net, neigh: introduce interval_probe_time for periodic probe
Subsystems (4)
v3 cover net, neigh: introduce interval_probe_time for periodic probe
This series adds a new option interval_probe_time in net, neigh for periodic probe, and add a limitation to prevent it set to 0
Yuwei Wang (2):
sysctl: add proc_dointvec_jiffies_minmax
net, neigh: introduce interval_probe_time for periodic probe
include/linux/sysctl.h | 2 ++ include/net/neighbour.h | 1 + include/uapi/linux/neighbour.h | 1 + include/uapi/linux/sysctl.h | 37 +++++++++++++++++----------------- kernel/sysctl.c | 36 +++++++++++++++++++++++++++++++++ net/core/neighbour.c | 33 ++++++++++++++++++++++++++++-- net/decnet/dn_neigh.c | 1 + net/ipv4/arp.c | 1 + net/ipv6/ndisc.c | 1 + 9 files changed, 93 insertions(+), 20 deletions(-)
base-commit: da6e113ff010815fdd21ee1e9af2e8d179a2680f
Signature · 1 line
2.34.1
PATCH v3 1/2 sysctl: add proc_dointvec_jiffies_minmax
add proc_dointvec_jiffies_minmax to fit jiffies param with a limited range of values
--- include/linux/sysctl.h | 2 ++ kernel/sysctl.c | 36 ++++++++++++++++++++++++++++++++++++ 2 files changed, 38 insertions(+)
include/linux/sysctl.h +2 -0
@@ -75,6 +75,8 @@ int proc_douintvec_minmax(struct ctl_table *table, int write, void *buffer,int proc_dou8vec_minmax (struct ctl_table *table,int write,void *buffer,size_t *lenp,loff_t *ppos);int proc_dointvec_jiffies (struct ctl_table *,int ,void *,size_t *,loff_t *); +int proc_dointvec_jiffies_minmax (struct ctl_table *table,int write, +void *buffer,size_t *lenp,loff_t *ppos);int proc_dointvec_userhz_jiffies (struct ctl_table *,int ,void *,size_t *,loff_t *); int proc_dointvec_ms_jiffies(struct ctl_table *, int, void *, size_t *,
kernel/sysctl.c +36 -0
@@ -1189,6 +1189,31 @@ static int do_proc_dointvec_jiffies_conv(bool *negp, unsigned long *lvalp,return 0 ; } +static int do_proc_dointvec_jiffies_minmax_conv (bool *negp,unsigned long *lvalp, +int *valp,int write,void *data) +{ +int tmp, ret; +struct do_proc_dointvec_minmax_conv_param *param = data; +/* +* If writing, first do so via a temporary local int so we can +* bounds-check it before touching *valp. +*/ +int *ip = write ? &tmp : valp; + + ret =do_proc_dointvec_jiffies_conv (negp, lvalp, ip, write, data); +if (ret) +return ret; + +if (write) { +if ((param->min && *param->min > tmp) || + (param->max && *param->max < tmp)) +return -EINVAL ; + *valp = tmp; +} + +return 0 ; +} +static int do_proc_dointvec_userhz_jiffies_conv (bool *negp,unsigned long *lvalp,int *valp,int write,void *data)
