,9 +210,7 @@ static int unix_diag_dump(struct sk_buff *skb, struct netlink_callback *cb) num = 0; spin_lock(&unix_table_locks[slot]); spin_lock(&net->unx.table.locks[slot]); - sk_for_each(sk, &unix_socket_table[slot]) { - if (!net_eq(sock_net(sk), net)) - continue; + sk_for_each(sk, &net->unx.table.buckets[slot]) { if (num < s_num) goto next; if (!(req->udiag_states & (1 << sk->sk_state))) @@ -246,13 +244,14 @@ static struct sock *unix_lookup_by_ino(struct net *net, unsigned int ino) for (i = 0; i < UNIX_HASH_SIZE; i++) { spin_lock(&unix_table_locks[i]); spin_lock(&net->unx.table.locks[i]); - sk_for_each(sk, &unix_socket_table[i]) + sk_for_each(sk, &net->unx.table.buckets[i]) { if (ino == sock_i_ino(sk)) { sock_hold(sk); spin_unlock(&net->unx.table.locks[i]); spin_unlock(&unix_table_locks[i]); return sk; } + } spin_unlock(&net->unx.table.locks[i]); spin_unlock(&unix_table_locks[i]); } @@ -277,8 +276,6 @@ static int unix_diag_get_exact(struct sk_buff *in_skb, err = -ENOENT; if (sk == NULL) goto out_nosk; - if (!net_eq(sock_net(sk), net)) - goto out; err = sock_diag_check_cookie(sk, req->udiag_cookie); if (err) -- cgit From 2f7ca90a0188b57a54d3b1159eb7874427a7e07a Mon Sep 17 00:00:00 2001 From: Kuniyuki Iwashima Date: Tue, 21 Jun 2022 10:19:13 -0700 Subject: af_unix: Remove unix_table_locks. unix_table_locks are to protect the global hash table, unix_socket_table. The previous commit removed it, so let's clean up the unnecessary locks. Here is a test result on EC2 c5.9xlarge where 10 processes run concurrently in different netns and bind 100,000 sockets for each. without this series : 1m 38s with this series : 11s It is ~10x faster because the global hash table is split into 10 netns in this case. Signed-off-by: Kuniyuki Iwashima Signed-off-by: David S. Miller --- net/unix/diag.c | 8 +------- 1 file changed, 1 insertion(+), 7 deletions(-) (limited to 'net/unix/diag.c') diff --git a/net/unix/diag.c b/net/unix/diag.c index 4d0f0ca6a1eb..105f522a89fe 100644 --- a/net/unix/diag.c +++ b/net/unix/diag.c @@ -13,7 +13,7 @@ static int sk_diag_dump_name(struct sock *sk, struct sk_buff *nlskb) { - /* might or might not have unix_table_locks */ + /* might or might not have a hash table lock */ struct unix_address *addr = smp_load_acquire(&unix_sk(sk)->addr); if (!addr) @@ -208,7 +208,6 @@ static int unix_diag_dump(struct sk_buff *skb, struct netlink_callback *cb) struct sock *sk; num = 0; - spin_lock(&unix_table_locks[slot]); spin_lock(&net->unx.table.locks[slot]); sk_for_each(sk, &net->unx.table.buckets[slot]) { if (num < s_num) @@ -220,14 +219,12 @@ static int unix_diag_dump(struct sk_buff *skb, struct netlink_callback *cb) cb->nlh->nlmsg_seq, NLM_F_MULTI) < 0) { spin_unlock(&net->unx.table.locks[slot]); - spin_unlock(&unix_table_locks[slot]); goto done; } next: num++; } spin_unlock(&net->unx.table.locks[slot]); - spin_unlock(&unix_table_locks[slot]); } done: cb->args[0] = slot; @@ -242,18 +239,15 @@ static struct sock *unix_lookup_by_ino(struct net *net, unsigned int ino) int i; for (i = 0; i < UNIX_HASH_SIZE; i++) { - spin_lock(&unix_table_locks[i]); spin_lock(&net->unx.table.locks[i]); sk_for_each(sk, &net->unx.table.buckets[i]) { if (ino == sock_i_ino(sk)) { sock_hold(sk); spin_unlock(&net->unx.table.locks[i]); - spin_unlock(&unix_table_locks[i]); return sk; } } spin_unlock(&net->unx.table.locks[i]); - spin_unlock(&unix_table_locks[i]); } return NULL; } -- cgit From b3abe42e94900bdd045c472f9c9be620ba5ce553 Mon Sep 17 00:00:00 2001 From: Kuniyuki Iwashima Date: Sun, 27 Nov 2022 10:24:11 +0900 Subject: af_unix: Get user_ns from in_skb in unix_diag_get_exact(). Wei Chen reported a NULL deref in sk_user_ns() [0][1], and Paolo diagnosed the root cause: in unix_diag_get_exact(), the newly allocated skb does not have sk. [2] We must get the user_ns from the NETLINK_CB(in_skb).sk and pass it to sk_diag_fill(). [0]: BUG: kernel NULL pointer dereference, address: 0000000000000270 #PF: supervisor read access in kernel mode #PF: error_code(0x0000) - not-present page PGD 12bbce067 P4D 12bbce067 PUD 12bc40067 PMD 0 Oops: 0000 [#1] PREEMPT SMP CPU: 0 PID: 27942 Comm: syz-executor.0 Not tainted 6.1.0-rc5-next-20221118 #2 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.13.0-48-gd9c812dda519-prebuilt.qemu.org 04/01/2014 RIP: 0010:sk_user_ns include/net/sock.h:920 [inline] RIP: 0010:sk_diag_dump_uid net/unix/diag.c:119 [inline] RIP: 0010:sk_diag_fill+0x77d/0x890 net/unix/diag.c:170 Code: 89 ef e8 66 d4 2d fd c7 44 24 40 00 00 00 00 49 8d 7c 24 18 e8 54 d7 2d fd 49 8b 5c 24 18 48 8d bb 70 02 00 00 e8 43 d7 2d fd <48> 8b 9b 70 02 00 00 48 8d 7b 10 e8 33 d7 2d fd 48 8b 5b 10 48 8d RSP: 0018:ffffc90000d67968 EFLAGS: 00010246 RAX: ffff88812badaa48 RBX: 0000000000000000 RCX: ffffffff840d481d RDX: 0000000000000465 RSI: 0000000000000000 RDI: 0000000000000270 RBP: ffffc90000d679a8 R08: 0000000000000277 R09: 0000000000000000 R10: 0001ffffffffffff R11: 0001c90000d679a8 R12: ffff88812ac03800 R13: ffff88812c87c400 R14: ffff88812ae42210 R15: ffff888103026940 FS: 00007f08b4e6f700(0000) GS:ffff88813bc00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000270 CR3: 000000012c58b000 CR4: 00000000003506f0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: unix_diag_get_exact net/unix/diag.c:285 [inline] unix_diag_handler_dump+0x3f9/0x500 net/unix/diag.c:317 __sock_diag_cmd net/core/sock_diag.c:235 [inline] sock_diag_rcv_msg+0x237/0x250 net/core/sock_diag.c:266 netlink_rcv_skb+0x13e/0x250 net/netlink/af_netlink.c:2564 sock_diag_rcv+0x24/0x40 net/core/sock_diag.c:277 netlink_unicast_kernel net/netlink/af_netlink.c:1330 [inline] netlink_unicast+0x5e9/0x6b0 net/netlink/af_netlink.c:1356 netlink_sendmsg+0x739/0x860 net/netlink/af_netlink.c:1932 sock_sendmsg_nosec net/socket.c:714 [inline] sock_sendmsg net/socket.c:734 [inline] ____sys_sendmsg+0x38f/0x500 net/socket.c:2476 ___sys_sendmsg net/socket.c:2530 [inline] __sys_sendmsg+0x197/0x230 net/socket.c:2559 __do_sys_sendmsg net/socket.c:2568 [inline] __se_sys_sendmsg net/socket.c:2566 [inline] __x64_sys_sendmsg+0x42/0x50 net/socket.c:2566 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x2b/0x70 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd RIP: 0033:0x4697f9 Code: f7 d8 64 89 02 b8 ff ff ff ff c3 66 0f 1f 44 00 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 bc ff ff ff f7 d8 64 89 01 48 RSP: 002b:00007f08b4e6ec48 EFLAGS: 00000246 ORIG_RAX: 000000000000002e RAX: ffffffffffffffda RBX: 000000000077bf80 RCX: 00000000004697f9 