static void xenvif_fill_frags(struct xenvif *vif, struct sk_buff *skb)
{
struct skb_shared_info *shinfo = skb_shinfo(skb);
int nr_frags = shinfo->nr_frags;
int i;
for (i = 0; i < nr_frags; i++) {
skb_frag_t *frag = shinfo->frags + i;
struct xen_netif_tx_request *txp;
struct page *page;
u16 pending_idx;
pending_idx = frag_get_pending_idx(frag);
txp = &vif->pending_tx_info[pending_idx].req;
page = virt_to_page(idx_to_kaddr(vif, pending_idx));
__skb_fill_page_desc(skb, i, page, txp->offset, txp->size);
skb->len += txp->size;
skb->data_len += txp->size;
skb->truesize += txp->size;
/* Take an extra reference to offset xenvif_idx_release */
get_page(vif->mmap_pages[pending_idx]);
xenvif_idx_release(vif, pending_idx, XEN_NETIF_RSP_OKAY);
}
}
/**
* Allocate a new skb that can hold data of length @len.
*
* An SKB is created complely headerless. The linear part of an SKB
* is set apart for headers, and stream data is placed in paged fragments.
* Lower layers will take care of prepending all required headers.
*/
struct sk_buff *
ss_skb_alloc_pages(size_t len)
{
int i_frag, nr_frags = DIV_ROUND_UP(len, PAGE_SIZE);
struct sk_buff *skb;
BUG_ON(nr_frags > MAX_SKB_FRAGS);
skb = ss_skb_alloc();
if (!skb)
return NULL;
for (i_frag = 0; i_frag < nr_frags; ++i_frag) {
struct page *page = alloc_page(GFP_ATOMIC);
if (!page) {
kfree_skb(skb);
return NULL;
}
#if LINUX_VERSION_CODE < KERNEL_VERSION(4,1,12)
/* See __skb_alloc_pages() in include/linux/skbuff.h. */
if (page->pfmemalloc)
skb->pfmemalloc = true;
#endif
__skb_fill_page_desc(skb, i_frag, page, 0, 0);
skb_shinfo(skb)->nr_frags++;
}
return skb;
}
/*
* Make room for @shift fragments starting with slot @i. Then make
* a new fragment in slot @i that can hold @size bytes, and it set up.
*/
static int
__new_pgfrag(struct sk_buff *skb, int size, int i, int shift, TfwStr *it)
{
int off = 0;
struct page *page = NULL;
skb_frag_t *frag;
BUG_ON(i > MAX_SKB_FRAGS);
/*
* Try to find room for @size bytes in SKB fragments.
* If none found, then allocate a new page for the fragment.
*/
frag = __lookup_pgfrag_room(skb, size);
if (frag) {
page = skb_frag_page(frag);
off = ss_skb_frag_len(frag);
__skb_frag_ref(frag); /* get_page(page); */
} else {
page = alloc_page(GFP_ATOMIC);
if (!page)
return -ENOMEM;
}
/* Make room for @shift fragments starting with slot @i. */
if (__extend_pgfrags(skb, i, shift, it)) {
if (frag)
__skb_frag_unref(frag); /* put_page(page); */
else
__free_page(page);
return -ENOMEM;
}
/*
* When the requested slot is right outside the range of the
* array of paged fragments, then the new fragment is put as
* the first fragment of the next SKB.
*/
if (i == MAX_SKB_FRAGS) {
i = 0;
skb = it->skb;
}
/* Set up the new fragment in slot @i to hold @size bytes. */
__skb_fill_page_desc(skb, i, page, off, size);
ss_skb_adjust_data_len(skb, size);
return 0;
}
static void xenvif_fill_frags(struct xenvif *vif, struct sk_buff *skb)
{
struct skb_shared_info *shinfo = skb_shinfo(skb);
int nr_frags = shinfo->nr_frags;
int i;
u16 prev_pending_idx = INVALID_PENDING_IDX;
for (i = 0; i < nr_frags; i++) {
skb_frag_t *frag = shinfo->frags + i;
struct xen_netif_tx_request *txp;
struct page *page;
u16 pending_idx;
pending_idx = frag_get_pending_idx(frag);
/* If this is not the first frag, chain it to the previous*/
if (prev_pending_idx == INVALID_PENDING_IDX)
skb_shinfo(skb)->destructor_arg =
&callback_param(vif, pending_idx);
else
callback_param(vif, prev_pending_idx).ctx =
&callback_param(vif, pending_idx);
callback_param(vif, pending_idx).ctx = NULL;
prev_pending_idx = pending_idx;
txp = &vif->pending_tx_info[pending_idx].req;
page = virt_to_page(idx_to_kaddr(vif, pending_idx));
__skb_fill_page_desc(skb, i, page, txp->offset, txp->size);
skb->len += txp->size;
skb->data_len += txp->size;
skb->truesize += txp->size;
/* Take an extra reference to offset network stack's put_page */
get_page(vif->mmap_pages[pending_idx]);
}
/* FIXME: __skb_fill_page_desc set this to true because page->pfmemalloc
* overlaps with "index", and "mapping" is not set. I think mapping
* should be set. If delivered to local stack, it would drop this
* skb in sk_filter unless the socket has the right to use it.
