Update Linux to v5.10.157
Sourced from [1]
[1] https://cdn.kernel.org/pub/linux/kernel/v5.x/linux-5.10.157.tar.xz
Signed-off-by: Olivier Deprez <olivier.deprez@arm.com>
Change-Id: I7b30d9e98d8c465d6b44de8e7433b4a40b3289ba
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index 12dd08a..541758c 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -425,7 +425,7 @@
if (sk->sk_sndbuf < sndmem)
WRITE_ONCE(sk->sk_sndbuf,
- min(sndmem, sock_net(sk)->ipv4.sysctl_tcp_wmem[2]));
+ min(sndmem, READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_wmem[2])));
}
/* 2. Tuning advertised window (window_clamp, rcv_ssthresh)
@@ -454,12 +454,13 @@
*/
/* Slow part of check#2. */
-static int __tcp_grow_window(const struct sock *sk, const struct sk_buff *skb)
+static int __tcp_grow_window(const struct sock *sk, const struct sk_buff *skb,
+ unsigned int skbtruesize)
{
struct tcp_sock *tp = tcp_sk(sk);
/* Optimize this! */
- int truesize = tcp_win_from_space(sk, skb->truesize) >> 1;
- int window = tcp_win_from_space(sk, sock_net(sk)->ipv4.sysctl_tcp_rmem[2]) >> 1;
+ int truesize = tcp_win_from_space(sk, skbtruesize) >> 1;
+ int window = tcp_win_from_space(sk, READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_rmem[2])) >> 1;
while (tp->rcv_ssthresh <= window) {
if (truesize <= skb->len)
@@ -471,7 +472,27 @@
return 0;
}
-static void tcp_grow_window(struct sock *sk, const struct sk_buff *skb)
+/* Even if skb appears to have a bad len/truesize ratio, TCP coalescing
+ * can play nice with us, as sk_buff and skb->head might be either
+ * freed or shared with up to MAX_SKB_FRAGS segments.
+ * Only give a boost to drivers using page frag(s) to hold the frame(s),
+ * and if no payload was pulled in skb->head before reaching us.
+ */
+static u32 truesize_adjust(bool adjust, const struct sk_buff *skb)
+{
+ u32 truesize = skb->truesize;
+
+ if (adjust && !skb_headlen(skb)) {
+ truesize -= SKB_TRUESIZE(skb_end_offset(skb));
+ /* paranoid check, some drivers might be buggy */
+ if (unlikely((int)truesize < (int)skb->len))
+ truesize = skb->truesize;
+ }
+ return truesize;
+}
+
+static void tcp_grow_window(struct sock *sk, const struct sk_buff *skb,
+ bool adjust)
{
struct tcp_sock *tp = tcp_sk(sk);
int room;
@@ -480,15 +501,16 @@
/* Check #1 */
if (room > 0 && !tcp_under_memory_pressure(sk)) {
+ unsigned int truesize = truesize_adjust(adjust, skb);
int incr;
/* Check #2. Increase window, if skb with such overhead
* will fit to rcvbuf in future.