RDX: 0000000000000000 RSI: 00000000200001c0 RDI: 0000000000000003 RBP: 00000000004d29e9 R08: 0000000000000000 R09: 0000000000000000 R10: 0000000000000000 R11: 0000000000000246 R12: 000000000077bf80 R13: 0000000000000000 R14: 000000000077bf80 R15: 00007ffdb36bc6c0 Modules linked in: CR2: 0000000000000270 [1]: https://lore.kernel.org/netdev/CAO4mrfdvyjFpokhNsiwZiP-wpdSD0AStcJwfKcKQdAALQ9_2Qw@mail.gmail.com/ [2]: https://lore.kernel.org/netdev/e04315e7c90d9a75613f3993c2baf2d344eef7eb.camel@redhat.com/ Fixes: cae9910e7344 ("net: Add UNIX_DIAG_UID to Netlink UNIX socket diagnostics.") Reported-by: syzbot Reported-by: Wei Chen Diagnosed-by: Paolo Abeni Signed-off-by: Kuniyuki Iwashima Signed-off-by: Paolo Abeni --- net/unix/diag.c | 20 ++++++++++++-------- 1 file changed, 12 insertions(+), 8 deletions(-) (limited to 'net/unix/diag.c') diff --git a/net/unix/diag.c b/net/unix/diag.c index 105f522a89fe..616b55c5b890 100644 --- a/net/unix/diag.c +++ b/net/unix/diag.c @@ -114,14 +114,16 @@ static int sk_diag_show_rqlen(struct sock *sk, struct sk_buff *nlskb) return nla_put(nlskb, UNIX_DIAG_RQLEN, sizeof(rql), &rql); } -static int sk_diag_dump_uid(struct sock *sk, struct sk_buff *nlskb) +static int sk_diag_dump_uid(struct sock *sk, struct sk_buff *nlskb, + struct user_namespace *user_ns) { - uid_t uid = from_kuid_munged(sk_user_ns(nlskb->sk), sock_i_uid(sk)); + uid_t uid = from_kuid_munged(user_ns, sock_i_uid(sk)); return nla_put(nlskb, UNIX_DIAG_UID, sizeof(uid_t), &uid); } static int sk_diag_fill(struct sock *sk, struct sk_buff *skb, struct unix_diag_req *req, - u32 portid, u32 seq, u32 flags, int sk_ino) + struct user_namespace *user_ns, + u32 portid, u32 seq, u32 flags, int sk_ino) { struct nlmsghdr *nlh; struct unix_diag_msg *rep; @@ -167,7 +169,7 @@ static int sk_diag_fill(struct sock *sk, struct sk_buff *skb, struct unix_diag_r goto out_nlmsg_trim; if ((req->udiag_show & UDIAG_SHOW_UID) && - sk_diag_dump_uid(sk, skb)) + sk_diag_dump_uid(sk, skb, user_ns)) goto out_nlmsg_trim; nlmsg_end(skb, nlh); @@ -179,7 +181,8 @@ out_nlmsg_trim: } static int sk_diag_dump(struct sock *sk, struct sk_buff *skb, struct unix_diag_req *req, - u32 portid, u32 seq, u32 flags) + struct user_namespace *user_ns, + u32 portid, u32 seq, u32 flags) { int sk_ino; @@ -190,7 +193,7 @@ static int sk_diag_dump(struct sock *sk, struct sk_buff *skb, struct unix_diag_r if (!sk_ino) return 0; - return sk_diag_fill(sk, skb, req, portid, seq, flags, sk_ino); + return sk_diag_fill(sk, skb, req, user_ns, portid, seq, flags, sk_ino); } static int unix_diag_dump(struct sk_buff *skb, struct netlink_callback *cb) @@ -214,7 +217,7 @@ static int unix_diag_dump(struct sk_buff *skb, struct netlink_callback *cb) goto next; if (!(req->udiag_states & (1 << sk->sk_state))) goto next; - if (sk_diag_dump(sk, skb, req, + if (sk_diag_dump(sk, skb, req, sk_user_ns(skb->sk), NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq, NLM_F_MULTI) < 0) { @@ -282,7 +285,8 @@ again: if (!rep) goto out; - err = sk_diag_fill(sk, rep, req, NETLINK_CB(in_skb).portid, + err = sk_diag_fill(sk, rep, req, sk_user_ns(NETLINK_CB(in_skb).sk), + NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, 0, req->udiag_ino); if (err < 0) { nlmsg_free(rep); -- cgit