*/
skb->pfmemalloc = false;
}
/**
* Delete @len (the value is positive now) bytes from @frag.
*
* @return 0 on success, -errno on failure.
* @return SKB in @it->skb if new SKB is allocated.
* @return pointer to data after the deleted area in @it->ptr.
* @return @it->flags is set if @it->ptr points to data in it->skb.
*/
static int
__split_pgfrag_del(struct sk_buff *skb, int i, int off, int len, TfwStr *it)
{
int tail_len;
struct sk_buff *skb_dst;
skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
struct skb_shared_info *si = skb_shinfo(skb);
SS_DBG("[%d]: %s: skb [%p] i [%d] off [%d] len [%d] fragsize [%d]\n",
smp_processor_id(), __func__,
skb, i, off, len, skb_frag_size(frag));
if (unlikely(off + len > skb_frag_size(frag))) {
SS_WARN("Attempt to delete too much\n");
return -EFAULT;
}
/* Fast path: delete a full fragment. */
if (!off && len == skb_frag_size(frag)) {
ss_skb_adjust_data_len(skb, -len);
__skb_frag_unref(frag);
if (i + 1 < si->nr_frags)
memmove(&si->frags[i], &si->frags[i + 1],
(si->nr_frags - i - 1) * sizeof(skb_frag_t));
--si->nr_frags;
goto lookup_next_ptr;
}
/* Fast path: delete the head part of a fragment. */
if (!off) {
frag->page_offset += len;
skb_frag_size_sub(frag, len);
ss_skb_adjust_data_len(skb, -len);
it->ptr = skb_frag_address(frag);
return 0;
}
/* Fast path: delete the tail part of a fragment. */
if (off + len == skb_frag_size(frag)) {
skb_frag_size_sub(frag, len);
ss_skb_adjust_data_len(skb, -len);
++i;
goto lookup_next_ptr;
}
/*
* Delete data in the middle of a fragment. After the data
* is deleted the fragment will contain only the head part,
* and the tail part is moved to another fragment.
* [frag @i] [frag @i+1 - tail data]
*
* Make room for a fragment right after the @i fragment
* to move the tail part of data there.
*/
if (__extend_pgfrags(skb, i + 1, 1, it))
return -EFAULT;
/* Find the SKB for tail data. */
skb_dst = (i < MAX_SKB_FRAGS - 1) ? skb : it->skb;
/* Calculate the length of the tail part. */
tail_len = skb_frag_size(frag) - off - len;
/* Trim the fragment with the head part. */
skb_frag_size_sub(frag, len + tail_len);
/* Make the fragment with the tail part. */
i = (i + 1) % MAX_SKB_FRAGS;
__skb_fill_page_desc(skb_dst, i, skb_frag_page(frag),
frag->page_offset + off + len, tail_len);
__skb_frag_ref(frag);
/* Adjust SKB data lengths. */
ss_skb_adjust_data_len(skb, -len);
if (skb != skb_dst) {
ss_skb_adjust_data_len(skb, -tail_len);
ss_skb_adjust_data_len(skb_dst, tail_len);
}
/* Get the SKB and the address of data after the deleted area. */
it->flags = (skb != skb_dst);
it->ptr = skb_frag_address(&skb_shinfo(skb_dst)->frags[i]);
return 0;
lookup_next_ptr:
/* Get the next fragment after the deleted fragment. */
if (i < si->nr_frags)
it->ptr = skb_frag_address(&si->frags[i]);
return 0;
}
/**
* Get room for @len bytes of data starting from offset @off
* in fragment @i.
*
* The room may be found in the preceding fragment if @off is zero.
* Otherwise, a new fragment is allocated and fragments around the
* fragment @i are rearranged so that data is not actually split
* and copied.
*
* Note: @off is always within the borders of fragment @i. It can
* point at the start of a fragment, but it can never point at the
* location right after the end of a fragment. In other words, @off
* can be zero, but it can not be equal to the size of fragment @i.
*
* @return 0 on success, -errno on failure.
* @return SKB in @it->skb if new SKB is allocated.
* @return pointer to the room for new data in @it->ptr.
* @return @it->flags is set if @it->ptr points to data in it->skb.
*/
static int
__split_pgfrag_add(struct sk_buff *skb, int i, int off, int len, TfwStr *it)
{
int tail_len;
struct sk_buff *skb_dst;
skb_frag_t *frag_dst, *frag = &skb_shinfo(skb)->frags[i];
SS_DBG("[%d]: %s: skb [%p] i [%d] off [%d] len [%d] fragsize [%d]\n",
smp_processor_id(), __func__,
skb, i, off, len, skb_frag_size(frag));
/*
* If @off is zero and there's a preceding page fragment,
* then try to append data to that fragment. Go for other
* solutions if there's no room.
*/
if (!off && i) {
frag_dst = __check_frag_room(skb, frag - 1, len);
if (frag_dst) {
/* Coalesce new data with the fragment. */
off = skb_frag_size(frag_dst);
skb_frag_size_add(frag_dst, len);
ss_skb_adjust_data_len(skb, len);
it->ptr = (char *)skb_frag_address(frag_dst) + off;
return 0;
}
}
/*
* Make a fragment that can hold @len bytes. If @off is
* zero, then data is added at the start of fragment @i.
