Skip to contentSkip to conversation

tcp: select sane initial rcvq_space.space for big MSS

PATCH Applied Eric Dumazet 5 participants 5 messages Dec 8, 2020 – Dec 10, 2020

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

PATCH 1/1 tcp: select sane initial rcvq_space.space for big MSS

From: Eric Dumazet <[email protected]>

Before commit a337531b942b ("tcp: up initial rmem to 128KB and SYN rwin to around 64KB") small tcp_rmem[1] values were overridden by tcp_fixup_rcvbuf() to accommodate various MSS.

This is no longer the case, and Hazem Mohamed Abuelfotoh reported
that DRS would not work for MTU 9000 endpoints receiving regular (1500 bytes) frames.

Root cause is that tcp_init_buffer_space() uses tp->rcv_wnd for upper limit of rcvq_space.space computation, while it can select later a smaller value for tp->rcv_ssthresh and tp->window_clamp.

ss -temoi on receiver would show :

skmem:(r0,rb131072,t0,tb46080,f0,w0,o0,bl0,d0) rcv_space:62496 rcv_ssthresh:56596

This means that TCP can not increase its window in tcp_grow_window(), and that DRS can never kick.

Fix this by making sure that rcvq_space.space is not bigger than number of bytes that can be held in TCP receive queue.

People unable/unwilling to change their kernel can work around this issue by selecting a bigger tcp_rmem[1] value as in :

echo "4096 196608 6291456" >/proc/sys/net/ipv4/tcp_rmem

Based on an initial report and patch from Hazem Mohamed Abuelfotoh
https://lore.kernel.org/netdev/[email protected]/

Fixes a337531b942b ("tcp: up initial rmem to 128KB and SYN rwin to around 64KB")
Fixes 041a14d26715 ("tcp: start receiver buffer autotuning sooner")
---
 net/ipv4/tcp_input.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
net/ipv4/tcp_input.c +2 -1
index 389d1b34024854a9bdcbe861d4820d1bfb495e24..ef4bdb038a4bbbd949868a01dc855bba0e90b9ca 100644
@@ -510,7 +510,6 @@ static void tcp_init_buffer_space(struct sock *sk)
 	if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK))
 		tcp_sndbuf_expand(sk);
 
-	tp->rcvq_space.space = min_t(u32, tp->rcv_wnd, TCP_INIT_CWND * tp->advmss);
 	tcp_mstamp_refresh(tp);
 	tp->rcvq_space.time = tp->tcp_mstamp;
 	tp->rcvq_space.seq = tp->copied_seq;
@@ -534,6 +533,8 @@ static void tcp_init_buffer_space(struct sock *sk)
 
 	tp->rcv_ssthresh = min(tp->rcv_ssthresh, tp->window_clamp);
 	tp->snd_cwnd_stamp = tcp_jiffies32;
+	tp->rcvq_space.space = min3(tp->rcv_ssthresh, tp->rcv_wnd,
+				    (u32)TCP_INIT_CWND * tp->advmss);
 }
 
 /* 4. Recalculate window clamp after socket hit its memory bounds. */
Signature · 1 line
2.29.2.576.ga3fc446d84-goog
Acked-by on v1 1/1
#1 lore raw
Quoted 28 lines · Eric Dumazet

From: Eric Dumazet <[email protected]>

Before commit a337531b942b ("tcp: up initial rmem to 128KB and SYN rwin to around 64KB") small tcp_rmem[1] values were overridden by tcp_fixup_rcvbuf() to accommodate various MSS.

This is no longer the case, and Hazem Mohamed Abuelfotoh reported
that DRS would not work for MTU 9000 endpoints receiving regular (1500 bytes) frames.

Root cause is that tcp_init_buffer_space() uses tp->rcv_wnd for upper limit of rcvq_space.space computation, while it can select later a smaller value for tp->rcv_ssthresh and tp->window_clamp.

ss -temoi on receiver would show :

skmem:(r0,rb131072,t0,tb46080,f0,w0,o0,bl0,d0) rcv_space:62496 rcv_ssthresh:56596

This means that TCP can not increase its window in tcp_grow_window(), and that DRS can never kick.

Fix this by making sure that rcvq_space.space is not bigger than number of bytes that can be held in TCP receive queue.

