Currently taskstats_user_cmd()/taskstats_exit() do:
1) allocate stats
2) fill stats
3) make a temporary copy on stack (236 bytes)
4) copy that copy to skb
5) free stats
With the help of nla_reserve() we can operate on skb->data directly,
thus avoiding all these steps except 2).
So, before this patch:
// copy *stats to skb->data
int mk_reply(skb, ..., struct taskstats *stats);
fill_pid(stats);
mk_reply(skb, ..., stats);
After:
// return a pointer to skb->data
struct taskstats *mk_reply(skb, ...);
stat = mk_reply(skb, ...);
fill_pid(stats);
Shrinks taskatsks.o by 162 bytes.
A stupid benchmark (send one million TASKSTATS_CMD_ATTR_PID) shows the
difference,
real user sys
before:
4.02 0.06 3.96
4.02 0.04 3.98
4.02 0.04 3.97
after:
3.86 0.08 3.78
3.88 0.10 3.77
3.89 0.09 3.80
but this looks suspiciously good.
Signed-off-by: Oleg Nesterov <[email protected]>
taskstats.c | 86 ++++++++++++++++++++++++++++++------------------------------
1 files changed, 44 insertions(+), 42 deletions(-)
--- STATS/kernel/taskstats.c~6_nla 2006-11-01 14:00:03.000000000 +0300
+++ STATS/kernel/taskstats.c 2006-11-01 21:14:39.000000000 +0300
@@ -190,6 +190,7 @@ static int fill_pid(pid_t pid, struct ta
} else
get_task_struct(tsk);
+ memset(stats, 0, sizeof(*stats));
/*
* Each accounting subsystem adds calls to its functions to
* fill in relevant parts of struct taskstsats as follows
@@ -232,6 +233,8 @@ static int fill_tgid(pid_t tgid, struct
if (first->signal->stats)
memcpy(stats, first->signal->stats, sizeof(*stats));
+ else
+ memset(stats, 0, sizeof(*stats));
tsk = first;
do {
@@ -348,9 +351,9 @@ static int parse(struct nlattr *na, cpum
return ret;
}
-static int mk_reply(struct sk_buff *skb, int type, u32 pid, struct taskstats *stats)
+static struct taskstats *mk_reply(struct sk_buff *skb, int type, u32 pid)
{
- struct nlattr *na;
+ struct nlattr *na, *ret;
int aggr;
aggr = TASKSTATS_TYPE_AGGR_TGID;
@@ -358,20 +361,23 @@ static int mk_reply(struct sk_buff *skb,
aggr = TASKSTATS_TYPE_AGGR_PID;
na = nla_nest_start(skb, aggr);
- NLA_PUT_U32(skb, type, pid);
- NLA_PUT_TYPE(skb, struct taskstats, TASKSTATS_TYPE_STATS, *stats);
+ if (nla_put(skb, type, sizeof(pid), &pid) < 0)
+ goto err;
+ ret = nla_reserve(skb, TASKSTATS_TYPE_STATS, sizeof(struct taskstats));
+ if (!ret)
+ goto err;
nla_nest_end(skb, na);
- return 0;
-nla_put_failure:
- return -1;
+ return nla_data(ret);
+err:
+ return NULL;
}
static int taskstats_user_cmd(struct sk_buff *skb, struct genl_info *info)
{
int rc = 0;
struct sk_buff *rep_skb;
- struct taskstats stats;
+ struct taskstats *stats;
void *reply;
size_t size;
cpumask_t mask;
@@ -394,36 +400,36 @@ static int taskstats_user_cmd(struct sk_
size = nla_total_size(sizeof(u32)) +
nla_total_size(sizeof(struct taskstats)) + nla_total_size(0);
- memset(&stats, 0, sizeof(stats));
rc = prepare_reply(info, TASKSTATS_CMD_NEW, &rep_skb, &reply, size);
if (rc < 0)
return rc;