* Make a fragment in slot @i, and the original fragment
* is shifted forward. If @off is not zero, then make
* a fragment in slot @i+1, and make an extra fragment
* in slot @i+2 to hold the tail data.
*/
if (__new_pgfrag(skb, len, i + !!off, 1 + !!off, it))
return -EFAULT;
/* If @off is zero, the job is done in __new_pgfrag(). */
if (!off) {
it->ptr = skb_frag_address(frag);
return 0;
}
/*
* If data is added in the middle of a fragment, then split
* the fragment. The head of the fragment stays there, and
* the tail of the fragment is moved to a new fragment.
* The fragment for new data is placed in between.
* [frag @i] [frag @i+1 - new data] [frag @i+2 - tail data]
* If @i is close to MAX_SKB_FRAGS, then new fragments may
* be located in another SKB.
*/
/* Find the SKB for tail data. */
skb_dst = (i < MAX_SKB_FRAGS - 2) ? skb : it->skb;
/* Calculate the length of the tail part. */
tail_len = skb_frag_size(frag) - off;
/* Trim the fragment with the head part. */
skb_frag_size_sub(frag, tail_len);
/* Make the fragment with the tail part. */
i = (i + 2) % MAX_SKB_FRAGS;
__skb_fill_page_desc(skb_dst, i, skb_frag_page(frag),
frag->page_offset + off, tail_len);
__skb_frag_ref(frag);
/* Adjust SKB data lengths. */
if (skb != skb_dst) {
ss_skb_adjust_data_len(skb, -tail_len);
ss_skb_adjust_data_len(skb_dst, tail_len);
}
/* Get the SKB and the address for new data. */
it->flags = !(i < MAX_SKB_FRAGS - 1);
frag_dst = it->flags ? &skb_shinfo(it->skb)->frags[0] : frag + 1;
it->ptr = skb_frag_address(frag_dst);
return 0;
}
/**
* The kernel may allocate a bit more memory for an SKB than what was
* requested (see ksize() call in __alloc_skb()). Use the extra memory
* if it's enough to hold @n bytes. Otherwise, allocate new linear data.
*
* @return 0 on success, -errno on failure.
* @return SKB in @it->skb if new SKB is allocated.
* @return pointer to the room for new data in @it->ptr if making room.
* @return pointer to data right after the deleted fragment in @it->ptr.
*/
static int
__split_linear_data(struct sk_buff *skb, char *pspt, int len, TfwStr *it)
{
int alloc = len > 0;
int tail_len = (char *)skb_tail_pointer(skb) - pspt;
struct page *page = virt_to_head_page(skb->head);
SS_DBG("[%d]: %s: skb [%p] pspt [%p] len [%d] tail_len [%d]\n",
smp_processor_id(), __func__, skb, pspt, len, tail_len);
BUG_ON(!skb->head_frag);
BUG_ON(tail_len <= 0);
BUG_ON(!(alloc | tail_len));
BUG_ON(-len > tail_len);
/*
* Quick and unlikely path: just advance the skb tail pointer.
* Note that this only works when we make room. When we remove,
* pspt points at the start of the data chunk to remove. In that
* case, tail_len can never be zero.
*/
if (unlikely(!tail_len && len <= ss_skb_tailroom(skb))) {
BUG_ON(len < 0);
it->ptr = ss_skb_put(skb, len);
return 0;
}
/*
* Quick and unlikely path: just move skb tail pointer backward.
* Note that this only works when we remove data, and the data
* is located exactly at the end of the linear part of an skb.
*/
if (unlikely((len < 0) && (tail_len == -len))) {
ss_skb_put(skb, len);
if (skb_is_nonlinear(skb))
it->ptr = skb_frag_address(&skb_shinfo(skb)->frags[0]);
return 0;
}
/*
* Data is inserted or deleted in the middle of the linear part,
* or there's insufficient room in the linear part of an SKB to
* insert @len bytes.
*
* Don't bother with skb tail room: if the linear part is large,
* then it's likely that we'll do some smaller data insertions
* later and go by the quick path above. Otherwise, the tail size
* is also small.
*
* The inserted data is placed in a fragment. The tail part is
* moved to yet another fragment. The linear part is trimmed to
* exclude the deleted data and the tail part.
*
* Do all allocations before moving the fragments to avoid complex
* rollback.
*/
if (alloc) {
if (__new_pgfrag(skb, len, 0, alloc + !!tail_len, it))
return -EFAULT;
} else {
if (__extend_pgfrags(skb, 0, 1, it))
return -EFAULT;
tail_len += len; /* @len is negative. */
}
if (tail_len) {
int tail_off = pspt - (char *)page_address(page);
/*
* Trim the linear part by |@len| bytes if data
* is deleted. Then trim it further to exclude
* the tail data. Finally, set up the fragment
* allotted above with the tail data.
