On 2024/10/09 17:05, Jason Wang wrote:
On Tue, Oct 8, 2024 at 2:55 PM Akihiko Odaki <akihiko.odaki@xxxxxxxxxx> wrote:Allow the guest to reuse the hash value to make receive steering consistent between the host and guest, and to save hash computation. Signed-off-by: Akihiko Odaki <akihiko.odaki@xxxxxxxxxx>I wonder if this would cause overhead when hash reporting is not enabled?
It only adds two branches in the data path. The first one is in tun_vnet_hash_report(), which determines to add the hash value to sk_buff. The second one is in tun_vnet_hdr_from_skb(), which determines to report the added hash value.
--- Documentation/networking/tuntap.rst | 7 +++ drivers/net/Kconfig | 1 + drivers/net/tap.c | 45 ++++++++++++++--Tile should be for tap as well or is this just for tun?
It is also for tap. I will update the title in v6.
drivers/net/tun.c | 46 ++++++++++++---- drivers/net/tun_vnet.h | 102 +++++++++++++++++++++++++++++++----- include/linux/if_tap.h | 2 + include/uapi/linux/if_tun.h | 48 +++++++++++++++++ 7 files changed, 223 insertions(+), 28 deletions(-) diff --git a/Documentation/networking/tuntap.rst b/Documentation/networking/tuntap.rst index 4d7087f727be..86b4ae8caa8a 100644 --- a/Documentation/networking/tuntap.rst +++ b/Documentation/networking/tuntap.rst @@ -206,6 +206,13 @@ enable is true we enable it, otherwise we disable it:: return ioctl(fd, TUNSETQUEUE, (void *)&ifr); } +3.4 Reference +------------- + +``linux/if_tun.h`` defines the interface described below: + +.. kernel-doc:: include/uapi/linux/if_tun.h + Universal TUN/TAP device driver Frequently Asked Question ========================================================= diff --git a/drivers/net/Kconfig b/drivers/net/Kconfig index 9920b3a68ed1..e2a7bd703550 100644 --- a/drivers/net/Kconfig +++ b/drivers/net/Kconfig @@ -395,6 +395,7 @@ config TUN tristate "Universal TUN/TAP device driver support" depends on INET select CRC32 + select SKB_EXTENSIONSThen we need this for macvtap at least as well?help TUN/TAP provides packet reception and transmission for user space programs. It can be viewed as a simple Point-to-Point or Ethernet diff --git a/drivers/net/tap.c b/drivers/net/tap.c index 9a34ceed0c2c..5e2fbe63ca47 100644 --- a/drivers/net/tap.c +++ b/drivers/net/tap.c @@ -179,6 +179,16 @@ static void tap_put_queue(struct tap_queue *q) sock_put(&q->sk); } +static struct virtio_net_hash *tap_add_hash(struct sk_buff *skb) +{ + return (struct virtio_net_hash *)skb->cb;Any reason that tap uses skb->cb but not skb extensions? (And is it safe to use that without cloning?)
tun adds virtio_net_hash to a skb in ndo_select_queue(), but it does not immediately put it into its ptr_ring; instead ndo_start_xmit() does so. It is hard to ensure that nobody modifies skb->cb between the two calls.
The situation is different for tap. tap_handle_frame() adds virtio_net_hash to a skb and immediately adds it in its ptr_ring so nobody should touch it between that.