*/
- if (tcp_win_from_space(sk, skb->truesize) <= skb->len)
+ if (tcp_win_from_space(sk, truesize) <= skb->len)
incr = 2 * tp->advmss;
else
- incr = __tcp_grow_window(sk, skb);
+ incr = __tcp_grow_window(sk, skb, truesize);
if (incr) {
incr = max_t(int, incr, 2 * skb->len);
@@ -503,7 +525,7 @@
*/
static void tcp_init_buffer_space(struct sock *sk)
{
- int tcp_app_win = sock_net(sk)->ipv4.sysctl_tcp_app_win;
+ int tcp_app_win = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_app_win);
struct tcp_sock *tp = tcp_sk(sk);
int maxwin;
@@ -543,16 +565,17 @@
struct tcp_sock *tp = tcp_sk(sk);
struct inet_connection_sock *icsk = inet_csk(sk);
struct net *net = sock_net(sk);
+ int rmem2;
icsk->icsk_ack.quick = 0;
+ rmem2 = READ_ONCE(net->ipv4.sysctl_tcp_rmem[2]);
- if (sk->sk_rcvbuf < net->ipv4.sysctl_tcp_rmem[2] &&
+ if (sk->sk_rcvbuf < rmem2 &&
!(sk->sk_userlocks & SOCK_RCVBUF_LOCK) &&
!tcp_under_memory_pressure(sk) &&
sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)) {
WRITE_ONCE(sk->sk_rcvbuf,
- min(atomic_read(&sk->sk_rmem_alloc),
- net->ipv4.sysctl_tcp_rmem[2]));
+ min(atomic_read(&sk->sk_rmem_alloc), rmem2));
}
if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf)
tp->rcv_ssthresh = min(tp->window_clamp, 2U * tp->advmss);
@@ -693,7 +716,7 @@
* <prev RTT . ><current RTT .. ><next RTT .... >
*/
- if (sock_net(sk)->ipv4.sysctl_tcp_moderate_rcvbuf &&
+ if (READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_moderate_rcvbuf) &&
!(sk->sk_userlocks & SOCK_RCVBUF_LOCK)) {
int rcvmem, rcvbuf;
u64 rcvwin, grow;
@@ -714,7 +737,7 @@
do_div(rcvwin, tp->advmss);
rcvbuf = min_t(u64, rcvwin * rcvmem,
- sock_net(sk)->ipv4.sysctl_tcp_rmem[2]);
+ READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_rmem[2]));
if (rcvbuf > sk->sk_rcvbuf) {
WRITE_ONCE(sk->sk_rcvbuf, rcvbuf);
@@ -782,7 +805,7 @@
tcp_ecn_check_ce(sk, skb);
if (skb->len >= 128)
- tcp_grow_window(sk, skb);
+ tcp_grow_window(sk, skb, true);
}
/* Called to compute a smoothed rtt estimate. The data fed to this
@@ -1011,7 +1034,7 @@
tp->undo_marker ? tp->undo_retrans : 0);
#endif
tp->reordering = min_t(u32, (metric + mss - 1) / mss,
- sock_net(sk)->ipv4.sysctl_tcp_max_reordering);
+ READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_max_reordering));
}
/* This exciting event is worth to be remembered. 8) */
@@ -1990,7 +2013,7 @@
return;
tp->reordering = min_t(u32, tp->packets_out + addend,
- sock_net(sk)->ipv4.sysctl_tcp_max_reordering);
+ READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_max_reordering));
tp->reord_seen++;
NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPRENOREORDER);
}
@@ -2055,7 +2078,8 @@
static bool tcp_is_rack(const struct sock *sk)
{
- return sock_net(sk)->ipv4.sysctl_tcp_recovery & TCP_RACK_LOSS_DETECTION;
+ return READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_recovery) &
+ TCP_RACK_LOSS_DETECTION;
}
/* If we detect SACK reneging, forget all SACK information
@@ -2099,6 +2123,7 @@
struct tcp_sock *tp = tcp_sk(sk);
struct net *net = sock_net(sk);
bool new_recovery = icsk->icsk_ca_state < TCP_CA_Recovery;
+ u8 reordering;
tcp_timeout_mark_lost(sk);
@@ -2119,10 +2144,12 @@
/* Timeout in disordered state after receiving substantial DUPACKs
* suggests that the degree of reordering is over-estimated.
