| Message ID | a35a5881c3280bd7a4fd1943a8b40b890e3bf280.1689316697.git.jamie.bainbridge@gmail.com (mailing list archive) |
|---|---|
| State | Rejected |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | tcp: Add memory pressure flag to sockstat | expand |
Jamie Bainbridge <jamie.bainbridge@gmail.com> wrote: >When tuning a system it can be helpful to know whether the protocol is >in memory pressure state or not. This can be determined by corresponding >the number of pages in "net.ipv4.tcp_mem" with the current allocation, >but a global variable already tracks this as the source of truth. > >Expose that variable in sockstat where other protocol memory usage is >already reported. > >Add "pressure" which is 0 in normal state and 1 under pressure: > > # grep TCP /proc/net/sockstat > TCP: inuse 5 orphan 0 tw 0 alloc 7 mem 1 pressure 0 > > # grep TCP /proc/net/sockstat > TCP: inuse 5 orphan 0 tw 0 alloc 7 mem 1 pressure 1 Isn't this already available in /proc/net/protocols? protocol size sockets memory press maxhdr slab module cl co di ac io in de sh ss gs se re sp bi br ha uh gp em [...] UDP 1472 7 6 NI 0 yes kernel y y y n y y y n y y y y y n n y y y n TCP 2512 5 1 no 320 yes kernel y y y y y y y y y y y y y n y y y y y -J >Tested by writing a large value to global variable tcp_memory_pressure >(it usually stores jiffies when memory pressure was entered) and not >just by code review or editing example output. > >Signed-off-by: Jamie Bainbridge <jamie.bainbridge@gmail.com> >--- > net/ipv4/proc.c | 7 ++++--- > 1 file changed, 4 insertions(+), 3 deletions(-) > >diff --git a/net/ipv4/proc.c b/net/ipv4/proc.c >index eaf1d3113b62f7dc93fdc7b7c4041140ac63bf69..f4c5ced2de49d5c6d7f5d7ccdaa76c89dcf8c932 100644 >--- a/net/ipv4/proc.c >+++ b/net/ipv4/proc.c >@@ -51,16 +51,17 @@ > static int sockstat_seq_show(struct seq_file *seq, void *v) > { > struct net *net = seq->private; >- int orphans, sockets; >+ int orphans, sockets, tcp_pressure; > > orphans = tcp_orphan_count_sum(); > sockets = proto_sockets_allocated_sum_positive(&tcp_prot); >+ tcp_pressure = READ_ONCE(tcp_memory_pressure) ? 1 : 0; > > socket_seq_show(seq); >- seq_printf(seq, "TCP: inuse %d orphan %d tw %d alloc %d mem %ld\n", >+ seq_printf(seq, "TCP: inuse %d orphan %d tw %d alloc %d mem %ld pressure %d\n", > sock_prot_inuse_get(net, &tcp_prot), orphans, > refcount_read(&net->ipv4.tcp_death_row.tw_refcount) - 1, >- sockets, proto_memory_allocated(&tcp_prot)); >+ sockets, proto_memory_allocated(&tcp_prot), tcp_pressure); > seq_printf(seq, "UDP: inuse %d mem %ld\n", > sock_prot_inuse_get(net, &udp_prot), > proto_memory_allocated(&udp_prot)); >-- >2.41.0 > > --- -Jay Vosburgh, jay.vosburgh@canonical.com
On Sat, 15 Jul 2023 at 08:45, Jay Vosburgh <jay.vosburgh@canonical.com> wrote: > > Jamie Bainbridge <jamie.bainbridge@gmail.com> wrote: > > >When tuning a system it can be helpful to know whether the protocol is > >in memory pressure state or not. This can be determined by corresponding > >the number of pages in "net.ipv4.tcp_mem" with the current allocation, > >but a global variable already tracks this as the source of truth. > > > >Expose that variable in sockstat where other protocol memory usage is > >already reported. > > > >Add "pressure" which is 0 in normal state and 1 under pressure: > > > > # grep TCP /proc/net/sockstat > > TCP: inuse 5 orphan 0 tw 0 alloc 7 mem 1 pressure 0 > > > > # grep TCP /proc/net/sockstat > > TCP: inuse 5 orphan 0 tw 0 alloc 7 mem 1 pressure 1 > > Isn't this already available in /proc/net/protocols? > > protocol size sockets memory press maxhdr slab module cl co di ac io in de sh ss gs se re sp bi br ha uh gp em > [...] > UDP 1472 7 6 NI 0 yes kernel y y y