+} + +static const struct virtio_net_hash *tap_find_hash(const struct sk_buff *skb) +{ + return (const struct virtio_net_hash *)skb->cb; +} + /* * Select a queue based on the rxq of the device on which this packet * arrived. If the incoming device is not mq, calculate a flow hash @@ -189,6 +199,7 @@ static void tap_put_queue(struct tap_queue *q) static struct tap_queue *tap_get_queue(struct tap_dev *tap, struct sk_buff *skb) { + struct flow_keys_basic keys_basic; struct tap_queue *queue = NULL; /* Access to taps array is protected by rcu, but access to numvtaps * isn't. Below we use it to lookup a queue, but treat it as a hint @@ -198,15 +209,32 @@ static struct tap_queue *tap_get_queue(struct tap_dev *tap, int numvtaps = READ_ONCE(tap->numvtaps); __u32 rxq; + *tap_add_hash(skb) = (struct virtio_net_hash) { .report = VIRTIO_NET_HASH_REPORT_NONE }; + if (!numvtaps) goto out; if (numvtaps == 1) goto single; + if (!skb->l4_hash && !skb->sw_hash) { + struct flow_keys keys; + + skb_flow_dissect_flow_keys(skb, &keys, FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); + rxq = flow_hash_from_keys(&keys); + keys_basic = (struct flow_keys_basic) { + .control = keys.control, + .basic = keys.basic + }; + } else { + skb_flow_dissect_flow_keys_basic(NULL, skb, &keys_basic, NULL, 0, 0, 0, + FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); + rxq = skb->hash; + } + /* Check if we can use flow to select a queue */ - rxq = skb_get_hash(skb); if (rxq) { + tun_vnet_hash_report(&tap->vnet_hash, skb, &keys_basic, rxq, tap_add_hash); queue = rcu_dereference(tap->taps[rxq % numvtaps]); goto out; } @@ -713,15 +741,16 @@ static ssize_t tap_put_user(struct tap_queue *q, int total; if (q->flags & IFF_VNET_HDR) { - struct virtio_net_hdr vnet_hdr; + struct virtio_net_hdr_v1_hash vnet_hdr; vnet_hdr_len = READ_ONCE(q->vnet_hdr_sz); - ret = tun_vnet_hdr_from_skb(q->flags, NULL, skb, &vnet_hdr); + ret = tun_vnet_hdr_from_skb(vnet_hdr_len, q->flags, NULL, skb, + tap_find_hash, &vnet_hdr); if (ret < 0) goto done; - ret = tun_vnet_hdr_put(vnet_hdr_len, iter, &vnet_hdr); + ret = tun_vnet_hdr_put(vnet_hdr_len, iter, &vnet_hdr, ret); if (ret < 0) goto done; } @@ -1025,7 +1054,13 @@ static long tap_ioctl(struct file *file, unsigned int cmd, return ret; default: - return tun_vnet_ioctl(&q->vnet_hdr_sz, &q->flags, cmd, sp); + rtnl_lock(); + tap = rtnl_dereference(q->tap); + ret = tun_vnet_ioctl(&q->vnet_hdr_sz, &q->flags, + tap ? &tap->vnet_hash : NULL, -EINVAL, + cmd, sp); + rtnl_unlock(); + return ret; } } diff --git a/drivers/net/tun.c b/drivers/net/tun.c index dd8799d19518..27308417b834 100644 --- a/drivers/net/tun.c +++ b/drivers/net/tun.c @@ -209,6 +209,7 @@ struct tun_struct { struct bpf_prog __rcu *xdp_prog; struct tun_prog __rcu *steering_prog; struct tun_prog __rcu *filter_prog; + struct tun_vnet_hash vnet_hash; struct ethtool_link_ksettings link_ksettings; /* init args */ struct file *file; @@ -451,6 +452,16 @@ static inline void tun_flow_save_rps_rxhash(struct tun_flow_entry *e, u32 hash) e->rps_rxhash = hash; } +static struct virtio_net_hash *tun_add_hash(struct sk_buff *skb) +{ + return skb_ext_add(skb, SKB_EXT_TUN_VNET_HASH); +} + +static const struct virtio_net_hash *tun_find_hash(const struct sk_buff *skb) +{ + return skb_ext_find(skb, SKB_EXT_TUN_VNET_HASH); +} + /* We try to identify a flow through its rxhash. The reason that * we do not check rxq no. is because some cards(e.g 82599), chooses * the rxq based on the txq where the last packet of the flow comes. As @@ -459,12 +470,17 @@ static inline void tun_flow_save_rps_rxhash(struct tun_flow_entry *e, u32 hash) */ static u16 tun_automq_select_queue(struct tun_struct *tun, struct sk_buff *skb) { + struct flow_keys keys; + struct flow_keys_basic keys_basic; struct tun_flow_entry *e; u32 txq, numqueues; numqueues = READ_ONCE(tun->numqueues); - txq = __skb_get_hash_symmetric(skb); + memset(&keys, 0, sizeof(keys)); + skb_flow_dissect(skb, &flow_keys_dissector_symmetric, &keys, 0); + + txq = flow_hash_from_keys(&keys); e = tun_flow_find(&tun->flows[tun_hashfn(txq)], txq); if (e) { tun_flow_save_rps_rxhash(e, txq); @@ -473,6 +489,13 @@ static u16 tun_automq_select_queue(struct tun_struct *tun, struct sk_buff *skb) txq = reciprocal_scale(txq, numqueues); } + keys_basic = (struct flow_keys_basic) { + .control = keys.control, + .basic = keys.basic + }; + tun_vnet_hash_report(&tun->vnet_hash, skb, &keys_basic, skb->l4_hash ? skb->hash : txq, + tun_add_hash);Is using txq required when not l4_hash is required by the virtio-spec?