People unable/unwilling to change their kernel can work around this issue by selecting a bigger tcp_rmem[1] value as in :

echo "4096 196608 6291456" >/proc/sys/net/ipv4/tcp_rmem

Quoted 7 lines

Based on an initial report and patch from Hazem Mohamed Abuelfotoh
https://lore.kernel.org/netdev/[email protected]/

Fixes a337531b942b ("tcp: up initial rmem to 128KB and SYN rwin to around 64KB")
Fixes 041a14d26715 ("tcp: start receiver buffer autotuning sooner")

Nice catch and fix! Thanks Eric!

Quoted 28 lines
---
 net/ipv4/tcp_input.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
net/ipv4/tcp_input.c +2 -2
index 389d1b34024854a9bdcbe861d4820d1bfb495e24..ef4bdb038a4bbbd949868a01dc855bba0e90b9ca 100644
@@ -510,7 +510,6 @@ static void tcp_init_buffer_space(struct sock *sk)
        if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK))
                tcp_sndbuf_expand(sk);
 
-       tp->rcvq_space.space = min_t(u32, tp->rcv_wnd, TCP_INIT_CWND * tp->advmss);
        tcp_mstamp_refresh(tp);
        tp->rcvq_space.time = tp->tcp_mstamp;
        tp->rcvq_space.seq = tp->copied_seq;
@@ -534,6 +533,8 @@ static void tcp_init_buffer_space(struct sock *sk)
 
        tp->rcv_ssthresh = min(tp->rcv_ssthresh, tp->window_clamp);
        tp->snd_cwnd_stamp = tcp_jiffies32;
+       tp->rcvq_space.space = min3(tp->rcv_ssthresh, tp->rcv_wnd,
+                                   (u32)TCP_INIT_CWND * tp->advmss);
 }
 
 /* 4. Recalculate window clamp after socket hit its memory bounds. */
--

2.29.2.576.ga3fc446d84-goog

maintainer Applied
#2 lore raw
In reply to Eric Dumazet

From: Eric Dumazet <[email protected]>
Date: Tue, 8 Dec 2020 08:21:31 -0800

Quoted 27 lines

From: Eric Dumazet <[email protected]>

Before commit a337531b942b ("tcp: up initial rmem to 128KB and SYN rwin to around 64KB") small tcp_rmem[1] values were overridden by tcp_fixup_rcvbuf() to accommodate various MSS.

This is no longer the case, and Hazem Mohamed Abuelfotoh reported
that DRS would not work for MTU 9000 endpoints receiving regular (1500 bytes) frames.

Root cause is that tcp_init_buffer_space() uses tp->rcv_wnd for upper limit of rcvq_space.space computation, while it can select later a smaller value for tp->rcv_ssthresh and tp->window_clamp.

ss -temoi on receiver would show :

skmem:(r0,rb131072,t0,tb46080,f0,w0,o0,bl0,d0) rcv_space:62496 rcv_ssthresh:56596

This means that TCP can not increase its window in tcp_grow_window(), and that DRS can never kick.

Fix this by making sure that rcvq_space.space is not bigger than number of bytes that can be held in TCP receive queue.

People unable/unwilling to change their kernel can work around this issue by selecting a bigger tcp_rmem[1] value as in :

echo "4096 196608 6291456" >/proc/sys/net/ipv4/tcp_rmem

Quoted 7 lines

Based on an initial report and patch from Hazem Mohamed Abuelfotoh
https://lore.kernel.org/netdev/[email protected]/

Fixes a337531b942b ("tcp: up initial rmem to 128KB and SYN rwin to around 64KB")
Fixes 041a14d26715 ("tcp: start receiver buffer autotuning sooner")

Applied, thanks Eric.

#3 lore raw

Hi Eric,

I don't see the patch in the stable queue. Can we add it to stable so we can cherry pick it in Amazon Linux kernel?

Thank you.

Hazem

On 09/12/2020, 00:29, "David Miller" <[email protected]> wrote:

CAUTION: This email originated from outside of the organization. Do not click links or open attachments unless you can confirm the sender and know the content is safe.

