net: af_packet: don't call tpacket_destruct_skb() until the skb is sent out
Subsystems (3)
PATCH v2 1/1 net: af_packet: don't call tpacket_destruct_skb() until the skb is sent out
Since skb->destructor() is used to account socket memory, and maybe called before the skb is sent out, a corrupt skb maybe sent out finally.
A new destructor is added into structure skb_shared_info(), and it won't be called until the last reference to the data of an skb is put. af_packet uses this destructor instead.
---
v2: avoid kmalloc/kfree
include/linux/skbuff.h | 3 ++- net/core/skbuff.c | 19 ++++++++++++++----- net/packet/af_packet.c | 24 +++++++++++------------- 3 files changed, 27 insertions(+), 19 deletions(-)
include/linux/skbuff.h +2 -1
@@ -191,6 +191,7 @@ struct skb_shared_info {__u8 tx_flags;struct sk_buff *frag_list;struct skb_shared_hwtstamps hwtstamps; +void (*destructor)(struct sk_buff *skb);/* * Warning : all fields before dataref are cleared in __alloc_skb()
@@ -199,7 +200,7 @@ struct skb_shared_info {/* Intermediate layers must ensure that destructor_arg * remains valid until skb destructor */ -void * destructor_arg; +void *destructor_arg[2 ];/* must be last field, see pskb_expand_head() */ skb_frag_t frags[MAX_SKB_FRAGS ]; };
net/core/skbuff.c +14 -5
@@ -332,10 +332,14 @@ static void skb_release_data(struct sk_buff *skb)if (!skb->cloned || !atomic_sub_return (skb->nohdr ? (1 <<SKB_DATAREF_SHIFT ) +1 :1 , &skb_shinfo (skb)->dataref)) { -if (skb_shinfo (skb)->nr_frags) { +struct skb_shared_info *shinfo =skb_shinfo (skb); + +if (shinfo->destructor) + shinfo->destructor (skb); +if (shinfo->nr_frags) {int i; -for (i =0 ; i <skb_shinfo (skb)->nr_frags; i++) -put_page (skb_shinfo (skb)->frags[i].page); +for (i =0 ; i < shinfo->nr_frags; i++) +put_page (shinfo->frags[i].page); }if (skb_has_frag_list (skb))
@@ -497,9 +501,12 @@ bool skb_recycle_check(struct sk_buff *skb, int skb_size)if (skb_shared (skb) ||skb_cloned (skb))return false; + shinfo =skb_shinfo (skb); +if (shinfo->destructor) +return false; +skb_release_head_state (skb); - shinfo =skb_shinfo (skb);memset (shinfo,0 , offsetof(struct skb_shared_info , dataref));atomic_set (&shinfo->dataref,1 );
@@ -799,7 +806,9 @@ int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail,memcpy ((struct skb_shared_info *)(data + size),skb_shinfo (skb), - offsetof(struct skb_shared_info, frags[skb_shinfo(skb)->nr_frags])); + offsetof(struct skb_shared_info , + frags[skb_shinfo(skb)->nr_frags])); +skb_shinfo (skb)->destructor =NULL ;/* Check if we can avoid taking references on fragments if we own * the last reference on skb->head. (see skb_release_data())
net/packet/af_packet.c +11 -13
@@ -825,19 +825,18 @@ ring_is_full:static void tpacket_destruct_skb (struct sk_buff *skb) { -struct packet_sock *po =pkt_sk (skb->sk); -void *ph; - -BUG_ON (skb ==NULL ); +struct packet_sock *po =pkt_sk (skb_shinfo (skb)->destructor_arg[0 ]);if (likely (po->tx_ring.pg_vec)) { - ph =skb_shinfo (skb)->destructor_arg; +void *ph =skb_shinfo (skb)->destructor_arg[1 ]; +BUG_ON (__packet_get_status (po, ph) !=TP_STATUS_SENDING );BUG_ON (atomic_read (&po->tx_ring.pending) ==0 );atomic_dec (&po->tx_ring.pending);__packet_set_status (po, ph,TP_STATUS_AVAILABLE ); } + skb->sk = &po->sk;sock_wfree (skb); }
@@ -862,7 +861,6 @@ static int tpacket_fill_skb(struct packet_sock *po, struct sk_buff *skb, skb->dev = dev; skb->priority = po->sk.sk_priority; skb->mark = po->sk.sk_mark; -skb_shinfo (skb)->destructor_arg = ph.raw;switch (po->tp_version) {case TPACKET_V2 :