*/
+ reordering = READ_ONCE(net->ipv4.sysctl_tcp_reordering);
if (icsk->icsk_ca_state <= TCP_CA_Disorder &&
- tp->sacked_out >= net->ipv4.sysctl_tcp_reordering)
+ tp->sacked_out >= reordering)
tp->reordering = min_t(unsigned int, tp->reordering,
- net->ipv4.sysctl_tcp_reordering);
+ reordering);
+
tcp_set_ca_state(sk, TCP_CA_Loss);
tp->high_seq = tp->snd_nxt;
tcp_ecn_queue_cwr(tp);
@@ -2131,7 +2158,7 @@
* loss recovery is underway except recurring timeout(s) on
* the same SND.UNA (sec 3.2). Disable F-RTO on path MTU probing
*/
- tp->frto = net->ipv4.sysctl_tcp_frto &&
+ tp->frto = READ_ONCE(net->ipv4.sysctl_tcp_frto) &&
(new_recovery || icsk->icsk_retransmits) &&
!inet_csk(sk)->icsk_mtup.probe_size;
}
@@ -2148,7 +2175,8 @@
*/
static bool tcp_check_sack_reneging(struct sock *sk, int flag)
{
- if (flag & FLAG_SACK_RENEGING) {
+ if (flag & FLAG_SACK_RENEGING &&
+ flag & FLAG_SND_UNA_ADVANCED) {
struct tcp_sock *tp = tcp_sk(sk);
unsigned long delay = max(usecs_to_jiffies(tp->srtt_us >> 4),
msecs_to_jiffies(10));
@@ -2469,6 +2497,21 @@
return tp->undo_marker && (!tp->undo_retrans || tcp_packet_delayed(tp));
}
+static bool tcp_is_non_sack_preventing_reopen(struct sock *sk)
+{
+ struct tcp_sock *tp = tcp_sk(sk);
+
+ if (tp->snd_una == tp->high_seq && tcp_is_reno(tp)) {
+ /* Hold old state until something *above* high_seq
+ * is ACKed. For Reno it is MUST to prevent false
+ * fast retransmits (RFC2582). SACK TCP is safe. */
+ if (!tcp_any_retrans_done(sk))
+ tp->retrans_stamp = 0;
+ return true;
+ }
+ return false;
+}
+
/* People celebrate: "We love our President!" */
static bool tcp_try_undo_recovery(struct sock *sk)
{
@@ -2491,14 +2534,8 @@
} else if (tp->rack.reo_wnd_persist) {
tp->rack.reo_wnd_persist--;
}
- if (tp->snd_una == tp->high_seq && tcp_is_reno(tp)) {
- /* Hold old state until something *above* high_seq
- * is ACKed. For Reno it is MUST to prevent false
- * fast retransmits (RFC2582). SACK TCP is safe. */
- if (!tcp_any_retrans_done(sk))
- tp->retrans_stamp = 0;
+ if (tcp_is_non_sack_preventing_reopen(sk))
return true;
- }
tcp_set_ca_state(sk, TCP_CA_Open);
tp->is_sack_reneg = 0;
return false;
@@ -2534,6 +2571,8 @@
NET_INC_STATS(sock_net(sk),
LINUX_MIB_TCPSPURIOUSRTOS);
inet_csk(sk)->icsk_retransmits = 0;
+ if (tcp_is_non_sack_preventing_reopen(sk))
+ return true;
if (frto_undo || tcp_is_sack(tp)) {
tcp_set_ca_state(sk, TCP_CA_Open);
tp->is_sack_reneg = 0;
@@ -2667,12 +2706,15 @@
{
struct tcp_sock *tp = tcp_sk(sk);
struct inet_connection_sock *icsk = inet_csk(sk);
+ u64 val;
- /* FIXME: breaks with very large cwnd */
tp->prior_ssthresh = tcp_current_ssthresh(sk);
- tp->snd_cwnd = tp->snd_cwnd *
- tcp_mss_to_mtu(sk, tp->mss_cache) /
- icsk->icsk_mtup.probe_size;
+
+ val = (u64)tp->snd_cwnd * tcp_mss_to_mtu(sk, tp->mss_cache);
+ do_div(val, icsk->icsk_mtup.probe_size);
+ WARN_ON_ONCE((u32)val != val);
+ tp->snd_cwnd = max_t(u32, 1U, val);
+
tp->snd_cwnd_cnt = 0;
tp->snd_cwnd_stamp = tcp_jiffies32;
tp->snd_ssthresh = tcp_current_ssthresh(sk);
@@ -2997,7 +3039,7 @@
static void tcp_update_rtt_min(struct sock *sk, u32 rtt_us, const int flag)
{
- u32 wlen = sock_net(sk)->ipv4.sysctl_tcp_min_rtt_wlen * HZ;
+ u32 wlen = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_min_rtt_wlen) * HZ;
struct tcp_sock *tp = tcp_sk(sk);
if ((flag & FLAG_ACK_MAYBE_DELAYED) && rtt_us > tcp_min_rtt(tp)) {
@@ -3408,7 +3450,8 @@
* new SACK or ECE mark may first advance cwnd here and later reduce
* cwnd in tcp_fastretrans_alert() based on more states.