@@ -1259,6 +1284,17 @@ int proc_dointvec_jiffies(struct ctl_table *table, int write, do_proc_dointvec_jiffies_conv,NULL); } +int proc_dointvec_jiffies_minmax (struct ctl_table *table,int write, +void *buffer,size_t *lenp,loff_t *ppos) +{ +struct do_proc_dointvec_minmax_conv_param param = { + .min = (int *) table->extra1, + .max = (int *) table->extra2, + }; +return do_proc_dointvec (table, write, buffer, lenp, ppos, + do_proc_dointvec_jiffies_minmax_conv, ¶m); +} +/** * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds * @table: the sysctl table
Signature · 1 line
2.34.1
Jakub Kicinski
maintainerRe: [PATCH net-next v3 1/2] sysctl: add proc_dointvec_jiffies_minmax
kernel/sysctl.c static int do_proc_dointvec_jiffies_conv(bool *negp, unsigned long *lvalp, · line 1216
Quoted 6 lines · On Thu, 9 Jun 2022 10:57:24 +0000 Yuwei Wang wrote:
kernel/sysctl.c +6
-0+ *valp = tmp; +} + +return 0 ; +} +
Looks whitespace damaged
PATCH v3 2/2 net, neigh: introduce interval_probe_time for periodic probe
In reply to Yuwei Wangcommit ed6cd6a17896 ("net, neigh: Set lower cap for neigh_managed_work rearming") fixed a case when DELAY_PROBE_TIME is configured to 0, the processing of the system work queue hog CPU to 100%, and further more we should introduce a new option used by periodic probe
---
v3:
- add limitation to prevent INTERVAL_PROBE_TIME to 0
- remove NETEVENT_INTERVAL_PROBE_TIME_UPDATE
- add .min to NDTPA_INTERVAL_PROBE_TIME
include/net/neighbour.h | 1 + include/uapi/linux/neighbour.h | 1 + include/uapi/linux/sysctl.h | 37 +++++++++++++++++----------------- net/core/neighbour.c | 33 ++++++++++++++++++++++++++++-- net/decnet/dn_neigh.c | 1 + net/ipv4/arp.c | 1 + net/ipv6/ndisc.c | 1 + 7 files changed, 55 insertions(+), 20 deletions(-)
include/net/neighbour.h +1 -0
@@ -48,6 +48,7 @@ enum {NEIGH_VAR_RETRANS_TIME ,NEIGH_VAR_BASE_REACHABLE_TIME ,NEIGH_VAR_DELAY_PROBE_TIME , +NEIGH_VAR_INTERVAL_PROBE_TIME ,NEIGH_VAR_GC_STALETIME ,NEIGH_VAR_QUEUE_LEN_BYTES , NEIGH_VAR_PROXY_QLEN,
include/uapi/linux/neighbour.h +1 -0
@@ -154,6 +154,7 @@ enum {NDTPA_QUEUE_LENBYTES ,/* u32 */ NDTPA_MCAST_REPROBES ,/* u32 */ NDTPA_PAD , +NDTPA_INTERVAL_PROBE_TIME ,/* u64, msecs */ __NDTPA_MAX };#define NDTPA_MAX (__NDTPA_MAX - 1)
include/uapi/linux/sysctl.h +19 -18
@@ -584,24 +584,25 @@ enum {/* /proc/sys/net/<protocol>/neigh/<dev> */ enum { -NET_NEIGH_MCAST_SOLICIT =1 , -NET_NEIGH_UCAST_SOLICIT =2 , -NET_NEIGH_APP_SOLICIT =3 , -NET_NEIGH_RETRANS_TIME =4 , -NET_NEIGH_REACHABLE_TIME =5 , -NET_NEIGH_DELAY_PROBE_TIME =6 , -NET_NEIGH_GC_STALE_TIME =7 , -NET_NEIGH_UNRES_QLEN =8 , -NET_NEIGH_PROXY_QLEN =9 , -NET_NEIGH_ANYCAST_DELAY =10 , -NET_NEIGH_PROXY_DELAY =11 , -NET_NEIGH_LOCKTIME =12 , -NET_NEIGH_GC_INTERVAL =13 , -NET_NEIGH_GC_THRESH1 =14 , -NET_NEIGH_GC_THRESH2 =15 , -NET_NEIGH_GC_THRESH3 =16 , -NET_NEIGH_RETRANS_TIME_MS =17 , -NET_NEIGH_REACHABLE_TIME_MS =18 , +NET_NEIGH_MCAST_SOLICIT =1 , +NET_NEIGH_UCAST_SOLICIT =2 , +NET_NEIGH_APP_SOLICIT =3 , +NET_NEIGH_RETRANS_TIME =4 , +NET_NEIGH_REACHABLE_TIME =5 , +NET_NEIGH_DELAY_PROBE_TIME =6 , +NET_NEIGH_GC_STALE_TIME =7 , +NET_NEIGH_UNRES_QLEN =8 , +NET_NEIGH_PROXY_QLEN =9 , +NET_NEIGH_ANYCAST_DELAY =10 , +NET_NEIGH_PROXY_DELAY =11 , +NET_NEIGH_LOCKTIME =12 , +NET_NEIGH_GC_INTERVAL =13 , +NET_NEIGH_GC_THRESH1 =14 , +NET_NEIGH_GC_THRESH2 =15 , +NET_NEIGH_GC_THRESH3 =16 , +NET_NEIGH_RETRANS_TIME_MS =17 , +NET_NEIGH_REACHABLE_TIME_MS =18 , +NET_NEIGH_INTERVAL_PROBE_TIME =19 , };/* /proc/sys/net/dccp */
net/core/neighbour.c +31 -2
@@ -1579,7 +1579,7 @@ static void neigh_managed_work(struct work_struct *work)list_for_each_entry (neigh, &tbl->managed_list, managed_list)neigh_event_send_probe (neigh,NULL , false);queue_delayed_work (system_power_efficient_wq, &tbl->managed_work, -max (NEIGH_VAR (&tbl->parms,DELAY_PROBE_TIME ),HZ )); +NEIGH_VAR (&tbl->parms,INTERVAL_PROBE_TIME ));write_unlock_bh (&tbl->lock); }
@@ -2100,7 +2100,9 @@ static int neightbl_fill_parms(struct sk_buff *skb, struct neigh_parms *parms)nla_put_msecs (skb,NDTPA_PROXY_DELAY ,NEIGH_VAR (parms,PROXY_DELAY ),NDTPA_PAD ) ||nla_put_msecs (skb,NDTPA_LOCKTIME , - NEIGH_VAR(parms, LOCKTIME), NDTPA_PAD)) +NEIGH_VAR (parms,LOCKTIME ),NDTPA_PAD ) || +nla_put_msecs (skb,NDTPA_INTERVAL_PROBE_TIME , + NEIGH_VAR(parms, INTERVAL_PROBE_TIME), NDTPA_PAD)) goto nla_put_failure;return nla_nest_end (skb, nest);
@@ -2255,6 +2257,7 @@ static const struct nla_policy nl_ntbl_parm_policy[NDTPA_MAX+1] = { [NDTPA_ANYCAST_DELAY] = { .type = NLA_U64 }, [NDTPA_PROXY_DELAY] = { .type = NLA_U64 }, [NDTPA_LOCKTIME] = { .type = NLA_U64 }, + [NDTPA_INTERVAL_PROBE_TIME] = { .type = NLA_U64, .min = 1 }, }; static int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh,
@@ -2373,6 +2376,10 @@ static int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh, nla_get_msecs(tbp[i]));call_netevent_notifiers (NETEVENT_DELAY_PROBE_TIME_UPDATE , p);break ; +case NDTPA_INTERVAL_PROBE_TIME : +NEIGH_VAR_SET (p,INTERVAL_PROBE_TIME , +nla_get_msecs (tbp[i])); +break ;case NDTPA_RETRANS_TIME :NEIGH_VAR_SET (p,RETRANS_TIME ,nla_get_msecs (tbp[i]));
@@ -3562,6 +3569,24 @@ static int neigh_proc_dointvec_zero_intmax(struct ctl_table *ctl, int write,return ret; } +static int neigh_proc_dointvec_jiffies_positive (struct ctl_table *ctl,int write, +void *buffer,size_t *lenp, +loff_t *ppos) +{ +struct ctl_table tmp = *ctl; +int ret; + +int min =HZ ; +int max =INT_MAX ; + + tmp.extra1 = &min; + tmp.extra2 = &max; + + ret =proc_dointvec_jiffies_minmax (&tmp, write, buffer, lenp, ppos); +neigh_proc_update (ctl, write); +return ret; +} +int neigh_proc_dointvec (struct ctl_table *ctl,int write,void *buffer,size_t *lenp,loff_t *ppos) {