From: Eric Dumazet <[email protected]>
Date: Tue, 8 Dec 2020 08:21:31 -0800

> From: Eric Dumazet <[email protected]>
>
> Before commit a337531b942b ("tcp: up initial rmem to 128KB and SYN rwin to around 64KB") > small tcp_rmem[1] values were overridden by tcp_fixup_rcvbuf() to accommodate various MSS. >
> This is no longer the case, and Hazem Mohamed Abuelfotoh reported
> that DRS would not work for MTU 9000 endpoints receiving regular (1500 bytes) frames. >
> Root cause is that tcp_init_buffer_space() uses tp->rcv_wnd for upper limit > of rcvq_space.space computation, while it can select later a smaller
> value for tp->rcv_ssthresh and tp->window_clamp.
>
> ss -temoi on receiver would show :
>
> skmem:(r0,rb131072,t0,tb46080,f0,w0,o0,bl0,d0) rcv_space:62496 rcv_ssthresh:56596 >
> This means that TCP can not increase its window in tcp_grow_window(),
> and that DRS can never kick.
>
> Fix this by making sure that rcvq_space.space is not bigger than number of bytes > that can be held in TCP receive queue.
>
> People unable/unwilling to change their kernel can work around this issue by > selecting a bigger tcp_rmem[1] value as in :
>
> echo "4096 196608 6291456" >/proc/sys/net/ipv4/tcp_rmem
>
> Based on an initial report and patch from Hazem Mohamed Abuelfotoh
> https://lore.kernel.org/netdev/[email protected]/
>
> Fixes: a337531b942b ("tcp: up initial rmem to 128KB and SYN rwin to around 64KB") > Fixes: 041a14d26715 ("tcp: start receiver buffer autotuning sooner")
> Reported-by: Hazem Mohamed Abuelfotoh <[email protected]>
> Signed-off-by: Eric Dumazet <[email protected]>

Applied, thanks Eric.

Amazon Web Services EMEA SARL, 38 avenue John F. Kennedy, L-1855 Luxembourg, R.C.S. Luxembourg B186284

Amazon Web Services EMEA SARL, Irish Branch, One Burlington Plaza, Burlington Road, Dublin 4, Ireland, branch registration number 908705

#4 lore raw
Quoted 5 lines · Hazem

Hi Eric,

I don't see the patch in the stable queue. Can we add it to stable so we can cherry pick it in Amazon Linux kernel?

No need for stable tags , as documented in
https://elixir.bootlin.com/linux/v5.9/source/Documentation/networking/netdev-FAQ.rst#L147

Quoted 58 lines

Thank you.

Hazem

On 09/12/2020, 00:29, "David Miller" <[email protected]> wrote:

CAUTION: This email originated from outside of the organization. Do not click links or open attachments unless you can confirm the sender and know the content is safe.

From: Eric Dumazet <[email protected]>
Date: Tue, 8 Dec 2020 08:21:31 -0800

> From: Eric Dumazet <[email protected]>
>
> Before commit a337531b942b ("tcp: up initial rmem to 128KB and SYN rwin to around 64KB") > small tcp_rmem[1] values were overridden by tcp_fixup_rcvbuf() to accommodate various MSS. >
> This is no longer the case, and Hazem Mohamed Abuelfotoh reported
> that DRS would not work for MTU 9000 endpoints receiving regular (1500 bytes) frames. >
> Root cause is that tcp_init_buffer_space() uses tp->rcv_wnd for upper limit > of rcvq_space.space computation, while it can select later a smaller
> value for tp->rcv_ssthresh and tp->window_clamp.
>
> ss -temoi on receiver would show :
>
> skmem:(r0,rb131072,t0,tb46080,f0,w0,o0,bl0,d0) rcv_space:62496 rcv_ssthresh:56596 >
> This means that TCP can not increase its window in tcp_grow_window(),
> and that DRS can never kick.
>
> Fix this by making sure that rcvq_space.space is not bigger than number of bytes > that can be held in TCP receive queue.
>
> People unable/unwilling to change their kernel can work around this issue by > selecting a bigger tcp_rmem[1] value as in :
>
> echo "4096 196608 6291456" >/proc/sys/net/ipv4/tcp_rmem
>
> Based on an initial report and patch from Hazem Mohamed Abuelfotoh
> https://lore.kernel.org/netdev/[email protected]/
>
> Fixes: a337531b942b ("tcp: up initial rmem to 128KB and SYN rwin to around 64KB") > Fixes: 041a14d26715 ("tcp: start receiver buffer autotuning sooner")
> Reported-by: Hazem Mohamed Abuelfotoh <[email protected]>
> Signed-off-by: Eric Dumazet <[email protected]>

Applied, thanks Eric.

Amazon Web Services EMEA SARL, 38 avenue John F. Kennedy, L-1855 Luxembourg, R.C.S. Luxembourg B186284

Amazon Web Services EMEA SARL, Irish Branch, One Burlington Plaza, Burlington Road, Dublin 4, Ireland, branch registration number 908705