@@ -884,9 +882,8 @@ static int tpacket_fill_skb(struct packet_sock *po, struct sk_buff *skb, to_write = tp_len;if (sock->type ==SOCK_DGRAM ) { - err =dev_hard_header (skb, dev,ntohs (proto), addr, -NULL , tp_len); -if (unlikely (err <0 )) +if (unlikely (dev_hard_header (skb, dev,ntohs (proto), addr, +NULL , tp_len) <0 ))return -EINVAL ; }else if (dev->hard_header_len) {/* net device doesn't like empty head */
@@ -897,8 +894,7 @@ static int tpacket_fill_skb(struct packet_sock *po, struct sk_buff *skb, }skb_push (skb, dev->hard_header_len); - err =skb_store_bits (skb,0 , data, - dev->hard_header_len); + err =skb_store_bits (skb,0 , data, dev->hard_header_len);if (unlikely (err))return err;
@@ -906,7 +902,6 @@ static int tpacket_fill_skb(struct packet_sock *po, struct sk_buff *skb, to_write -= dev->hard_header_len; } - err = -EFAULT ; page =virt_to_page (data); offset =offset_in_page (data); len_max =PAGE_SIZE - offset;
@@ -1028,7 +1023,10 @@ static int tpacket_snd(struct packet_sock *po, struct msghdr *msg) } } - skb->destructor = tpacket_destruct_skb; +skb_shinfo (skb)->destructor_arg[0 ] = &po->sk; +skb_shinfo (skb)->destructor_arg[1 ] = ph; + skb->destructor =NULL ; +skb_shinfo (skb)->destructor = tpacket_destruct_skb;__packet_set_status (po, ph,TP_STATUS_SENDING );atomic_inc (&po->tx_ring.pending);
net/core/skbuff.c int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, · line 814
Quoted 11 lines · Changli Gao
net/core/skbuff.c +3
-1@@ -799,7 +806,9 @@ int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail,memcpy ((struct skb_shared_info *)(data + size),skb_shinfo (skb), - offsetof(struct skb_shared_info, frags[skb_shinfo(skb)->nr_frags])); + offsetof(struct skb_shared_info , + frags[skb_shinfo(skb)->nr_frags])); +skb_shinfo (skb)->destructor =NULL ;/* Check if we can avoid taking references on fragments if we own * the last reference on skb->head. (see skb_release_data())
So it looks like pskb_expand_head will prevent the shinfo desctructor from being called, ever? If so, won't this break af_packet?
Signature · 1 line
MST
David Miller
maintainerFrom: "Michael S. Tsirkin" <[email protected]>
Date: Sun, 12 Sep 2010 14:13:49 +0200
Quoted 10 lines · Changli Gao
Quoted 9 lines
net/core/skbuff.c +3
-1@@ -799,7 +806,9 @@ int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail,memcpy ((struct skb_shared_info *)(data + size),skb_shinfo (skb), - offsetof(struct skb_shared_info, frags[skb_shinfo(skb)->nr_frags])); + offsetof(struct skb_shared_info , + frags[skb_shinfo(skb)->nr_frags])); +skb_shinfo (skb)->destructor =NULL ;
Quoted 7 lines
net/core/skbuff.c int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, · line 814
Quoted 2 lines
net/core/skbuff.c +0
-0/* Check if we can avoid taking references on fragments if we own * the last reference on skb->head. (see skb_release_data()) So it looks like
pskb_expand_headwill prevent the shinfo desctructor from being called, ever? If so, won't this breakaf_packet?From what I read, he is propagating it into the new SKB data blob
with expanded head area. It would get invoked when the skb's new data is put.
I am not sure this is correct, however.
Destructor register only cares about original data area, but what constitutes "original data" is ambiguous. In fact it seems impossible to catch the freeing of all parts properly.
When pskb_expand_head() is invoked we get new linear part, but non-linear part stays the same. However, entity which registered skb data destructor cares about old linear data lifetime, which we will no longer track after destructor is propagated only to the new shinfo.