*/
if (len < 0) {
tail_off -= len;
skb->tail += len;
skb->len += len;
}
skb->tail -= tail_len;
skb->data_len += tail_len;
skb->truesize += tail_len;
__skb_fill_page_desc(skb, alloc, page, tail_off, tail_len);
skb_frag_ref(skb, alloc); /* get_page(page); */
}
it->ptr = skb_frag_address(&skb_shinfo(skb)->frags[0]);
return 0;
}
static int __ip6_append_data(struct sock *sk,
struct flowi6 *fl6,
struct sk_buff_head *queue,
struct inet_cork *cork,
struct inet6_cork *v6_cork,
struct page_frag *pfrag,
int getfrag(void *from, char *to, int offset,
int len, int odd, struct sk_buff *skb),
void *from, int length, int transhdrlen,
unsigned int flags, int dontfrag)
{
struct sk_buff *skb, *skb_prev = NULL;
unsigned int maxfraglen, fragheaderlen, mtu, orig_mtu;
int exthdrlen = 0;
int dst_exthdrlen = 0;
int hh_len;
int copy;
int err;
int offset = 0;
__u8 tx_flags = 0;
u32 tskey = 0;
struct rt6_info *rt = (struct rt6_info *)cork->dst;
struct ipv6_txoptions *opt = v6_cork->opt;
int csummode = CHECKSUM_NONE;
unsigned int maxnonfragsize, headersize;
skb = skb_peek_tail(queue);
if (!skb) {
exthdrlen = opt ? opt->opt_flen : 0;
dst_exthdrlen = rt->dst.header_len - rt->rt6i_nfheader_len;
}
mtu = cork->fragsize;
orig_mtu = mtu;
hh_len = LL_RESERVED_SPACE(rt->dst.dev);
fragheaderlen = sizeof(struct ipv6hdr) + rt->rt6i_nfheader_len +
(opt ? opt->opt_nflen : 0);
maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen -
sizeof(struct frag_hdr);
headersize = sizeof(struct ipv6hdr) +
(opt ? opt->opt_flen + opt->opt_nflen : 0) +
(dst_allfrag(&rt->dst) ?
sizeof(struct frag_hdr) : 0) +
rt->rt6i_nfheader_len;
if (cork->length + length > mtu - headersize && dontfrag &&
(sk->sk_protocol == IPPROTO_UDP ||
sk->sk_protocol == IPPROTO_RAW)) {
ipv6_local_rxpmtu(sk, fl6, mtu - headersize +
sizeof(struct ipv6hdr));
goto emsgsize;
}
if (ip6_sk_ignore_df(sk))
maxnonfragsize = sizeof(struct ipv6hdr) + IPV6_MAXPLEN;
else
maxnonfragsize = mtu;
if (cork->length + length > maxnonfragsize - headersize) {
emsgsize:
ipv6_local_error(sk, EMSGSIZE, fl6,
mtu - headersize +
sizeof(struct ipv6hdr));
return -EMSGSIZE;
}
/* CHECKSUM_PARTIAL only with no extension headers and when
* we are not going to fragment
*/
if (transhdrlen && sk->sk_protocol == IPPROTO_UDP &&
headersize == sizeof(struct ipv6hdr) &&
length < mtu - headersize &&
!(flags & MSG_MORE) &&
rt->dst.dev->features & NETIF_F_V6_CSUM)
csummode = CHECKSUM_PARTIAL;
if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_RAW) {
sock_tx_timestamp(sk, &tx_flags);
if (tx_flags & SKBTX_ANY_SW_TSTAMP &&
sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)
tskey = sk->sk_tskey++;
}
/*
* Let's try using as much space as possible.
* Use MTU if total length of the message fits into the MTU.
* Otherwise, we need to reserve fragment header and
* fragment alignment (= 8-15 octects, in total).
*
* Note that we may need to "move" the data from the tail of
* of the buffer to the new fragment when we split
* the message.
*
* FIXME: It may be fragmented into multiple chunks
* at once if non-fragmentable extension headers
* are too large.
* --yoshfuji
*/
cork->length += length;
if (((length > mtu) ||
(skb && skb_is_gso(skb))) &&
(sk->sk_protocol == IPPROTO_UDP) &&
(rt->dst.dev->features & NETIF_F_UFO) &&
(sk->sk_type == SOCK_DGRAM) && !udp_get_no_check6_tx(sk)) {
err = ip6_ufo_append_data(sk, queue, getfrag, from, length,
hh_len, fragheaderlen, exthdrlen,
transhdrlen, mtu, flags, fl6);
if (err)
goto error;
return 0;
}
if (!skb)
goto alloc_new_skb;
while (length > 0) {
/* Check if the remaining data fits into current packet. */
copy = (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len;
if (copy < length)
copy = maxfraglen - skb->len;
if (copy <= 0) {
char *data;
unsigned int datalen;
unsigned int fraglen;
unsigned int fraggap;
unsigned int alloclen;
alloc_new_skb:
/* There's no room in the current skb */
if (skb)
fraggap = skb->len - maxfraglen;
else
fraggap = 0;
/* update mtu and maxfraglen if necessary */
if (!skb || !skb_prev)
ip6_append_data_mtu(&mtu, &maxfraglen,
fragheaderlen, skb, rt,
orig_mtu);
skb_prev = skb;
/*
* If remaining data exceeds the mtu,
* we know we need more fragment(s).