n y y y n y y y y y n n y y y n > TCP 2512 5 1 no 320 yes kernel y y y y y y y y y y y y y n y y y y y I was not aware of this, I was only looking for symbol usage of TCP and not the generic. This is perfect, thank you very much! Jamie > >Tested by writing a large value to global variable tcp_memory_pressure > >(it usually stores jiffies when memory pressure was entered) and not > >just by code review or editing example output. > > > >Signed-off-by: Jamie Bainbridge <jamie.bainbridge@gmail.com> > >--- > > net/ipv4/proc.c | 7 ++++--- > > 1 file changed, 4 insertions(+), 3 deletions(-) > > > >diff --git a/net/ipv4/proc.c b/net/ipv4/proc.c > >index eaf1d3113b62f7dc93fdc7b7c4041140ac63bf69..f4c5ced2de49d5c6d7f5d7ccdaa76c89dcf8c932 100644 > >--- a/net/ipv4/proc.c > >+++ b/net/ipv4/proc.c > >@@ -51,16 +51,17 @@ > > static int sockstat_seq_show(struct seq_file *seq, void *v) > > { > > struct net *net = seq->private; > >- int orphans, sockets; > >+ int orphans, sockets, tcp_pressure; > > > > orphans = tcp_orphan_count_sum(); > > sockets = proto_sockets_allocated_sum_positive(&tcp_prot); > >+ tcp_pressure = READ_ONCE(tcp_memory_pressure) ? 1 : 0; > > > > socket_seq_show(seq); > >- seq_printf(seq, "TCP: inuse %d orphan %d tw %d alloc %d mem %ld\n", > >+ seq_printf(seq, "TCP: inuse %d orphan %d tw %d alloc %d mem %ld pressure %d\n", > > sock_prot_inuse_get(net, &tcp_prot), orphans, > > refcount_read(&net->ipv4.tcp_death_row.tw_refcount) - 1, > >- sockets, proto_memory_allocated(&tcp_prot)); > >+ sockets, proto_memory_allocated(&tcp_prot), tcp_pressure); > > seq_printf(seq, "UDP: inuse %d mem %ld\n", > > sock_prot_inuse_get(net, &udp_prot), > > proto_memory_allocated(&udp_prot)); > >-- > >2.41.0 > > > > > > --- > -Jay Vosburgh, jay.vosburgh@canonical.com
diff --git a/net/ipv4/proc.c b/net/ipv4/proc.c index eaf1d3113b62f7dc93fdc7b7c4041140ac63bf69..f4c5ced2de49d5c6d7f5d7ccdaa76c89dcf8c932 100644 --- a/net/ipv4/proc.c +++ b/net/ipv4/proc.c @@ -51,16 +51,17 @@ static int sockstat_seq_show(struct seq_file *seq, void *v) { struct net *net = seq->private; - int orphans, sockets; + int orphans, sockets, tcp_pressure; orphans = tcp_orphan_count_sum(); sockets = proto_sockets_allocated_sum_positive(&tcp_prot); + tcp_pressure = READ_ONCE(tcp_memory_pressure) ? 1 : 0; socket_seq_show(seq); - seq_printf(seq, "TCP: inuse %d orphan %d tw %d alloc %d mem %ld\n", + seq_printf(seq, "TCP: inuse %d orphan %d tw %d alloc %d mem %ld pressure %d\n", sock_prot_inuse_get(net, &tcp_prot), orphans, refcount_read(&net->ipv4.tcp_death_row.tw_refcount) - 1, - sockets, proto_memory_allocated(&tcp_prot)); + sockets, proto_memory_allocated(&tcp_prot), tcp_pressure); seq_printf(seq, "UDP: inuse %d mem %ld\n", sock_prot_inuse_get(net, &udp_prot), proto_memory_allocated(&udp_prot));
When tuning a system it can be helpful to know whether the protocol is in memory pressure state or not. This can be determined by corresponding the number of pages in "net.ipv4.tcp_mem" with the current allocation, but a global variable already tracks this as the source of truth. Expose that variable in sockstat where other protocol memory usage is already reported. Add "pressure" which is 0 in normal state and 1 under pressure: # grep TCP /proc/net/sockstat TCP: inuse 5 orphan 0 tw 0 alloc 7 mem 1 pressure 0 # grep TCP /proc/net/sockstat TCP: inuse 5 orphan 0 tw 0 alloc 7 mem 1 pressure 1 Tested by writing a large value to global variable tcp_memory_pressure (it usually stores jiffies when memory pressure was entered) and not just by code review or editing example output. Signed-off-by: Jamie Bainbridge <jamie.bainbridge@gmail.com> --- net/ipv4/proc.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-)