It is a limitation of the implementation. A hardware driver may set a hash value with skb_set_hash(), which takes enum pkt_hash_types. The enum is defined as follows:
enum pkt_hash_types { PKT_HASH_TYPE_NONE, /* Undefined type */ PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */ PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */ PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */ };A hash value with PKT_HASH_TYPE_L2 must be ignored as the virtio spec does not have a corresponding hash type. The virtio spec has corresponding hash types for PKT_HASH_TYPE_L3 and PKT_HASH_TYPE_L4 so we should report them with the virtio_net header.
However, skb only tells whether the hash is PKT_HASH_TYPE_L4 or not. So tun reports skb->hash with a L4 hash type if the hash is PKT_HASH_TYPE_L4. Otherwise it ignores skb->hash and uses the hash value computed its own because it cannot tell if skb->hash is PKT_HASH_TYPE_L2 or PKT_HASH_TYPE_L3.
+ return txq; } @@ -1990,10 +2013,8 @@ static ssize_t tun_put_user_xdp(struct tun_struct *tun, size_t total; if (tun->flags & IFF_VNET_HDR) { - struct virtio_net_hdr gso = { 0 }; - vnet_hdr_sz = READ_ONCE(tun->vnet_hdr_sz); - ret = tun_vnet_hdr_put(vnet_hdr_sz, iter, &gso); + ret = tun_vnet_hdr_put(vnet_hdr_sz, iter, NULL, 0); if (ret < 0) return ret; } @@ -2018,7 +2039,6 @@ static ssize_t tun_put_user(struct tun_struct *tun, int vlan_offset = 0; int vlan_hlen = 0; int vnet_hdr_sz = 0; - int ret; if (skb_vlan_tag_present(skb)) vlan_hlen = VLAN_HLEN; @@ -2043,13 +2063,15 @@ static ssize_t tun_put_user(struct tun_struct *tun, } if (vnet_hdr_sz) { - struct virtio_net_hdr gso; + struct virtio_net_hdr_v1_hash gso; + int ret; - ret = tun_vnet_hdr_from_skb(tun->flags, tun->dev, skb, &gso); + ret = tun_vnet_hdr_from_skb(vnet_hdr_sz, tun->flags, tun->dev, skb, + tun_find_hash, &gso); if (ret < 0) goto done; - ret = tun_vnet_hdr_put(vnet_hdr_sz, iter, &gso); + ret = tun_vnet_hdr_put(vnet_hdr_sz, iter, &gso, ret); if (ret < 0) goto done; } @@ -3055,9 +3077,10 @@ static long __tun_chr_ioctl(struct file *file, unsigned int cmd, goto unlock; } - ret = -EBADFD; - if (!tun) + if (!tun) { + ret = tun_vnet_ioctl(NULL, NULL, NULL, -EBADFD, cmd, argp);This seems not elegant (passing three NULL pointers). Any reason we can't just modify __tun_chr_ioctl() instead of introducing things like tun_vnet_ioctl()?
tun_vnet_ioctl() is introduced with patch "tun: Unify vnet implementation". We can abandon unifying the ioctl handling if the interface looks too awkward.