*/
datalen = length + fraggap;
if (datalen > (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen)
datalen = maxfraglen - fragheaderlen - rt->dst.trailer_len;
if ((flags & MSG_MORE) &&
!(rt->dst.dev->features&NETIF_F_SG))
alloclen = mtu;
else
alloclen = datalen + fragheaderlen;
alloclen += dst_exthdrlen;
if (datalen != length + fraggap) {
/*
* this is not the last fragment, the trailer
* space is regarded as data space.
*/
datalen += rt->dst.trailer_len;
}
alloclen += rt->dst.trailer_len;
fraglen = datalen + fragheaderlen;
/*
* We just reserve space for fragment header.
* Note: this may be overallocation if the message
* (without MSG_MORE) fits into the MTU.
*/
alloclen += sizeof(struct frag_hdr);
if (transhdrlen) {
skb = sock_alloc_send_skb(sk,
alloclen + hh_len,
(flags & MSG_DONTWAIT), &err);
} else {
skb = NULL;
if (atomic_read(&sk->sk_wmem_alloc) <=
2 * sk->sk_sndbuf)
skb = sock_wmalloc(sk,
alloclen + hh_len, 1,
sk->sk_allocation);
if (unlikely(!skb))
err = -ENOBUFS;
}
if (!skb)
goto error;
/*
* Fill in the control structures
*/
skb->protocol = htons(ETH_P_IPV6);
skb->ip_summed = csummode;
skb->csum = 0;
/* reserve for fragmentation and ipsec header */
skb_reserve(skb, hh_len + sizeof(struct frag_hdr) +
dst_exthdrlen);
/* Only the initial fragment is time stamped */
skb_shinfo(skb)->tx_flags = tx_flags;
tx_flags = 0;
skb_shinfo(skb)->tskey = tskey;
tskey = 0;
/*
* Find where to start putting bytes
*/
data = skb_put(skb, fraglen);
skb_set_network_header(skb, exthdrlen);
data += fragheaderlen;
skb->transport_header = (skb->network_header +
fragheaderlen);
if (fraggap) {
skb->csum = skb_copy_and_csum_bits(
skb_prev, maxfraglen,
data + transhdrlen, fraggap, 0);
skb_prev->csum = csum_sub(skb_prev->csum,
skb->csum);
data += fraggap;
pskb_trim_unique(skb_prev, maxfraglen);
}
copy = datalen - transhdrlen - fraggap;
if (copy < 0) {
err = -EINVAL;
kfree_skb(skb);
goto error;
} else if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) {
err = -EFAULT;
kfree_skb(skb);
goto error;
}
offset += copy;
length -= datalen - fraggap;
transhdrlen = 0;
exthdrlen = 0;
dst_exthdrlen = 0;
/*
* Put the packet on the pending queue
*/
__skb_queue_tail(queue, skb);
continue;
}
if (copy > length)
copy = length;
if (!(rt->dst.dev->features&NETIF_F_SG)) {
unsigned int off;
off = skb->len;
if (getfrag(from, skb_put(skb, copy),
offset, copy, off, skb) < 0) {
__skb_trim(skb, off);
err = -EFAULT;
goto error;
}
} else {
int i = skb_shinfo(skb)->nr_frags;
err = -ENOMEM;
if (!sk_page_frag_refill(sk, pfrag))
goto error;
if (!skb_can_coalesce(skb, i, pfrag->page,
pfrag->offset)) {
err = -EMSGSIZE;
if (i == MAX_SKB_FRAGS)
goto error;
__skb_fill_page_desc(skb, i, pfrag->page,
pfrag->offset, 0);
skb_shinfo(skb)->nr_frags = ++i;
get_page(pfrag->page);
}
copy = min_t(int, copy, pfrag->size - pfrag->offset);
if (getfrag(from,
page_address(pfrag->page) + pfrag->offset,
offset, copy, skb->len, skb) < 0)
goto error_efault;
pfrag->offset += copy;
skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy);
skb->len += copy;
skb->data_len += copy;
skb->truesize += copy;
atomic_add(copy, &sk->sk_wmem_alloc);
}
offset += copy;
length -= copy;
}
return 0;
error_efault:
err = -EFAULT;
error:
cork->length -= length;
IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
return err;
}
int ip6_append_data(struct sock *sk, int getfrag(void *from, char *to,
int offset, int len, int odd, struct sk_buff *skb),
void *from, int length, int transhdrlen,
int hlimit, int tclass, struct ipv6_txoptions *opt, struct flowi6 *fl6,
struct rt6_info *rt, unsigned int flags, int dontfrag)
{
struct inet_sock *inet = inet_sk(sk);
struct ipv6_pinfo *np = inet6_sk(sk);
struct inet_cork *cork;
struct sk_buff *skb, *skb_prev = NULL;
unsigned int maxfraglen, fragheaderlen, mtu;
int exthdrlen;
int dst_exthdrlen;
int hh_len;
int copy;
int err;
int offset = 0;
__u8 tx_flags = 0;
if (flags&MSG_PROBE)
return 0;
cork = &inet->cork.base;
if (skb_queue_empty(&sk->sk_write_queue)) {
/*
* setup for corking
*/
if (opt) {
if (WARN_ON(np->cork.opt))
return -EINVAL;
np->cork.opt = kzalloc(opt->tot_len, sk->sk_allocation);
if (unlikely(np->cork.opt == NULL))
return -ENOBUFS;
np->cork.opt->tot_len = opt->tot_len;
np->cork.opt->opt_flen = opt->opt_flen;
np->cork.opt->opt_nflen = opt->opt_nflen;
np->cork.opt->dst0opt = ip6_opt_dup(opt->dst0opt,
sk->sk_allocation);
if (opt->dst0opt && !np->cork.opt->dst0opt)
return -ENOBUFS;
np->cork.opt->dst1opt = ip6_opt_dup(opt->dst1opt,
sk->sk_allocation);
if (opt->dst1opt && !np->cork.opt->dst1opt)
return -ENOBUFS;
np->cork.opt->hopopt = ip6_opt_dup(opt->hopopt,
sk->sk_allocation);
if (opt->hopopt && !np->cork.opt->hopopt)
return -ENOBUFS;
np->cork.opt->srcrt = ip6_rthdr_dup(opt->srcrt,
sk->sk_allocation);
if (opt->srcrt && !np->cork.opt->srcrt)
return -ENOBUFS;
/* need source address above miyazawa*/
}
dst_hold(&rt->dst);
cork->dst = &rt->dst;
inet->cork.fl.u.ip6 = *fl6;
np->cork.hop_limit = hlimit;
np->cork.tclass = tclass;
if (rt->dst.flags & DST_XFRM_TUNNEL)
mtu = np->pmtudisc == IPV6_PMTUDISC_PROBE ?