*/
- if (tcp_sk(sk)->reordering > sock_net(sk)->ipv4.sysctl_tcp_reordering)
+ if (tcp_sk(sk)->reordering >
+ READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_reordering))
return flag & FLAG_FORWARD_PROGRESS;
return flag & FLAG_DATA_ACKED;
@@ -3520,7 +3563,8 @@
if (*last_oow_ack_time) {
s32 elapsed = (s32)(tcp_jiffies32 - *last_oow_ack_time);
- if (0 <= elapsed && elapsed < net->ipv4.sysctl_tcp_invalid_ratelimit) {
+ if (0 <= elapsed &&
+ elapsed < READ_ONCE(net->ipv4.sysctl_tcp_invalid_ratelimit)) {
NET_INC_STATS(net, mib_idx);
return true; /* rate-limited: don't send yet! */
}
@@ -3567,11 +3611,11 @@
/* Then check host-wide RFC 5961 rate limit. */
now = jiffies / HZ;
- if (now != challenge_timestamp) {
- u32 ack_limit = net->ipv4.sysctl_tcp_challenge_ack_limit;
+ if (now != READ_ONCE(challenge_timestamp)) {
+ u32 ack_limit = READ_ONCE(net->ipv4.sysctl_tcp_challenge_ack_limit);
u32 half = (ack_limit + 1) >> 1;
- challenge_timestamp = now;
+ WRITE_ONCE(challenge_timestamp, now);
WRITE_ONCE(challenge_count, half + prandom_u32_max(ack_limit));
}
count = READ_ONCE(challenge_count);
@@ -3814,7 +3858,8 @@
tcp_process_tlp_ack(sk, ack, flag);
if (tcp_ack_is_dubious(sk, flag)) {
- if (!(flag & (FLAG_SND_UNA_ADVANCED | FLAG_NOT_DUP))) {
+ if (!(flag & (FLAG_SND_UNA_ADVANCED |
+ FLAG_NOT_DUP | FLAG_DSACKING_ACK))) {
num_dupack = 1;
/* Consider if pure acks were aggregated in tcp_add_backlog() */
if (!(flag & FLAG_DATA))
@@ -3999,7 +4044,7 @@
break;
case TCPOPT_WINDOW:
if (opsize == TCPOLEN_WINDOW && th->syn &&
- !estab && net->ipv4.sysctl_tcp_window_scaling) {
+ !estab && READ_ONCE(net->ipv4.sysctl_tcp_window_scaling)) {
__u8 snd_wscale = *(__u8 *)ptr;
opt_rx->wscale_ok = 1;
if (snd_wscale > TCP_MAX_WSCALE) {
@@ -4015,7 +4060,7 @@
case TCPOPT_TIMESTAMP:
if ((opsize == TCPOLEN_TIMESTAMP) &&
((estab && opt_rx->tstamp_ok) ||
- (!estab && net->ipv4.sysctl_tcp_timestamps))) {
+ (!estab && READ_ONCE(net->ipv4.sysctl_tcp_timestamps)))) {
opt_rx->saw_tstamp = 1;
opt_rx->rcv_tsval = get_unaligned_be32(ptr);
opt_rx->rcv_tsecr = get_unaligned_be32(ptr + 4);
@@ -4023,7 +4068,7 @@
break;
case TCPOPT_SACK_PERM:
if (opsize == TCPOLEN_SACK_PERM && th->syn &&
- !estab && net->ipv4.sysctl_tcp_sack) {
+ !estab && READ_ONCE(net->ipv4.sysctl_tcp_sack)) {
opt_rx->sack_ok = TCP_SACK_SEEN;
tcp_sack_reset(opt_rx);
}
@@ -4358,7 +4403,7 @@
{
struct tcp_sock *tp = tcp_sk(sk);
- if (tcp_is_sack(tp) && sock_net(sk)->ipv4.sysctl_tcp_dsack) {
+ if (tcp_is_sack(tp) && READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_dsack)) {
int mib_idx;
if (before(seq, tp->rcv_nxt))
@@ -4405,7 +4450,7 @@
NET_INC_STATS(sock_net(sk), LINUX_MIB_DELAYEDACKLOST);