@@ -3658,6 +3683,9 @@ static int neigh_proc_base_reachable_time(struct ctl_table *ctl, int write,#define NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY (attr, name) \ NEIGH_SYSCTL_ENTRY(attr, attr, name, 0644, neigh_proc_dointvec_userhz_jiffies) +#define NEIGH_SYSCTL_JIFFIES_POSITIVE_ENTRY (attr, name) \ + NEIGH_SYSCTL_ENTRY(attr, attr, name, 0644, neigh_proc_dointvec_jiffies_positive) +#define NEIGH_SYSCTL_MS_JIFFIES_REUSED_ENTRY (attr, data_attr, name) \ NEIGH_SYSCTL_ENTRY(attr, data_attr, name, 0644, neigh_proc_dointvec_ms_jiffies)
@@ -3676,6 +3704,7 @@ static struct neigh_sysctl_table {NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY (RETRANS_TIME ,"retrans_time" ),NEIGH_SYSCTL_JIFFIES_ENTRY (BASE_REACHABLE_TIME ,"base_reachable_time" ),NEIGH_SYSCTL_JIFFIES_ENTRY (DELAY_PROBE_TIME ,"delay_first_probe_time" ), +NEIGH_SYSCTL_JIFFIES_POSITIVE_ENTRY (INTERVAL_PROBE_TIME ,"interval_probe_time" ),NEIGH_SYSCTL_JIFFIES_ENTRY (GC_STALETIME ,"gc_stale_time" ),NEIGH_SYSCTL_ZERO_INTMAX_ENTRY (QUEUE_LEN_BYTES ,"unres_qlen_bytes" ), NEIGH_SYSCTL_ZERO_INTMAX_ENTRY(PROXY_QLEN, "proxy_qlen"),
net/decnet/dn_neigh.c +1 -0
@@ -94,6 +94,7 @@ struct neigh_table dn_neigh_table = { [NEIGH_VAR_RETRANS_TIME] = 1 * HZ, [NEIGH_VAR_BASE_REACHABLE_TIME] = 30 * HZ, [NEIGH_VAR_DELAY_PROBE_TIME] = 5 * HZ, + [NEIGH_VAR_INTERVAL_PROBE_TIME] = 5 * HZ, [NEIGH_VAR_GC_STALETIME] = 60 * HZ, [NEIGH_VAR_QUEUE_LEN_BYTES] = SK_WMEM_MAX, [NEIGH_VAR_PROXY_QLEN] = 0,
net/ipv4/arp.c +1 -0
@@ -168,6 +168,7 @@ struct neigh_table arp_tbl = { [NEIGH_VAR_RETRANS_TIME] = 1 * HZ, [NEIGH_VAR_BASE_REACHABLE_TIME] = 30 * HZ, [NEIGH_VAR_DELAY_PROBE_TIME] = 5 * HZ, + [NEIGH_VAR_INTERVAL_PROBE_TIME] = 5 * HZ, [NEIGH_VAR_GC_STALETIME] = 60 * HZ, [NEIGH_VAR_QUEUE_LEN_BYTES] = SK_WMEM_MAX, [NEIGH_VAR_PROXY_QLEN] = 64,
net/ipv6/ndisc.c +1 -0
@@ -128,6 +128,7 @@ struct neigh_table nd_tbl = { [NEIGH_VAR_RETRANS_TIME] = ND_RETRANS_TIMER, [NEIGH_VAR_BASE_REACHABLE_TIME] = ND_REACHABLE_TIME, [NEIGH_VAR_DELAY_PROBE_TIME] = 5 * HZ, + [NEIGH_VAR_INTERVAL_PROBE_TIME] = 5 * HZ, [NEIGH_VAR_GC_STALETIME] = 60 * HZ, [NEIGH_VAR_QUEUE_LEN_BYTES] = SK_WMEM_MAX, [NEIGH_VAR_PROXY_QLEN] = 64,
Signature · 1 line
2.34.1
Jakub Kicinski
maintainerQuoted 6 lines · On Thu, 9 Jun 2022 10:57:25 +0000 Yuwei Wang wrote:
commit
ed6cd6a17896("net, neigh: Set lower cap forneigh_managed_workrearming") fixed a case whenDELAY_PROBE_TIMEis configured to 0, the processing of the system work queue hog CPU to 100%, and further more we should introduce a new option used by periodic probeSigned-off-by Yuwei Wang <[email protected]>
The new sysctl needs to be documented inDocumentation/networking/ip-sysctl
Quoted 11 lines · Jakub Kicinski
Quoted 6 lines · On Thu, 9 Jun 2022 10:57:25 +0000 Yuwei Wang wrote:
commit
