Skip to contentSkip to conversation

memcg: net: track network throttling due to memcg memory pressure

PATCH Applied Shakeel Butt 2 participants 3 messages Oct 16, 2025

Subsystems (7)
lore
Share & tools
lore

Shakeel Butt

author maintainer
#0 lore raw

PATCH v2 1/1 memcg: net: track network throttling due to memcg memory pressure

The kernel can throttle network sockets if the memory cgroup associated with the corresponding socket is under memory pressure. The throttling actions include clamping the transmit window, failing to expand receive or send buffers, aggressively prune out-of-order receive queue, FIN deferred to a retransmitted packet and more. Let's add memcg metric to indicate track such throttling actions.

At the moment memcg memory pressure is defined through vmpressure and in future it may be defined using PSI or we may add more flexible way for the users to define memory pressure, maybe through ebpf. However the potential throttling actions will remain the same, so this newly introduced metric will continue to track throttling actions irrespective of how memcg memory pressure is defined.

---
Changes since v1:
- renamed socks_throttled & MEMCG_SOCKS_THROTTLED as suggested by Roman
http://lore.kernel.org/[email protected]

 Documentation/admin-guide/cgroup-v2.rst | 4 ++++
 include/linux/memcontrol.h              | 1 +
 include/net/sock.h                      | 6 +++++-
 kernel/cgroup/cgroup.c                  | 1 +
 mm/memcontrol.c                         | 3 +++
 5 files changed, 14 insertions(+), 1 deletion(-)
Documentation/admin-guide/cgroup-v2.rst +4 -0
index 0e6c67ac585a..3345961c30ac 100644
@@ -1515,6 +1515,10 @@ The following nested keys are defined.
           oom_group_kill
                 The number of times a group OOM has occurred.
 
+          sock_throttled
+                The number of times network sockets associated with
+                this cgroup are throttled.
+
   memory.events.local
 	Similar to memory.events but the fields in the file are local
 	to the cgroup i.e. not hierarchical. The file modified event
include/linux/memcontrol.h +1 -0
index 7ed15f858dc4..e0240560cea4 100644
@@ -52,6 +52,7 @@ enum memcg_memory_event {
 	MEMCG_SWAP_HIGH,
 	MEMCG_SWAP_MAX,
 	MEMCG_SWAP_FAIL,
+	MEMCG_SOCK_THROTTLED,
 	MEMCG_NR_MEMORY_EVENTS,
 };
 
include/net/sock.h +5 -1
index 60bcb13f045c..ff7d49af1619 100644
@@ -2635,8 +2635,12 @@ static inline bool mem_cgroup_sk_under_memory_pressure(const struct sock *sk)
 #endif /* CONFIG_MEMCG_V1 */
 
 	do {
-		if (time_before64(get_jiffies_64(), mem_cgroup_get_socket_pressure(memcg)))
+		if (time_before64(get_jiffies_64(),
+				  mem_cgroup_get_socket_pressure(memcg))) {
+			memcg_memory_event(mem_cgroup_from_sk(sk),
+					   MEMCG_SOCK_THROTTLED);
 			return true;
+		}
 	} while ((memcg = parent_mem_cgroup(memcg)));
 
 	return false;
kernel/cgroup/cgroup.c +1 -0
index fdee387f0d6b..8df671c59987 100644
@@ -4704,6 +4704,7 @@ void cgroup_file_notify(struct cgroup_file *cfile)
 	}
 	spin_unlock_irqrestore(&cgroup_file_kn_lock, flags);
 }
+EXPORT_SYMBOL_GPL(cgroup_file_notify);
 
 /**
  * cgroup_file_show - show or hide a hidden cgroup file
mm/memcontrol.c +3 -0
index 3ae5cbcaed75..976412c8196e 100644
@@ -81,6 +81,7 @@ struct cgroup_subsys memory_cgrp_subsys __read_mostly;
 EXPORT_SYMBOL(memory_cgrp_subsys);
 
 struct mem_cgroup *root_mem_cgroup __read_mostly;
+EXPORT_SYMBOL(root_mem_cgroup);
 
 /* Active memory cgroup to use from an interrupt context */
 DEFINE_PER_CPU(struct mem_cgroup *, int_active_memcg);
@@ -4463,6 +4464,8 @@ static void __memory_events_show(struct seq_file *m, atomic_long_t *events)
 		   atomic_long_read(&events[MEMCG_OOM_KILL]));
 	seq_printf(m, "oom_group_kill %lu\n",
 		   atomic_long_read(&events[MEMCG_OOM_GROUP_KILL]));
+	seq_printf(m, "sock_throttled %lu\n",
+		   atomic_long_read(&events[MEMCG_SOCK_THROTTLED]));
 }
 
 static int memory_events_show(struct seq_file *m, void *v)
Signature · 1 line
2.47.3
maintainer
#1 lore raw
Quoted 21 lines · On Thu, 16 Oct 2025 09:10:35 -0700 Shakeel Butt <[email protected]> wrote:

The kernel can throttle network sockets if the memory cgroup associated with the corresponding socket is under memory pressure. The throttling actions include clamping the transmit window, failing to expand receive or send buffers, aggressively prune out-of-order receive queue, FIN deferred to a retransmitted packet and more. Let's add memcg metric to indicate track such throttling actions.