tcp_enter_quickack_mode(sk, TCP_MAX_QUICKACKS);
- if (tcp_is_sack(tp) && sock_net(sk)->ipv4.sysctl_tcp_dsack) {
+ if (tcp_is_sack(tp) && READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_dsack)) {
u32 end_seq = TCP_SKB_CB(skb)->end_seq;
tcp_rcv_spurious_retrans(sk, skb);
@@ -4751,7 +4796,7 @@
* and trigger fast retransmit.
*/
if (tcp_is_sack(tp))
- tcp_grow_window(sk, skb);
+ tcp_grow_window(sk, skb, true);
kfree_skb_partial(skb, fragstolen);
skb = NULL;
goto add_sack;
@@ -4839,7 +4884,7 @@
* and trigger fast retransmit.
*/
if (tcp_is_sack(tp))
- tcp_grow_window(sk, skb);
+ tcp_grow_window(sk, skb, false);
skb_condense(skb);
skb_set_owner_r(skb, sk);
}
@@ -5370,7 +5415,17 @@
sk->sk_write_space(sk);
}
-static void tcp_check_space(struct sock *sk)
+/* Caller made space either from:
+ * 1) Freeing skbs in rtx queues (after tp->snd_una has advanced)
+ * 2) Sent skbs from output queue (and thus advancing tp->snd_nxt)
+ *
+ * We might be able to generate EPOLLOUT to the application if:
+ * 1) Space consumed in output/rtx queues is below sk->sk_sndbuf/2
+ * 2) notsent amount (tp->write_seq - tp->snd_nxt) became
+ * small enough that tcp_stream_memory_free() decides it
+ * is time to generate EPOLLOUT.
+ */
+void tcp_check_space(struct sock *sk)
{
/* pairs with tcp_poll() */
smp_mb();
@@ -5420,7 +5475,7 @@
}
if (!tcp_is_sack(tp) ||
- tp->compressed_ack >= sock_net(sk)->ipv4.sysctl_tcp_comp_sack_nr)
+ tp->compressed_ack >= READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_comp_sack_nr))
goto send_now;
if (tp->compressed_ack_rcv_nxt != tp->rcv_nxt) {
@@ -5441,11 +5496,12 @@
if (tp->srtt_us && tp->srtt_us < rtt)
rtt = tp->srtt_us;
- delay = min_t(unsigned long, sock_net(sk)->ipv4.sysctl_tcp_comp_sack_delay_ns,
+ delay = min_t(unsigned long,
+ READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_comp_sack_delay_ns),
rtt * (NSEC_PER_USEC >> 3)/20);
sock_hold(sk);
hrtimer_start_range_ns(&tp->compressed_ack_timer, ns_to_ktime(delay),
- sock_net(sk)->ipv4.sysctl_tcp_comp_sack_slack_ns,
+ READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_comp_sack_slack_ns),
HRTIMER_MODE_REL_PINNED_SOFT);
}
@@ -5473,7 +5529,7 @@
struct tcp_sock *tp = tcp_sk(sk);
u32 ptr = ntohs(th->urg_ptr);
- if (ptr && !sock_net(sk)->ipv4.sysctl_tcp_stdurg)
+ if (ptr && !READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_stdurg))
ptr--;
ptr += ntohl(th->seq);
@@ -5722,7 +5778,7 @@
trace_tcp_probe(sk, skb);
tcp_mstamp_refresh(tp);
- if (unlikely(!sk->sk_rx_dst))
+ if (unlikely(!rcu_access_pointer(sk->sk_rx_dst)))
inet_csk(sk)->icsk_af_ops->sk_rx_dst_set(sk, skb);
/*
* Header prediction.