So we need to do something different here. I bet original code overriding socket destructor semantics had a similar problem.
Changli, I have one other minor request, please name this something like "shinfo->data_destructor" and "shinfo->data_destructor_arg".
I think that will make it easier for other humans to understand :)
Thank you.
Quoted 38 lines · David Miller
From: "Michael S. Tsirkin" <[email protected]>
Date: Sun, 12 Sep 2010 14:13:49 +0200Quoted 10 lines · Changli Gao
Quoted 9 lines
net/core/skbuff.c +3
-1@@ -799,7 +806,9 @@ int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail,memcpy ((struct skb_shared_info *)(data + size),skb_shinfo (skb), - offsetof(struct skb_shared_info, frags[skb_shinfo(skb)->nr_frags])); + offsetof(struct skb_shared_info , + frags[skb_shinfo(skb)->nr_frags])); +skb_shinfo (skb)->destructor =NULL ;Quoted 7 lines
net/core/skbuff.c int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, · line 814
Quoted 2 lines
net/core/skbuff.c +0
-0/* Check if we can avoid taking references on fragments if we own * the last reference on skb->head. (see skb_release_data()) So it looks like
pskb_expand_headwill prevent the shinfo desctructor from being called, ever? If so, won't this breakaf_packet?From what I read, he is propagating it into the new SKB data blob
with expanded head area. It would get invoked when the skb's new data is put.
I am not sure this is correct, however.
Destructor register only cares about original data area, but what constitutes "original data" is ambiguous. In fact it seems impossible to catch the freeing of all parts properly.
When
pskb_expand_head()is invoked we get new linear part, but non-linear part stays the same. However, entity which registered skb data destructor cares about old linear data lifetime, which we will no longer track after destructor is propagated only to the new shinfo.So we need to do something different here. I bet original code overriding socket destructor semantics had a similar problem.
Quoted 6 lines
Changli, I have one other minor request, please name this something like "
shinfo->data_destructor" and "shinfo->data_destructor_arg".I think that will make it easier for other humans to understand :)
Thank you.
Hmm, and there's another issue I think I see here:destructor_arg now points to a socket.
What happens if the skb gets queued on an interface for a very long time (as can be the case with e.g. tap), and meanwhile
you try to kill the task that owns the socket, which
will try to destroy the socket?
Original code handles this by relevant devices orphaning an skb if it's queued indefinitely.
Signature · 1 line
MST
Quoted 7 lines · David Miller
Changli, I have one other minor request, please name this something like "
shinfo->data_destructor" and "shinfo->data_destructor_arg".I think that will make it easier for other humans to understand :)
OK. Thanks.
But there is another issue, when splice() is involved. If we splice the skbs generated by AF_PACKET socket to a pipe, the fragment pages will be hold by the pipe, but the skbs are freed, and AF_PACKET socket will been told that the corresponding TX ring buffers are available for the other uses wrongly.
Signature · 2 lines
Regards, Changli Gao([email protected])
Quoted 11 lines · Michael S. Tsirkin
Hmm, and there's another issue I think I see here:
destructor_argnow points to a socket.
What happens if the skb gets queued on an interface for a very long time (as can be the case with e.g. tap), and meanwhile
you try to kill the task that owns the socket, which
will try to destroy the socket?Original code handles this by relevant devices orphaning an skb if it's queued indefinitely.
I don't think the skb_orphan() there is used to destroy the socket in time, but notify the socket that skbs are sent out and it can send new skbs.
Signature · 2 lines
Regards, Changli Gao([email protected])
Quoted 9 lines · Changli Gao
Quoted 8 lines · Michael S. Tsirkin
Hmm, and there's another issue I think I see here:
destructor_argnow points to a socket.
What happens if the skb gets queued on an interface for a very long time (as can be the case with e.g. tap), and meanwhile
you try to kill the task that owns the socket, which
will try to destroy the socket?
Quoted 7 lines
Quoted 3 lines
Original code handles this by relevant devices orphaning an skb if it's queued indefinitely.
I don't think the
skb_orphan()there is used to destroy the socket in time, but notify the socket that skbs are sent out and it can send new skbs.
Well, the result is that we drop a socket reference from the skb, so it becomes possible to free the socket.
Quoted 3 lines
Signature · 2 lines
Regards, Changli Gao([email protected])