ed6cd6a17896("net, neigh: Set lower cap forneigh_managed_workrearming") fixed a case whenDELAY_PROBE_TIMEis configured to 0, the processing of the system work queue hog CPU to 100%, and further more we should introduce a new option used by periodic probeSigned-off-by Yuwei Wang <[email protected]>The new sysctl needs to be documented in
Documentation/networking/ip-sysctl
Oops……I will fix the whitespace and add the doc in the next version
Quoted 21 lines · Yuwei Wang
commit
ed6cd6a17896("net, neigh: Set lower cap forneigh_managed_workrearming") fixed a case whenDELAY_PROBE_TIMEis configured to 0, the processing of the system work queue hog CPU to 100%, and further more we should introduce a new option used by periodic probeSigned-off-by Yuwei Wang <[email protected]>---
v3:
- add limitation to preventINTERVAL_PROBE_TIMEto 0
- removeNETEVENT_INTERVAL_PROBE_TIME_UPDATE
- add .min toNDTPA_INTERVAL_PROBE_TIMEinclude/net/neighbour.h | 1 + include/uapi/linux/neighbour.h | 1 + include/uapi/linux/sysctl.h | 37 +++++++++++++++++----------------- net/core/neighbour.c | 33 ++++++++++++++++++++++++++++-- net/decnet/dn_neigh.c | 1 + net/ipv4/arp.c | 1 + net/ipv6/ndisc.c | 1 + 7 files changed, 55 insertions(+), 20 deletions(-)
[snip]
Quoted 12 lines
include/uapi/linux/neighbour.h +1
-0@@ -154,6 +154,7 @@ enum {NDTPA_QUEUE_LENBYTES ,/* u32 */ NDTPA_MCAST_REPROBES ,/* u32 */ NDTPA_PAD , +NDTPA_INTERVAL_PROBE_TIME ,/* u64, msecs */ __NDTPA_MAX };#define NDTPA_MAX (__NDTPA_MAX - 1)
[snip]
Quoted 26 lines
/*
/proc/sys/net/dccp*/net/core/neighbour.c +4
-2@@ -1579,7 +1579,7 @@ static void neigh_managed_work(struct work_struct *work)list_for_each_entry (neigh, &tbl->managed_list, managed_list)neigh_event_send_probe (neigh,NULL , false);queue_delayed_work (system_power_efficient_wq, &tbl->managed_work, -max (NEIGH_VAR (&tbl->parms,DELAY_PROBE_TIME ),HZ )); +NEIGH_VAR (&tbl->parms,INTERVAL_PROBE_TIME ));write_unlock_bh (&tbl->lock); }@@ -2100,7 +2100,9 @@ static int neightbl_fill_parms(struct sk_buff *skb, struct neigh_parms *parms)nla_put_msecs (skb,NDTPA_PROXY_DELAY ,NEIGH_VAR (parms,PROXY_DELAY ),NDTPA_PAD ) ||nla_put_msecs (skb,NDTPA_LOCKTIME , - NEIGH_VAR(parms, LOCKTIME), NDTPA_PAD)) +NEIGH_VAR (parms,LOCKTIME ),NDTPA_PAD ) || +nla_put_msecs (skb,NDTPA_INTERVAL_PROBE_TIME , + NEIGH_VAR(parms, INTERVAL_PROBE_TIME), NDTPA_PAD)) goto nla_put_failure;return nla_nest_end (skb, nest);
net/core/neighbour.c static const struct nla_policy nl_ntbl_parm_policy[NDTPA_MAX+1] = { · line 2260
Quoted 4 lines
net/core/neighbour.c +1
-0@@ -2255,6 +2257,7 @@ static const struct nla_policy nl_ntbl_parm_policy[NDTPA_MAX+1] = { [NDTPA_ANYCAST_DELAY] = { .type = NLA_U64 }, [NDTPA_PROXY_DELAY] = { .type = NLA_U64 }, [NDTPA_LOCKTIME] = { .type = NLA_U64 }, + [NDTPA_INTERVAL_PROBE_TIME] = { .type = NLA_U64, .min = 1 },
shouldn't the min be MSEC_PER_SEC (1 sec minimum) ?