@@ -6669,11 +6725,14 @@
{
struct request_sock_queue *queue = &inet_csk(sk)->icsk_accept_queue;
const char *msg = "Dropping request";
- bool want_cookie = false;
struct net *net = sock_net(sk);
+ bool want_cookie = false;
+ u8 syncookies;
+
+ syncookies = READ_ONCE(net->ipv4.sysctl_tcp_syncookies);
#ifdef CONFIG_SYN_COOKIES
- if (net->ipv4.sysctl_tcp_syncookies) {
+ if (syncookies) {
msg = "Sending cookies";
want_cookie = true;
__NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPREQQFULLDOCOOKIES);
@@ -6681,8 +6740,7 @@
#endif
__NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPREQQFULLDROP);
- if (!queue->synflood_warned &&
- net->ipv4.sysctl_tcp_syncookies != 2 &&
+ if (!queue->synflood_warned && syncookies != 2 &&
xchg(&queue->synflood_warned, 1) == 0)
net_info_ratelimited("%s: Possible SYN flooding on port %d. %s. Check SNMP counters.\n",
proto, sk->sk_num, msg);
@@ -6731,7 +6789,7 @@
struct tcp_sock *tp = tcp_sk(sk);
u16 mss;
- if (sock_net(sk)->ipv4.sysctl_tcp_syncookies != 2 &&
+ if (READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_syncookies) != 2 &&
!inet_csk_reqsk_queue_is_full(sk))
return 0;
@@ -6765,13 +6823,15 @@
bool want_cookie = false;
struct dst_entry *dst;
struct flowi fl;
+ u8 syncookies;
+
+ syncookies = READ_ONCE(net->ipv4.sysctl_tcp_syncookies);
/* TW buckets are converted to open requests without
* limitations, they conserve resources and peer is
* evidently real one.
*/
- if ((net->ipv4.sysctl_tcp_syncookies == 2 ||
- inet_csk_reqsk_queue_is_full(sk)) && !isn) {
+ if ((syncookies == 2 || inet_csk_reqsk_queue_is_full(sk)) && !isn) {
want_cookie = tcp_syn_flood_action(sk, rsk_ops->slab_name);
if (!want_cookie)
goto drop;
@@ -6825,10 +6885,12 @@
goto drop_and_free;
if (!want_cookie && !isn) {
+ int max_syn_backlog = READ_ONCE(net->ipv4.sysctl_max_syn_backlog);
+
/* Kill the following clause, if you dislike this way. */
- if (!net->ipv4.sysctl_tcp_syncookies &&
- (net->ipv4.sysctl_max_syn_backlog - inet_csk_reqsk_queue_len(sk) <
- (net->ipv4.sysctl_max_syn_backlog >> 2)) &&
+ if (!syncookies &&
+ (max_syn_backlog - inet_csk_reqsk_queue_len(sk) <
+ (max_syn_backlog >> 2)) &&
!tcp_peer_is_proven(req, dst)) {
/* Without syncookies last quarter of
* backlog is filled with destinations,