+ rc = -EINVAL;
if (info->attrs[TASKSTATS_CMD_ATTR_PID]) {
u32 pid = nla_get_u32(info->attrs[TASKSTATS_CMD_ATTR_PID]);
- rc = fill_pid(pid, NULL, &stats);
- if (rc < 0)
- goto err;
+ stats = mk_reply(rep_skb, TASKSTATS_TYPE_PID, pid);
+ if (!stats)
+ goto nla_err;
- if (mk_reply(rep_skb, TASKSTATS_TYPE_PID, pid, &stats))
- goto nla_put_failure;
+ rc = fill_pid(pid, NULL, stats);
+ if (rc < 0)
+ goto nla_err;
} else if (info->attrs[TASKSTATS_CMD_ATTR_TGID]) {
u32 tgid = nla_get_u32(info->attrs[TASKSTATS_CMD_ATTR_TGID]);
- rc = fill_tgid(tgid, NULL, &stats);
- if (rc < 0)
- goto err;
+ stats = mk_reply(rep_skb, TASKSTATS_TYPE_TGID, tgid);
+ if (!stats)
+ goto nla_err;
- if (mk_reply(rep_skb, TASKSTATS_TYPE_TGID, tgid, &stats))
- goto nla_put_failure;
- } else {
- rc = -EINVAL;
+ rc = fill_tgid(tgid, NULL, stats);
+ if (rc < 0)
+ goto nla_err;
+ } else
goto err;
- }
return send_reply(rep_skb, info->snd_pid);
-nla_put_failure:
- rc = genlmsg_cancel(rep_skb, reply);
+nla_err:
+ genlmsg_cancel(rep_skb, reply);
err:
nlmsg_free(rep_skb);
return rc;
@@ -458,7 +464,7 @@ void taskstats_exit(struct task_struct *
{
int rc;
struct listener_list *listeners;
- struct taskstats *tidstats;
+ struct taskstats *stats;
struct sk_buff *rep_skb;
void *reply;
size_t size;
@@ -485,20 +491,17 @@ void taskstats_exit(struct task_struct *
if (list_empty(&listeners->list))
return;
- tidstats = kmem_cache_zalloc(taskstats_cache, SLAB_KERNEL);
- if (!tidstats)
- return;
-
rc = prepare_reply(NULL, TASKSTATS_CMD_NEW, &rep_skb, &reply, size);
if (rc < 0)
- goto free_stats;
+ return;
- rc = fill_pid(tsk->pid, tsk, tidstats);
- if (rc < 0)
- goto err_skb;
+ stats = mk_reply(rep_skb, TASKSTATS_TYPE_PID, tsk->pid);
+ if (!stats)
+ goto nla_err;
- if (mk_reply(rep_skb, TASKSTATS_TYPE_PID, tsk->pid, tidstats))
- goto nla_put_failure;
+ rc = fill_pid(tsk->pid, tsk, stats);
+ if (rc < 0)
+ goto nla_err;
/*
* Doesn't matter if tsk is the leader or the last group member leaving
@@ -506,20 +509,19 @@ void taskstats_exit(struct task_struct *
if (!is_thread_group || !group_dead)
goto send;
- if (mk_reply(rep_skb, TASKSTATS_TYPE_TGID, tsk->tgid, tsk->signal->stats))
- goto nla_put_failure;
+ stats = mk_reply(rep_skb, TASKSTATS_TYPE_TGID, tsk->tgid);
+ if (!stats)
+ goto nla_err;
+
+ memcpy(stats, tsk->signal->stats, sizeof(*stats));
send:
send_cpu_listeners(rep_skb, listeners);
-free_stats:
- kmem_cache_free(taskstats_cache, tidstats);
return;
-nla_put_failure:
+nla_err:
genlmsg_cancel(rep_skb, reply);
-err_skb:
nlmsg_free(rep_skb);
- goto free_stats;
}
static struct genl_ops taskstats_ops = {
Oleg Nesterov wrote:
> Currently taskstats_user_cmd()/taskstats_exit() do:
>
> 1) allocate stats
> 2) fill stats
> 3) make a temporary copy on stack (236 bytes)
> 4) copy that copy to skb
> 5) free stats
>
> With the help of nla_reserve() we can operate on skb->data directly,
> thus avoiding all these steps except 2).