Quoted 15 lines
net/core/neighbour.c +4
-0}; static int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh,@@ -2373,6 +2376,10 @@ static int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh, nla_get_msecs(tbp[i]));call_netevent_notifiers (NETEVENT_DELAY_PROBE_TIME_UPDATE , p);break ; +case NDTPA_INTERVAL_PROBE_TIME : +NEIGH_VAR_SET (p,INTERVAL_PROBE_TIME , +nla_get_msecs (tbp[i])); +break ;case NDTPA_RETRANS_TIME :NEIGH_VAR_SET (p,RETRANS_TIME ,nla_get_msecs (tbp[i]));
net/core/neighbour.c static int neigh_proc_dointvec_zero_intmax(struct ctl_table *ctl, int write, · line 3574
Quoted 6 lines
net/core/neighbour.c +3
-0@@ -3562,6 +3569,24 @@ static int neigh_proc_dointvec_zero_intmax(struct ctl_table *ctl, int write,return ret; } +static int neigh_proc_dointvec_jiffies_positive (struct ctl_table *ctl,int write, +void *buffer,size_t *lenp, +loff_t *ppos)
Do we need the proc entry to be in jiffies when the netlink option is in ms? Why not make it directly in ms (with _ms similar to other neigh _ms time options) ?
IMO, it would be better to be consistent with the netlink option which sets it in ms.
It seems the _ms options were added later and usually people want a more understandable value, I haven't seen anyone wanting a jiffies version of a ms interval variable. :)
net/core/neighbour.c static int neigh_proc_dointvec_zero_intmax(struct ctl_table *ctl, int write, · line 3583
Quoted 9 lines
net/core/neighbour.c +9
-0+{ +struct ctl_table tmp = *ctl; +int ret; + +int min =HZ ; +int max =INT_MAX ; + + tmp.extra1 = &min; + tmp.extra2 = &max;
hmm, I don't think these min/max match the netlink attribute's min/max.
net/core/neighbour.c static int neigh_proc_base_reachable_time(struct ctl_table *ctl, int write, · line 3685
Quoted 13 lines
net/core/neighbour.c +6
-0+ + ret =proc_dointvec_jiffies_minmax (&tmp, write, buffer, lenp, ppos); +neigh_proc_update (ctl, write); +return ret; +} +int neigh_proc_dointvec (struct ctl_table *ctl,int write,void *buffer,size_t *lenp,loff_t *ppos) {@@ -3658,6 +3683,9 @@ static int neigh_proc_base_reachable_time(struct ctl_table *ctl, int write,#define NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY (attr, name) \ NEIGH_SYSCTL_ENTRY(attr, attr, name, 0644, neigh_proc_dointvec_userhz_jiffies)
[snip]
Cheers,
Nik
Quoted 7 lines · Nikolay Aleksandrov
net/core/neighbour.c static const struct nla_policy nl_ntbl_parm_policy[NDTPA_MAX+1] = { · line 2260
Quoted 5 lines
net/core/neighbour.c +1
-0@@ -2255,6 +2257,7 @@ static const struct nla_policy nl_ntbl_parm_policy[NDTPA_MAX+1] = { [NDTPA_ANYCAST_DELAY] = { .type = NLA_U64 }, [NDTPA_PROXY_DELAY] = { .type = NLA_U64 }, [NDTPA_LOCKTIME] = { .type = NLA_U64 }, + [NDTPA_INTERVAL_PROBE_TIME] = { .type = NLA_U64, .min = 1 },shouldn't the min be
MSEC_PER_SEC(1 sec minimum) ?
thanks, I will make it match the option ;)
Quoted 5 lines
net/core/neighbour.c static int neigh_proc_dointvec_zero_intmax(struct ctl_table *ctl, int write, · line 3574
Quoted 4 lines
net/core/neighbour.c +3
-0+static int neigh_proc_dointvec_jiffies_positive (struct ctl_table *ctl,int write, +void *buffer,size_t *lenp, +loff_t *ppos)
Quoted 8 lines
Do we need the proc entry to be in jiffies when the netlink option is in ms? Why not make it directly in ms (with
_mssimilar to other neigh_mstime options) ?IMO, it would be better to be consistent with the netlink option which sets it in ms.