rt->dst.dev->mtu : dst_mtu(&rt->dst);
else
mtu = np->pmtudisc == IPV6_PMTUDISC_PROBE ?
rt->dst.dev->mtu : dst_mtu(rt->dst.path);
if (np->frag_size < mtu) {
if (np->frag_size)
mtu = np->frag_size;
}
cork->fragsize = mtu;
if (dst_allfrag(rt->dst.path))
cork->flags |= IPCORK_ALLFRAG;
cork->length = 0;
exthdrlen = (opt ? opt->opt_flen : 0);
length += exthdrlen;
transhdrlen += exthdrlen;
dst_exthdrlen = rt->dst.header_len - rt->rt6i_nfheader_len;
} else {
rt = (struct rt6_info *)cork->dst;
fl6 = &inet->cork.fl.u.ip6;
opt = np->cork.opt;
transhdrlen = 0;
exthdrlen = 0;
dst_exthdrlen = 0;
mtu = cork->fragsize;
}
hh_len = LL_RESERVED_SPACE(rt->dst.dev);
fragheaderlen = sizeof(struct ipv6hdr) + rt->rt6i_nfheader_len +
(opt ? opt->opt_nflen : 0);
maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen - sizeof(struct frag_hdr);
if (mtu <= sizeof(struct ipv6hdr) + IPV6_MAXPLEN) {
if (cork->length + length > sizeof(struct ipv6hdr) + IPV6_MAXPLEN - fragheaderlen) {
ipv6_local_error(sk, EMSGSIZE, fl6, mtu-exthdrlen);
return -EMSGSIZE;
}
}
/* For UDP, check if TX timestamp is enabled */
if (sk->sk_type == SOCK_DGRAM)
sock_tx_timestamp(sk, &tx_flags);
/*
* Let's try using as much space as possible.
* Use MTU if total length of the message fits into the MTU.
* Otherwise, we need to reserve fragment header and
* fragment alignment (= 8-15 octects, in total).
*
* Note that we may need to "move" the data from the tail of
* of the buffer to the new fragment when we split
* the message.
*
* FIXME: It may be fragmented into multiple chunks
* at once if non-fragmentable extension headers
* are too large.
* --yoshfuji
*/
if ((length > mtu) && dontfrag && (sk->sk_protocol == IPPROTO_UDP ||
sk->sk_protocol == IPPROTO_RAW)) {
ipv6_local_rxpmtu(sk, fl6, mtu-exthdrlen);
return -EMSGSIZE;
}
skb = skb_peek_tail(&sk->sk_write_queue);
cork->length += length;
if (((length > mtu) ||
(skb && skb_has_frags(skb))) &&
(sk->sk_protocol == IPPROTO_UDP) &&
(rt->dst.dev->features & NETIF_F_UFO)) {
err = ip6_ufo_append_data(sk, getfrag, from, length,
hh_len, fragheaderlen,
transhdrlen, mtu, flags, rt);
if (err)
goto error;
return 0;
}
if (!skb)
goto alloc_new_skb;
while (length > 0) {
/* Check if the remaining data fits into current packet. */
copy = (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len;
if (copy < length)
copy = maxfraglen - skb->len;
if (copy <= 0) {
char *data;
unsigned int datalen;
unsigned int fraglen;
unsigned int fraggap;
unsigned int alloclen;
alloc_new_skb:
/* There's no room in the current skb */
if (skb)
fraggap = skb->len - maxfraglen;
else
fraggap = 0;
/* update mtu and maxfraglen if necessary */
if (skb == NULL || skb_prev == NULL)
ip6_append_data_mtu(&mtu, &maxfraglen,
fragheaderlen, skb, rt,
np->pmtudisc ==
IPV6_PMTUDISC_PROBE);
skb_prev = skb;
/*
* If remaining data exceeds the mtu,
* we know we need more fragment(s).