goto unlock; + } netif_info(tun, drv, tun->dev, "tun_chr_ioctl cmd %u\n", cmd); @@ -3256,7 +3279,8 @@ static long __tun_chr_ioctl(struct file *file, unsigned int cmd, break; default: - ret = tun_vnet_ioctl(&tun->vnet_hdr_sz, &tun->flags, cmd, argp); + ret = tun_vnet_ioctl(&tun->vnet_hdr_sz, &tun->flags, + &tun->vnet_hash, -EINVAL, cmd, argp); } if (do_notify) diff --git a/drivers/net/tun_vnet.h b/drivers/net/tun_vnet.h index c40bde0fdf8c..589a97dd7d02 100644 --- a/drivers/net/tun_vnet.h +++ b/drivers/net/tun_vnet.h @@ -6,6 +6,9 @@ #define TUN_VNET_LE 0x80000000 #define TUN_VNET_BE 0x40000000 +typedef struct virtio_net_hash *(*tun_vnet_hash_add)(struct sk_buff *); +typedef const struct virtio_net_hash *(*tun_vnet_hash_find)(const struct sk_buff *); + static inline bool tun_vnet_legacy_is_little_endian(unsigned int flags) { return !(IS_ENABLED(CONFIG_TUN_VNET_CROSS_LE) && (flags & TUN_VNET_BE)) && @@ -59,18 +62,31 @@ static inline __virtio16 cpu_to_tun_vnet16(unsigned int flags, u16 val) } static inline long tun_vnet_ioctl(int *sz, unsigned int *flags, - unsigned int cmd, int __user *sp) + struct tun_vnet_hash *hash, long fallback, + unsigned int cmd, void __user *argp) { + static const struct tun_vnet_hash cap = { + .flags = TUN_VNET_HASH_REPORT, + .types = VIRTIO_NET_SUPPORTED_HASH_TYPES + };Let's find a way to reuse virtio-net uAPI instead of introducing new stuff to stress the management layer.
I found include/uapi/linux/virtio_net.h inappropriate for this ioctl. It has the following structure for hash reporting: struct virtio_net_hash_config { __le32 hash_types; /* for compatibility with virtio_net_rss_config */ __le16 reserved[4]; __u8 hash_key_length; __u8 hash_key_data[/* hash_key_length */]; }; We do not care hash_key_length and hash_key_data.It has the following structure for RSS and hash reporting when RSS is enabled:
struct virtio_net_rss_config { __le32 hash_types; __le16 indirection_table_mask; __le16 unclassified_queue; __le16 indirection_table[1/* + indirection_table_mask */]; __le16 max_tx_vq; __u8 hash_key_length; __u8 hash_key_data[/* hash_key_length */]; };This structure is hard to use as it has data members after the indirection_table flexible array. max_tx_vq is not our interest either.
I tested the usability of ioctl by actually using it in QEMU. The RFC patch series is available at:
https://patchew.org/QEMU/20240915-hash-v3-0-79cb08d28647@xxxxxxxxxx/
+ struct tun_vnet_hash hash_buf; + int __user *sp = argp; int s; switch (cmd) { case TUNGETVNETHDRSZ: + if (!sz) + return -EBADFD; + s = *sz; if (put_user(s, sp)) return -EFAULT; return 0; case TUNSETVNETHDRSZ: + if (!sz) + return -EBADFD; + if (get_user(s, sp)) return -EFAULT; if (s < (int)sizeof(struct virtio_net_hdr)) @@ -80,12 +96,18 @@ static inline long tun_vnet_ioctl(int *sz, unsigned int *flags, return 0; case TUNGETVNETLE: + if (!flags) + return -EBADFD; + s = !!