It seems the
_msoptions were added later and usually people want a more understandable value, I haven't seen anyone wanting a jiffies version of a ms interval variable. :)
It was in jiffies because this entry was separated from DELAY_PROBE_TIME, it keeps nearly all the things the same as DELAY_PROBE_TIME, they are both configured by seconds and read to jiffies, was ms in netlink attribute,
I think it's ok to keep this consistency, and is there a demand required to configure it by ms?
If there is that demand, we can make it configured as ms.
Quoted 11 lines
net/core/neighbour.c static int neigh_proc_dointvec_zero_intmax(struct ctl_table *ctl, int write, · line 3583
Quoted 9 lines
net/core/neighbour.c +9
-0+{ +struct ctl_table tmp = *ctl; +int ret; + +int min =HZ ; +int max =INT_MAX ; + + tmp.extra1 = &min; + tmp.extra2 = &max;hmm, I don't think these min/max match the netlink attribute's min/max.
thanks, I will make it match the attribute ;)
Quoted 17 lines
Quoted 11 lines
net/core/neighbour.c +6
-0+ + ret =proc_dointvec_jiffies_minmax (&tmp, write, buffer, lenp, ppos); +neigh_proc_update (ctl, write); +return ret; +} +int neigh_proc_dointvec (struct ctl_table *ctl,int write,void *buffer,size_t *lenp,loff_t *ppos) {net/core/neighbour.c static int neigh_proc_base_reachable_time(struct ctl_table *ctl, int write, · line 3685
Quoted 3 lines
net/core/neighbour.c +0
-0@@ -3658,6 +3683,9 @@ static int neigh_proc_base_reachable_time(struct ctl_table *ctl, int write,#define NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY (attr, name) \ NEIGH_SYSCTL_ENTRY(attr, attr, name, 0644, neigh_proc_dointvec_userhz_jiffies)[snip]
Cheers,
Nik
Thanks,
Yuwei Wang
Quoted 25 lines · Yuwei Wang
Quoted 7 lines · Nikolay Aleksandrov
net/core/neighbour.c static const struct nla_policy nl_ntbl_parm_policy[NDTPA_MAX+1] = { · line 2260
Quoted 5 lines
net/core/neighbour.c +1
-0@@ -2255,6 +2257,7 @@ static const struct nla_policy nl_ntbl_parm_policy[NDTPA_MAX+1] = { [NDTPA_ANYCAST_DELAY] = { .type = NLA_U64 }, [NDTPA_PROXY_DELAY] = { .type = NLA_U64 }, [NDTPA_LOCKTIME] = { .type = NLA_U64 }, + [NDTPA_INTERVAL_PROBE_TIME] = { .type = NLA_U64, .min = 1 },shouldn't the min be
MSEC_PER_SEC(1 sec minimum) ?thanks, I will make it match the option ;)
Quoted 5 lines
net/core/neighbour.c static int neigh_proc_dointvec_zero_intmax(struct ctl_table *ctl, int write, · line 3574
Quoted 4 lines
net/core/neighbour.c +3
-0+static int neigh_proc_dointvec_jiffies_positive (struct ctl_table *ctl,int write, +void *buffer,size_t *lenp, +loff_t *ppos)Quoted 8 lines
Do we need the proc entry to be in jiffies when the netlink option is in ms? Why not make it directly in ms (with
_mssimilar to other neigh_mstime options) ?IMO, it would be better to be consistent with the netlink option which sets it in ms.
It seems the
_msoptions were added later and usually people want a more understandable value, I haven't seen anyone wanting a jiffies version of a ms interval variable. :)
Quoted 8 lines
It was in jiffies because this entry was separated from
DELAY_PROBE_TIME, it keeps nearly all the things the same asDELAY_PROBE_TIME, they are both configured by seconds and read to jiffies, wasmsin netlink attribute,
I think it's ok to keep this consistency, and is there a demand required to configure it by ms?