*/
datalen = length + fraggap;
if (datalen > (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen)
datalen = maxfraglen - fragheaderlen - rt->dst.trailer_len;
if ((flags & MSG_MORE) &&
!(rt->dst.dev->features&NETIF_F_SG))
alloclen = mtu;
else
alloclen = datalen + fragheaderlen;
alloclen += dst_exthdrlen;
if (datalen != length + fraggap) {
/*
* this is not the last fragment, the trailer
* space is regarded as data space.
*/
datalen += rt->dst.trailer_len;
}
alloclen += rt->dst.trailer_len;
fraglen = datalen + fragheaderlen;
/*
* We just reserve space for fragment header.
* Note: this may be overallocation if the message
* (without MSG_MORE) fits into the MTU.
*/
alloclen += sizeof(struct frag_hdr);
if (transhdrlen) {
skb = sock_alloc_send_skb(sk,
alloclen + hh_len,
(flags & MSG_DONTWAIT), &err);
} else {
skb = NULL;
if (atomic_read(&sk->sk_wmem_alloc) <=
2 * sk->sk_sndbuf)
skb = sock_wmalloc(sk,
alloclen + hh_len, 1,
sk->sk_allocation);
if (unlikely(skb == NULL))
err = -ENOBUFS;
else {
/* Only the initial fragment
* is time stamped.
*/
tx_flags = 0;
}
}
if (skb == NULL)
goto error;
/*
* Fill in the control structures
*/
skb->ip_summed = CHECKSUM_NONE;
skb->csum = 0;
/* reserve for fragmentation and ipsec header */
skb_reserve(skb, hh_len + sizeof(struct frag_hdr) +
dst_exthdrlen);
if (sk->sk_type == SOCK_DGRAM)
skb_shinfo(skb)->tx_flags = tx_flags;
/*
* Find where to start putting bytes
*/
data = skb_put(skb, fraglen);
skb_set_network_header(skb, exthdrlen);
data += fragheaderlen;
skb->transport_header = (skb->network_header +
fragheaderlen);
if (fraggap) {
skb->csum = skb_copy_and_csum_bits(
skb_prev, maxfraglen,
data + transhdrlen, fraggap, 0);
skb_prev->csum = csum_sub(skb_prev->csum,
skb->csum);
data += fraggap;
pskb_trim_unique(skb_prev, maxfraglen);
}
copy = datalen - transhdrlen - fraggap;
if (copy < 0) {
err = -EINVAL;
kfree_skb(skb);
goto error;
} else if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) {
err = -EFAULT;
kfree_skb(skb);
goto error;
}
offset += copy;
length -= datalen - fraggap;
transhdrlen = 0;
exthdrlen = 0;
dst_exthdrlen = 0;
/*
* Put the packet on the pending queue
*/
__skb_queue_tail(&sk->sk_write_queue, skb);
continue;
}
if (copy > length)
copy = length;
if (!(rt->dst.dev->features&NETIF_F_SG)) {
unsigned int off;
off = skb->len;
if (getfrag(from, skb_put(skb, copy),
offset, copy, off, skb) < 0) {
__skb_trim(skb, off);
err = -EFAULT;
goto error;
}
} else {
int i = skb_shinfo(skb)->nr_frags;
struct page_frag *pfrag = sk_page_frag(sk);
err = -ENOMEM;
if (!sk_page_frag_refill(sk, pfrag))
goto error;
if (!skb_can_coalesce(skb, i, pfrag->page,
pfrag->offset)) {
err = -EMSGSIZE;
if (i == MAX_SKB_FRAGS)
goto error;
__skb_fill_page_desc(skb, i, pfrag->page,
pfrag->offset, 0);
skb_shinfo(skb)->nr_frags = ++i;
get_page(pfrag->page);
}
copy = min_t(int, copy, pfrag->size - pfrag->offset);
if (getfrag(from,
page_address(pfrag->page) + pfrag->offset,
offset, copy, skb->len, skb) < 0)
goto error_efault;
pfrag->offset += copy;
skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy);
skb->len += copy;
skb->data_len += copy;
skb->truesize += copy;
atomic_add(copy, &sk->sk_wmem_alloc);
}
offset += copy;
length -= copy;
}
return 0;
error_efault:
err = -EFAULT;
error:
cork->length -= length;
IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
return err;
}
static void xennet_alloc_rx_buffers(struct net_device *dev)
{
unsigned short id;
struct netfront_info *np = netdev_priv(dev);
struct sk_buff *skb;
struct page *page;
int i, batch_target, notify;
RING_IDX req_prod = np->rx.req_prod_pvt;
grant_ref_t ref;
unsigned long pfn;
void *vaddr;
struct xen_netif_rx_request *req;
if (unlikely(!netif_carrier_ok(dev)))
return;
/*
* Allocate skbuffs greedily, even though we batch updates to the
* receive ring. This creates a less bursty demand on the memory
* allocator, so should reduce the chance of failed allocation requests
* both for ourself and for other kernel subsystems.
