[RFC net-next io_uring 04/11] net: shrink struct ubuf_info

From: Pavel Begunkov
Date: Wed Aug 10 2022 - 11:52:26 EST

We can benefit from a smaller struct ubuf_info, so leave only mandatory
fields and let users to decide how they want to extend it. Convert
MSG_ZEROCOPY to struct ubuf_info_msgzc and remove duplicated fields.
This reduces the size from 48 bytes to just 16.

Signed-off-by: Pavel Begunkov <asml.silence@xxxxxxxxx>
---
include/linux/skbuff.h | 22 ++++------------------
net/core/skbuff.c | 38 +++++++++++++++++++++-----------------
net/ipv4/ip_output.c | 2 +-
net/ipv4/tcp.c | 2 +-
net/ipv6/ip6_output.c | 2 +-
5 files changed, 28 insertions(+), 38 deletions(-)

diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index f8ac3678dab8..afd7400d7f62 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -533,25 +533,8 @@ enum {
struct ubuf_info {
void (*callback)(struct sk_buff *, struct ubuf_info *,
bool zerocopy_success);
- union {
- struct {
- unsigned long desc;
- void *ctx;
- };
- struct {
- u32 id;
- u16 len;
- u16 zerocopy:1;
- u32 bytelen;
- };
- };
refcount_t refcnt;
u8 flags;
-
- struct mmpin {
- struct user_struct *user;
- unsigned int num_pg;
- } mmp;
};

struct ubuf_info_msgzc {
@@ -570,7 +553,10 @@ struct ubuf_info_msgzc {
};
};

- struct mmpin mmp;
+ struct mmpin {
+ struct user_struct *user;
+ unsigned int num_pg;
+ } mmp;
};

#define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index 974bbbbe7138..b047a773acd7 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -1183,7 +1183,7 @@ EXPORT_SYMBOL_GPL(mm_unaccount_pinned_pages);

static struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size)
{
- struct ubuf_info *uarg;
+ struct ubuf_info_msgzc *uarg;
struct sk_buff *skb;

WARN_ON_ONCE(!in_task());
@@ -1201,19 +1201,19 @@ static struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size)
return NULL;
}

- uarg->callback = msg_zerocopy_callback;
+ uarg->ubuf.callback = msg_zerocopy_callback;
uarg->id = ((u32)atomic_inc_return(&sk->sk_zckey)) - 1;
uarg->len = 1;
uarg->bytelen = size;
uarg->zerocopy = 1;
- uarg->flags = SKBFL_ZEROCOPY_FRAG | SKBFL_DONT_ORPHAN;
- refcount_set(&uarg->refcnt, 1);
+ uarg->ubuf.flags = SKBFL_ZEROCOPY_FRAG | SKBFL_DONT_ORPHAN;
+ refcount_set(&uarg->ubuf.refcnt, 1);
sock_hold(sk);

- return uarg;
+ return &uarg->ubuf;
}

-static inline struct sk_buff *skb_from_uarg(struct ubuf_info *uarg)
+static inline struct sk_buff *skb_from_uarg(struct ubuf_info_msgzc *uarg)
{
return container_of((void *)uarg, struct sk_buff, cb);
}
@@ -1222,6 +1222,7 @@ struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
struct ubuf_info *uarg)
{
if (uarg) {
+ struct ubuf_info_msgzc *uarg_zc;
const u32 byte_limit = 1 << 19; /* limit to a few TSO */
u32 bytelen, next;

@@ -1237,8 +1238,9 @@ struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
return NULL;
}

- bytelen = uarg->bytelen + size;
- if (uarg->len == USHRT_MAX - 1 || bytelen > byte_limit) {
+ uarg_zc = uarg_to_msgzc(uarg);
+ bytelen = uarg_zc->bytelen + size;
+ if (uarg_zc->len == USHRT_MAX - 1 || bytelen > byte_limit) {
/* TCP can create new skb to attach new uarg */
if (sk->sk_type == SOCK_STREAM)
goto new_alloc;
@@ -1246,11 +1248,11 @@ struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
}

next = (u32)atomic_read(&sk->sk_zckey);
- if ((u32)(uarg->id + uarg->len) == next) {
- if (mm_account_pinned_pages(&uarg->mmp, size))
+ if ((u32)(uarg_zc->id + uarg_zc->len) == next) {
+ if (mm_account_pinned_pages(&uarg_zc->mmp, size))
return NULL;
- uarg->len++;
- uarg->bytelen = bytelen;
+ uarg_zc->len++;
+ uarg_zc->bytelen = bytelen;
atomic_set(&sk->sk_zckey, ++next);

/* no extra ref when appending to datagram (MSG_MORE) */
@@ -1286,7 +1288,7 @@ static bool skb_zerocopy_notify_extend(struct sk_buff *skb, u32 lo, u16 len)
return true;
}

-static void __msg_zerocopy_callback(struct ubuf_info *uarg)
+static void __msg_zerocopy_callback(struct ubuf_info_msgzc *uarg)
{
struct sk_buff *tail, *skb = skb_from_uarg(uarg);
struct sock_exterr_skb *serr;
@@ -1339,19 +1341,21 @@ static void __msg_zerocopy_callback(struct ubuf_info *uarg)
void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg,
bool success)
{
- uarg->zerocopy = uarg->zerocopy & success;
+ struct ubuf_info_msgzc *uarg_zc = uarg_to_msgzc(uarg);
+
+ uarg_zc->zerocopy = uarg_zc->zerocopy & success;

if (refcount_dec_and_test(&uarg->refcnt))
- __msg_zerocopy_callback(uarg);
+ __msg_zerocopy_callback(uarg_zc);
}
EXPORT_SYMBOL_GPL(msg_zerocopy_callback);

void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref)
{
- struct sock *sk = skb_from_uarg(uarg)->sk;
+ struct sock *sk = skb_from_uarg(uarg_to_msgzc(uarg))->sk;

atomic_dec(&sk->sk_zckey);
- uarg->len--;
+ uarg_to_msgzc(uarg)->len--;

if (have_uref)
msg_zerocopy_callback(NULL, uarg, true);
diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c
index d7bd1daf022b..546897a4b4fa 100644
--- a/net/ipv4/ip_output.c
+++ b/net/ipv4/ip_output.c
@@ -1043,7 +1043,7 @@ static int __ip_append_data(struct sock *sk,
paged = true;
zc = true;
} else {
- uarg->zerocopy = 0;
+ uarg_to_msgzc(uarg)->zerocopy = 0;
skb_zcopy_set(skb, uarg, &extra_uref);
}
}
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 970e9a2cca4a..3152da8f4763 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -1239,7 +1239,7 @@ int tcp_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t size)
}
zc = sk->sk_route_caps & NETIF_F_SG;
if (!zc)
- uarg->zerocopy = 0;
+ uarg_to_msgzc(uarg)->zerocopy = 0;
}
}

diff --git a/net/ipv6/ip6_output.c b/net/ipv6/ip6_output.c
index 897ca4f9b791..6d4f01a0cf6e 100644
--- a/net/ipv6/ip6_output.c
+++ b/net/ipv6/ip6_output.c
@@ -1568,7 +1568,7 @@ static int __ip6_append_data(struct sock *sk,
paged = true;
zc = true;
} else {
- uarg->zerocopy = 0;
+ uarg_to_msgzc(uarg)->zerocopy = 0;
skb_zcopy_set(skb, uarg, &extra_uref);
}
}
--
2.37.0