(*flags & TUN_VNET_LE); if (put_user(s, sp)) return -EFAULT; return 0; case TUNSETVNETLE: + if (!flags) + return -EBADFD; + if (get_user(s, sp)) return -EFAULT; if (s) @@ -95,16 +117,56 @@ static inline long tun_vnet_ioctl(int *sz, unsigned int *flags, return 0; case TUNGETVNETBE: + if (!flags) + return -EBADFD; + return tun_vnet_get_be(*flags, sp); case TUNSETVNETBE: + if (!flags) + return -EBADFD; + return tun_vnet_set_be(flags, sp); + case TUNGETVNETHASHCAP: + return copy_to_user(argp, &cap, sizeof(cap)) ? -EFAULT : 0; + + case TUNSETVNETHASH: + if (!hash) + return -EBADFD; + + if (copy_from_user(&hash_buf, argp, sizeof(hash_buf))) + return -EFAULT; + + *hash = hash_buf; + return 0; + default: - return -EINVAL; + return fallback; } } +static inline void tun_vnet_hash_report(const struct tun_vnet_hash *hash, + struct sk_buff *skb, + const struct flow_keys_basic *keys, + u32 value, + tun_vnet_hash_add vnet_hash_add) +{ + struct virtio_net_hash *report; + + if (!(hash->flags & TUN_VNET_HASH_REPORT)) + return; + + report = vnet_hash_add(skb); + if (!report) + return; + + *report = (struct virtio_net_hash) { + .report = virtio_net_hash_report(hash->types, keys), + .value = value + }; +} + static inline int tun_vnet_hdr_get(int sz, unsigned int flags, struct iov_iter *from, struct virtio_net_hdr *hdr) @@ -130,15 +192,15 @@ static inline int tun_vnet_hdr_get(int sz, unsigned int flags, } static inline int tun_vnet_hdr_put(int sz, struct iov_iter *iter, - const struct virtio_net_hdr *hdr) + const void *hdr, int content_sz) { if (iov_iter_count(iter) < sz) return -EINVAL; - if (copy_to_iter(hdr, sizeof(*hdr), iter) != sizeof(*hdr)) + if (copy_to_iter(hdr, content_sz, iter) != content_sz) return -EFAULT; - if (iov_iter_zero(sz - sizeof(*hdr), iter) != sz - sizeof(*hdr)) + if (iov_iter_zero(sz - content_sz, iter) != sz - content_sz) return -EFAULT; return 0; @@ -151,32 +213,48 @@ static inline int tun_vnet_hdr_to_skb(unsigned int flags, return virtio_net_hdr_to_skb(skb, hdr, tun_vnet_is_little_endian(flags)); } -static inline int tun_vnet_hdr_from_skb(unsigned int flags, +static inline int tun_vnet_hdr_from_skb(int sz, unsigned int flags, const struct net_device *dev, const struct sk_buff *skb, - struct virtio_net_hdr *hdr) + tun_vnet_hash_find vnet_hash_find, + struct virtio_net_hdr_v1_hash *hdr) { int vlan_hlen = skb_vlan_tag_present(skb) ? VLAN_HLEN : 0; + const struct virtio_net_hash *report = sz < sizeof(struct virtio_net_hdr_v1_hash) ? + NULL : vnet_hash_find(skb); + int content_sz; + + if (report) { + content_sz = sizeof(struct virtio_net_hdr_v1_hash); + + *hdr = (struct virtio_net_hdr_v1_hash) { + .hdr = { .num_buffers = __cpu_to_virtio16(true, 1) }, + .hash_value = cpu_to_le32(report->value), + .hash_report = cpu_to_le16(report->report) + }; + } else { + content_sz = sizeof(struct virtio_net_hdr); + } - if (virtio_net_hdr_from_skb(skb, hdr, + if (virtio_net_hdr_from_skb(skb, (struct virtio_net_hdr *)hdr, tun_vnet_is_little_endian(flags), true, vlan_hlen)) { struct skb_shared_info *sinfo = skb_shinfo(skb); if (net_ratelimit()) { netdev_err(dev, "unexpected GSO type: 0x%x, gso_size %d, hdr_len %d\n", - sinfo->gso_type, tun_vnet16_to_cpu(flags, hdr->gso_size), - tun_vnet16_to_cpu(flags, hdr->hdr_len)); + sinfo->gso_type, tun_vnet16_to_cpu(flags, hdr->hdr.gso_size), + tun_vnet16_to_cpu(flags, hdr->hdr.hdr_len)); print_hex_dump(KERN_ERR, "tun: ", DUMP_PREFIX_NONE, 16, 1, skb->head, - min(tun_vnet16_to_cpu(flags, hdr->hdr_len), 64), true); + min(tun_vnet16_to_cpu(flags, hdr->hdr.hdr_len), 64), true); } WARN_ON_ONCE(1); return -EINVAL; } - return 0; + return content_sz; } #endif /* TUN_VNET_H */ diff --git a/include/linux/if_tap.h b/include/linux/if_tap.h index 553552fa635c..5bbb343a6dba 