At the moment memcg memory pressure is defined through vmpressure and in future it may be defined using PSI or we may add more flexible way for the users to define memory pressure, maybe through ebpf. However the potential throttling actions will remain the same, so this newly introduced metric will continue to track throttling actions irrespective of how memcg memory pressure is defined.

...

include/net/sock.h +0 -0
@@ -2635,8 +2635,12 @@ static inline bool mem_cgroup_sk_under_memory_pressure(const struct sock *sk)
 #endif /* CONFIG_MEMCG_V1 */
 

include/net/sock.h static inline bool mem_cgroup_sk_under_memory_pressure(const struct sock *sk) · line 2645

Quoted 10 lines
include/net/sock.h +5 -1
 	do {
-		if (time_before64(get_jiffies_64(), mem_cgroup_get_socket_pressure(memcg)))
+		if (time_before64(get_jiffies_64(),
+				  mem_cgroup_get_socket_pressure(memcg))) {
+			memcg_memory_event(mem_cgroup_from_sk(sk),
+					   MEMCG_SOCK_THROTTLED);
 			return true;
+		}
 	} while ((memcg = parent_mem_cgroup(memcg)));
 

Totally OT, but that's one bigass inlined function. A quick test indicates that uninlining just this function reduces the size of tcp_input.o and tcp_output.o nicely. x86_64 defconfig:

   text	   data	    bss	    dec	    hex	filename
  52130	   1686	      0	  53816	   d238	net/ipv4/tcp_input.o
  32335	   1221	      0	  33556	   8314	net/ipv4/tcp_output.o
   text	   data	    bss	    dec	    hex	filename
  51346	   1494	      0	  52840	   ce68	net/ipv4/tcp_input.o
  31911	   1125	      0	  33036	   810c	net/ipv4/tcp_output.o
author maintainer
#2 lore raw
Quoted 38 lines · Andrew Morton
Quoted 21 lines · On Thu, 16 Oct 2025 09:10:35 -0700 Shakeel Butt <[email protected]> wrote:

The kernel can throttle network sockets if the memory cgroup associated with the corresponding socket is under memory pressure. The throttling actions include clamping the transmit window, failing to expand receive or send buffers, aggressively prune out-of-order receive queue, FIN deferred to a retransmitted packet and more. Let's add memcg metric to indicate track such throttling actions.

At the moment memcg memory pressure is defined through vmpressure and in future it may be defined using PSI or we may add more flexible way for the users to define memory pressure, maybe through ebpf. However the potential throttling actions will remain the same, so this newly introduced metric will continue to track throttling actions irrespective of how memcg memory pressure is defined.

...

include/net/sock.h +0 -0
@@ -2635,8 +2635,12 @@ static inline bool mem_cgroup_sk_under_memory_pressure(const struct sock *sk)
 #endif /* CONFIG_MEMCG_V1 */
 

include/net/sock.h static inline bool mem_cgroup_sk_under_memory_pressure(const struct sock *sk) · line 2645

Quoted 10 lines
include/net/sock.h +5 -1
 	do {
-		if (time_before64(get_jiffies_64(), mem_cgroup_get_socket_pressure(memcg)))
+		if (time_before64(get_jiffies_64(),
+				  mem_cgroup_get_socket_pressure(memcg))) {
+			memcg_memory_event(mem_cgroup_from_sk(sk),
+					   MEMCG_SOCK_THROTTLED);
 			return true;
+		}
 	} while ((memcg = parent_mem_cgroup(memcg)));
 

Totally OT, but that's one bigass inlined function. A quick test indicates that uninlining just this function reduces the size of tcp_input.o and tcp_output.o nicely. x86_64 defconfig:

Quoted 8 lines
   text	   data	    bss	    dec	    hex	filename
  52130	   1686	      0	  53816	   d238	net/ipv4/tcp_input.o
  32335	   1221	      0	  33556	   8314	net/ipv4/tcp_output.o
   text	   data	    bss	    dec	    hex	filename
  51346	   1494	      0	  52840	   ce68	net/ipv4/tcp_input.o
  31911	   1125	      0	  33036	   810c	net/ipv4/tcp_output.o

Nice find and this inlining might be hurting instead of helping. I will look into it if no one else comes to it before me.