*/
batch_target = np->rx_target - (req_prod - np->rx.rsp_cons);
for (i = skb_queue_len(&np->rx_batch); i < batch_target; i++) {
skb = __netdev_alloc_skb(dev, RX_COPY_THRESHOLD + NET_IP_ALIGN,
GFP_ATOMIC | __GFP_NOWARN);
if (unlikely(!skb))
goto no_skb;
/* Align ip header to a 16 bytes boundary */
skb_reserve(skb, NET_IP_ALIGN);
page = alloc_page(GFP_ATOMIC | __GFP_NOWARN);
if (!page) {
kfree_skb(skb);
no_skb:
/* Any skbuffs queued for refill? Force them out. */
if (i != 0)
goto refill;
/* Could not allocate any skbuffs. Try again later. */
mod_timer(&np->rx_refill_timer,
jiffies + (HZ/10));
break;
}
__skb_fill_page_desc(skb, 0, page, 0, 0);
skb_shinfo(skb)->nr_frags = 1;
__skb_queue_tail(&np->rx_batch, skb);
}
/* Is the batch large enough to be worthwhile? */
if (i < (np->rx_target/2)) {
if (req_prod > np->rx.sring->req_prod)
goto push;
return;
}
/* Adjust our fill target if we risked running out of buffers. */
if (((req_prod - np->rx.sring->rsp_prod) < (np->rx_target / 4)) &&
((np->rx_target *= 2) > np->rx_max_target))
np->rx_target = np->rx_max_target;
refill:
for (i = 0; ; i++) {
skb = __skb_dequeue(&np->rx_batch);
if (skb == NULL)
break;
skb->dev = dev;
id = xennet_rxidx(req_prod + i);
BUG_ON(np->rx_skbs[id]);
np->rx_skbs[id] = skb;
ref = gnttab_claim_grant_reference(&np->gref_rx_head);
BUG_ON((signed short)ref < 0);
np->grant_rx_ref[id] = ref;
pfn = page_to_pfn(skb_frag_page(&skb_shinfo(skb)->frags[0]));
vaddr = page_address(skb_frag_page(&skb_shinfo(skb)->frags[0]));
req = RING_GET_REQUEST(&np->rx, req_prod + i);
gnttab_grant_foreign_access_ref(ref,
np->xbdev->otherend_id,
pfn_to_mfn(pfn),
0);
req->id = id;
req->gref = ref;
}
wmb(); /* barrier so backend seens requests */
/* Above is a suitable barrier to ensure backend will see requests. */
np->rx.req_prod_pvt = req_prod + i;
push:
RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&np->rx, notify);
if (notify)
notify_remote_via_irq(np->netdev->irq);
}
static void xennet_alloc_rx_buffers(struct net_device *dev)
{
unsigned short id;
struct netfront_info *np = netdev_priv(dev);
struct sk_buff *skb;
struct page *page;
int i, batch_target, notify;
RING_IDX req_prod = np->rx.req_prod_pvt;
grant_ref_t ref;
unsigned long pfn;
void *vaddr;
struct xen_netif_rx_request *req;
if (unlikely(!netif_carrier_ok(dev)))
return;
batch_target = np->rx_target - (req_prod - np->rx.rsp_cons);
for (i = skb_queue_len(&np->rx_batch); i < batch_target; i++) {
skb = __netdev_alloc_skb(dev, RX_COPY_THRESHOLD + NET_IP_ALIGN,
GFP_ATOMIC | __GFP_NOWARN);
if (unlikely(!skb))
goto no_skb;
skb_reserve(skb, NET_IP_ALIGN);
page = alloc_page(GFP_ATOMIC | __GFP_NOWARN);
if (!page) {
kfree_skb(skb);
no_skb:
if (i != 0)
goto refill;
mod_timer(&np->rx_refill_timer,
jiffies + (HZ/10));
break;
}
__skb_fill_page_desc(skb, 0, page, 0, 0);
skb_shinfo(skb)->nr_frags = 1;
__skb_queue_tail(&np->rx_batch, skb);
}
if (i < (np->rx_target/2)) {
if (req_prod > np->rx.sring->req_prod)
goto push;
return;
}
if (((req_prod - np->rx.sring->rsp_prod) < (np->rx_target / 4)) &&
((np->rx_target *= 2) > np->rx_max_target))
np->rx_target = np->rx_max_target;
refill:
for (i = 0; ; i++) {
skb = __skb_dequeue(&np->rx_batch);
if (skb == NULL)
break;
skb->dev = dev;
id = xennet_rxidx(req_prod + i);
BUG_ON(np->rx_skbs[id]);
np->rx_skbs[id] = skb;
ref = gnttab_claim_grant_reference(&np->gref_rx_head);
BUG_ON((signed short)ref < 0);
np->grant_rx_ref[id] = ref;
pfn = page_to_pfn(skb_frag_page(&skb_shinfo(skb)->frags[0]));
vaddr = page_address(skb_frag_page(&skb_shinfo(skb)->frags[0]));
req = RING_GET_REQUEST(&np->rx, req_prod + i);
gnttab_grant_foreign_access_ref(ref,
np->xbdev->otherend_id,
pfn_to_mfn(pfn),
0);
req->id = id;
req->gref = ref;
}
wmb();
np->rx.req_prod_pvt = req_prod + i;
push:
RING_PUSH_REQUESTS_AND_CHECK_NOTIFY(&np->rx, notify);
if (notify)
notify_remote_via_irq(np->netdev->irq);
}