100644 --- a/include/linux/if_tap.h +++ b/include/linux/if_tap.h @@ -4,6 +4,7 @@ #include <net/sock.h> #include <linux/skb_array.h> +#include <uapi/linux/if_tun.h> struct file; struct socket; @@ -43,6 +44,7 @@ struct tap_dev { int numqueues; netdev_features_t tap_features; int minor; + struct tun_vnet_hash vnet_hash; void (*update_features)(struct tap_dev *tap, netdev_features_t features); void (*count_tx_dropped)(struct tap_dev *tap); diff --git a/include/uapi/linux/if_tun.h b/include/uapi/linux/if_tun.h index 287cdc81c939..d11e79b4e0dc 100644 --- a/include/uapi/linux/if_tun.h +++ b/include/uapi/linux/if_tun.h @@ -62,6 +62,34 @@ #define TUNSETCARRIER _IOW('T', 226, int) #define TUNGETDEVNETNS _IO('T', 227) +/** + * define TUNGETVNETHASHCAP - ioctl to get virtio_net hashing capability. + * + * The argument is a pointer to &struct tun_vnet_hash which will store the + * maximal virtio_net hashing configuration. + */ +#define TUNGETVNETHASHCAP _IOR('T', 228, struct tun_vnet_hash) + +/** + * define TUNSETVNETHASH - ioctl to configure virtio_net hashing + * + * The argument is a pointer to &struct tun_vnet_hash. + * + * The %TUN_VNET_HASH_REPORT flag set with this ioctl will be effective only + * after calling the %TUNSETVNETHDRSZ ioctl with a number greater than or equal + * to the size of &struct virtio_net_hdr_v1_hash.I think we don't need & here.
Documentation/doc-guide/kernel-doc.rst says &struct is a token for struct cross-reference.
+ * + * The members added to the legacy header by %TUN_VNET_HASH_REPORT flag will + * always be little-endian. + * + * This ioctl results in %EBADFD if the underlying device is deleted. It affects + * all queues attached to the same device. + * + * This ioctl currently has no effect on XDP packets and packets with + * queue_mapping set by TC.This needs to be fixed?
We don't use a hash value to select a queue in such a case so there is no point to report one.
+ */ +#define TUNSETVNETHASH _IOW('T', 229, struct tun_vnet_hash) + /* TUNSETIFF ifr flags */ #define IFF_TUN 0x0001 #define IFF_TAP 0x0002 @@ -115,4 +143,24 @@ struct tun_filter { __u8 addr[][ETH_ALEN]; }; +/** + * define TUN_VNET_HASH_REPORT - Request virtio_net hash reporting for vhost + */ +#define TUN_VNET_HASH_REPORT 0x0001 + +/** + * struct tun_vnet_hash - virtio_net hashing configuration + * @flags: + * Bitmask consists of %TUN_VNET_HASH_REPORT and %TUN_VNET_HASH_RSSCould we reuse TUNGETIFF by introduce new IFF_XXX stuffs?
That's certainly doable though I'm a bit worrying that exhausting all bits of IFF_XXX.
+ * @pad: + * Should be filled with zero before passing to %TUNSETVNETHASH + * @types: + * Bitmask of allowed hash typesWhat are they?
They are defined in the virtio spec and include/uapi/linux/virtio_net.h contains them:
#define VIRTIO_NET_RSS_HASH_TYPE_IPv4 (1 << 0) #define VIRTIO_NET_RSS_HASH_TYPE_TCPv4 (1 << 1) #define VIRTIO_NET_RSS_HASH_TYPE_UDPv4 (1 << 2) #define VIRTIO_NET_RSS_HASH_TYPE_IPv6 (1 << 3) #define VIRTIO_NET_RSS_HASH_TYPE_TCPv6 (1 << 4) #define VIRTIO_NET_RSS_HASH_TYPE_UDPv6 (1 << 5) #define VIRTIO_NET_RSS_HASH_TYPE_IP_EX (1 << 6) #define VIRTIO_NET_RSS_HASH_TYPE_TCP_EX (1 << 7) #define VIRTIO_NET_RSS_HASH_TYPE_UDP_EX (1 << 8)
+ */ +struct tun_vnet_hash { + __u16 flags; + __u8 pad[2]; + __u32 types; +}; + #endif /* _UAPI__IF_TUN_H */ -- 2.46.2Thanks