Good point.
>
> So, before this patch:
>
> // copy *stats to skb->data
> int mk_reply(skb, ..., struct taskstats *stats);
>
> fill_pid(stats);
> mk_reply(skb, ..., stats);
>
> After:
> // return a pointer to skb->data
> struct taskstats *mk_reply(skb, ...);
>
> stat = mk_reply(skb, ...);
> fill_pid(stats);
>
> Shrinks taskatsks.o by 162 bytes.
>
> A stupid benchmark (send one million TASKSTATS_CMD_ATTR_PID) shows the
> difference,
>
> real user sys
> before:
> 4.02 0.06 3.96
> 4.02 0.04 3.98
> 4.02 0.04 3.97
> after:
> 3.86 0.08 3.78
> 3.88 0.10 3.77
> 3.89 0.09 3.80
>
> but this looks suspiciously good.
>
> Signed-off-by: Oleg Nesterov <[email protected]>
Acked-by: Shailabh Nagar <[email protected]>
>
> taskstats.c | 86 ++++++++++++++++++++++++++++++------------------------------
> 1 files changed, 44 insertions(+), 42 deletions(-)
>
> --- STATS/kernel/taskstats.c~6_nla 2006-11-01 14:00:03.000000000 +0300
> +++ STATS/kernel/taskstats.c 2006-11-01 21:14:39.000000000 +0300
> @@ -190,6 +190,7 @@ static int fill_pid(pid_t pid, struct ta
> } else
> get_task_struct(tsk);
>
> + memset(stats, 0, sizeof(*stats));
> /*
> * Each accounting subsystem adds calls to its functions to
> * fill in relevant parts of struct taskstsats as follows
> @@ -232,6 +233,8 @@ static int fill_tgid(pid_t tgid, struct
>
> if (first->signal->stats)
> memcpy(stats, first->signal->stats, sizeof(*stats));
> + else
> + memset(stats, 0, sizeof(*stats));
>
> tsk = first;
> do {
> @@ -348,9 +351,9 @@ static int parse(struct nlattr *na, cpum
> return ret;
> }
>
> -static int mk_reply(struct sk_buff *skb, int type, u32 pid, struct taskstats *stats)
> +static struct taskstats *mk_reply(struct sk_buff *skb, int type, u32 pid)
> {
> - struct nlattr *na;
> + struct nlattr *na, *ret;
> int aggr;
>
> aggr = TASKSTATS_TYPE_AGGR_TGID;
> @@ -358,20 +361,23 @@ static int mk_reply(struct sk_buff *skb,
> aggr = TASKSTATS_TYPE_AGGR_PID;
>
> na = nla_nest_start(skb, aggr);
> - NLA_PUT_U32(skb, type, pid);
> - NLA_PUT_TYPE(skb, struct taskstats, TASKSTATS_TYPE_STATS, *stats);
> + if (nla_put(skb, type, sizeof(pid), &pid) < 0)
> + goto err;
> + ret = nla_reserve(skb, TASKSTATS_TYPE_STATS, sizeof(struct taskstats));
> + if (!ret)
> + goto err;
> nla_nest_end(skb, na);
>
> - return 0;
> -nla_put_failure:
> - return -1;
> + return nla_data(ret);
> +err:
> + return NULL;
> }
>
> static int taskstats_user_cmd(struct sk_buff *skb, struct genl_info *info)
> {
> int rc = 0;
> struct sk_buff *rep_skb;
> - struct taskstats stats;
> + struct taskstats *stats;
> void *reply;
> size_t size;
> cpumask_t mask;
> @@ -394,36 +400,36 @@ static int taskstats_user_cmd(struct sk_
> size = nla_total_size(sizeof(u32)) +
> nla_total_size(sizeof(struct taskstats)) + nla_total_size(0);
>
> - memset(&stats, 0, sizeof(stats));
> rc = prepare_reply(info, TASKSTATS_CMD_NEW, &rep_skb, &reply, size);