If there is that demand, we can make it configured as ms.
no, no demand, just out of user-friendliness :) but I get it keeping it as jiffies is also fine
Quoted 34 lines
Quoted 11 lines
net/core/neighbour.c static int neigh_proc_dointvec_zero_intmax(struct ctl_table *ctl, int write, · line 3583
Quoted 9 lines
net/core/neighbour.c +9
-0+{ +struct ctl_table tmp = *ctl; +int ret; + +int min =HZ ; +int max =INT_MAX ; + + tmp.extra1 = &min; + tmp.extra2 = &max;hmm, I don't think these min/max match the netlink attribute's min/max.
thanks, I will make it match the attribute ;)
Quoted 17 lines
Quoted 11 lines
net/core/neighbour.c +6
-0+ + ret =proc_dointvec_jiffies_minmax (&tmp, write, buffer, lenp, ppos); +neigh_proc_update (ctl, write); +return ret; +} +int neigh_proc_dointvec (struct ctl_table *ctl,int write,void *buffer,size_t *lenp,loff_t *ppos) {net/core/neighbour.c static int neigh_proc_base_reachable_time(struct ctl_table *ctl, int write, · line 3685
Quoted 3 lines
net/core/neighbour.c +0
-0@@ -3658,6 +3683,9 @@ static int neigh_proc_base_reachable_time(struct ctl_table *ctl, int write,#define NEIGH_SYSCTL_USERHZ_JIFFIES_ENTRY (attr, name) \ NEIGH_SYSCTL_ENTRY(attr, attr, name, 0644, neigh_proc_dointvec_userhz_jiffies)[snip]
Cheers,
NikThanks,
Yuwei Wang
Jakub Kicinski
maintainerQuoted 17 lines · On Wed, 15 Jun 2022 18:39:53 +0300 Nikolay Aleksandrov wrote:
Quoted 9 lines
Quoted 8 lines
Do we need the proc entry to be in jiffies when the netlink option is in ms? Why not make it directly in ms (with
_mssimilar to other neigh_mstime options) ?IMO, it would be better to be consistent with the netlink option which sets it in ms.
It seems the
_msoptions were added later and usually people want a more understandable value, I haven't seen anyone wanting a jiffies version of a ms interval variable. :)Quoted 7 lines
It was in jiffies because this entry was separated from
DELAY_PROBE_TIME, it keeps nearly all the things the same asDELAY_PROBE_TIME, they are both configured by seconds and read to jiffies, wasmsin netlink attribute,
I think it's ok to keep this consistency, and is there a demand required to configure it by ms?
If there is that demand, we can make it configured as ms.
Quoted 2 lines
no, no demand, just out of user-friendliness :) but I get it keeping it as jiffies is also fine
+1 to using ms
Quoted 19 lines · Jakub Kicinski
Quoted 17 lines · On Wed, 15 Jun 2022 18:39:53 +0300 Nikolay Aleksandrov wrote:
Quoted 16 lines
Quoted 8 lines
Do we need the proc entry to be in jiffies when the netlink option is in ms? Why not make it directly in ms (with
_mssimilar to other neigh_mstime options) ?IMO, it would be better to be consistent with the netlink option which sets it in ms.
It seems the
_msoptions were added later and usually people want a more understandable value, I haven't seen anyone wanting a jiffies version of a ms interval variable. :)It was in jiffies because this entry was separated from
DELAY_PROBE_TIME, it keeps nearly all the things the same asDELAY_PROBE_TIME, they are both configured by seconds and read to jiffies, wasmsin netlink attribute,
I think it's ok to keep this consistency, and is there a demand required to configure it by ms?
If there is that demand, we can make it configured as ms.
Quoted 4 lines
Quoted 2 lines
no, no demand, just out of user-friendliness :) but I get it keeping it as jiffies is also fine
+1 to using ms
If no one is against this, we can make configured as ms?
so, there will be updates as follow in the next version patch:
- make this option configured as ms, and rename it to interval_probe_time
- add documentation to Documentation/networking/ip-sysctl
- fix damaged whitespace
- fix missing proc_*_jiffies_minmax on CONFIG_PROC_SYSCTL is not defined
Quoted 2 lines · Yuwei Wang
- make this option configured as ms, and rename it to
interval_probe_time
typo: make this option configured as ms, and rename it to interval_probe_time_ms
Quoted 3 lines
- add documentation to
Documentation/networking/ip-sysctl
- fix damaged whitespace
- fix missingproc_*_jiffies_minmaxonCONFIG_PROC_SYSCTLis not defined