> if (rc < 0)
> return rc;
>
> + rc = -EINVAL;
> if (info->attrs[TASKSTATS_CMD_ATTR_PID]) {
> u32 pid = nla_get_u32(info->attrs[TASKSTATS_CMD_ATTR_PID]);
> - rc = fill_pid(pid, NULL, &stats);
> - if (rc < 0)
> - goto err;
> + stats = mk_reply(rep_skb, TASKSTATS_TYPE_PID, pid);
> + if (!stats)
> + goto nla_err;
>
> - if (mk_reply(rep_skb, TASKSTATS_TYPE_PID, pid, &stats))
> - goto nla_put_failure;
> + rc = fill_pid(pid, NULL, stats);
> + if (rc < 0)
> + goto nla_err;
> } else if (info->attrs[TASKSTATS_CMD_ATTR_TGID]) {
> u32 tgid = nla_get_u32(info->attrs[TASKSTATS_CMD_ATTR_TGID]);
> - rc = fill_tgid(tgid, NULL, &stats);
> - if (rc < 0)
> - goto err;
> + stats = mk_reply(rep_skb, TASKSTATS_TYPE_TGID, tgid);
> + if (!stats)
> + goto nla_err;
>
> - if (mk_reply(rep_skb, TASKSTATS_TYPE_TGID, tgid, &stats))
> - goto nla_put_failure;
> - } else {
> - rc = -EINVAL;
> + rc = fill_tgid(tgid, NULL, stats);
> + if (rc < 0)
> + goto nla_err;
> + } else
> goto err;
> - }
>
> return send_reply(rep_skb, info->snd_pid);
>
> -nla_put_failure:
> - rc = genlmsg_cancel(rep_skb, reply);
> +nla_err:
> + genlmsg_cancel(rep_skb, reply);
> err:
> nlmsg_free(rep_skb);
> return rc;
> @@ -458,7 +464,7 @@ void taskstats_exit(struct task_struct *
> {
> int rc;
> struct listener_list *listeners;
> - struct taskstats *tidstats;
> + struct taskstats *stats;
> struct sk_buff *rep_skb;
> void *reply;
> size_t size;
> @@ -485,20 +491,17 @@ void taskstats_exit(struct task_struct *
> if (list_empty(&listeners->list))
> return;
>
> - tidstats = kmem_cache_zalloc(taskstats_cache, SLAB_KERNEL);
> - if (!tidstats)
> - return;
> -
> rc = prepare_reply(NULL, TASKSTATS_CMD_NEW, &rep_skb, &reply, size);
> if (rc < 0)
> - goto free_stats;
> + return;
>
> - rc = fill_pid(tsk->pid, tsk, tidstats);
> - if (rc < 0)
> - goto err_skb;
> + stats = mk_reply(rep_skb, TASKSTATS_TYPE_PID, tsk->pid);
> + if (!stats)
> + goto nla_err;
>
> - if (mk_reply(rep_skb, TASKSTATS_TYPE_PID, tsk->pid, tidstats))
> - goto nla_put_failure;
> + rc = fill_pid(tsk->pid, tsk, stats);
> + if (rc < 0)
> + goto nla_err;
>
> /*
> * Doesn't matter if tsk is the leader or the last group member leaving
> @@ -506,20 +509,19 @@ void taskstats_exit(struct task_struct *
> if (!is_thread_group || !group_dead)
> goto send;
>
> - if (mk_reply(rep_skb, TASKSTATS_TYPE_TGID, tsk->tgid, tsk->signal->stats))
> - goto nla_put_failure;
> + stats = mk_reply(rep_skb, TASKSTATS_TYPE_TGID, tsk->tgid);
> + if (!stats)
> + goto nla_err;
> +
> + memcpy(stats, tsk->signal->stats, sizeof(*stats));
>
> send:
> send_cpu_listeners(rep_skb, listeners);
> -free_stats:
> - kmem_cache_free(taskstats_cache, tidstats);
> return;
>
> -nla_put_failure:
> +nla_err:
> genlmsg_cancel(rep_skb, reply);
> -err_skb:
> nlmsg_free(rep_skb);
> - goto free_stats;
> }
>
> static struct genl_ops taskstats_ops = {
>