As Alexander points out, when we are trying to recycle a cloned/expanded
SKB we might trigger a race. The recycling code relies on the
pp_recycle bit to trigger, which we carry that over to cloned SKBs.
When that cloned SKB gets expanded, we are creating 2 separate instances
accessing the page frags. Since the skb_release_data() will first try to
recycle the frags, there's a potential race between the original and
cloned SKB.
Fix this by explicitly making the cloned/expanded SKB not recyclable.
If the original SKB is freed first the pages are released.
If it is released after the clone/expended skb then it can still be
recycled.
Fixes: 6a5bcd84e886 ("page_pool: Allow drivers to hint on SKB recycling")
Reported-by: Alexander Duyck <[email protected]>
Signed-off-by: Ilias Apalodimas <[email protected]>
---
net/core/skbuff.c | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index 12aabcda6db2..0cb53c05ed76 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -1718,6 +1718,13 @@ int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail,
}
off = (data + nhead) - skb->head;
+ /* If it's a cloned skb we expand with frags attached we must prohibit
+ * the recycling code from running, otherwise we might trigger a race
+ * while trying to recycle the fragments from the original and cloned
+ * skb
+ */
+ if (skb_cloned(skb))
+ skb->pp_recycle = 0;
skb->head = data;
skb->head_frag = 0;
skb->data += off;
--
2.32.0.rc0
+cc Alexander on his gmail address since the Intel one bounced.
Alexander want me to respin it with you gmail address on the Reported-by?
Sorry for the noise
/Ilias
On Thu, 8 Jul 2021 at 19:24, Ilias Apalodimas
<[email protected]> wrote:
>
> As Alexander points out, when we are trying to recycle a cloned/expanded
> SKB we might trigger a race. The recycling code relies on the
> pp_recycle bit to trigger, which we carry that over to cloned SKBs.
> When that cloned SKB gets expanded, we are creating 2 separate instances
> accessing the page frags. Since the skb_release_data() will first try to
> recycle the frags, there's a potential race between the original and
> cloned SKB.
>
> Fix this by explicitly making the cloned/expanded SKB not recyclable.
> If the original SKB is freed first the pages are released.
> If it is released after the clone/expended skb then it can still be
> recycled.
>
> Fixes: 6a5bcd84e886 ("page_pool: Allow drivers to hint on SKB recycling")
> Reported-by: Alexander Duyck <[email protected]>
> Signed-off-by: Ilias Apalodimas <[email protected]>
> ---
> net/core/skbuff.c | 7 +++++++
> 1 file changed, 7 insertions(+)
>
> diff --git a/net/core/skbuff.c b/net/core/skbuff.c
> index 12aabcda6db2..0cb53c05ed76 100644
> --- a/net/core/skbuff.c
> +++ b/net/core/skbuff.c
> @@ -1718,6 +1718,13 @@ int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail,
> }
> off = (data + nhead) - skb->head;
>
> + /* If it's a cloned skb we expand with frags attached we must prohibit
> + * the recycling code from running, otherwise we might trigger a race
> + * while trying to recycle the fragments from the original and cloned
> + * skb
> + */
> + if (skb_cloned(skb))
> + skb->pp_recycle = 0;
> skb->head = data;
> skb->head_frag = 0;
> skb->data += off;
> --
> 2.32.0.rc0
>
On Thu, Jul 8, 2021 at 9:31 AM Ilias Apalodimas
<[email protected]> wrote:
>
> +cc Alexander on his gmail address since the Intel one bounced.
>
> Alexander want me to respin it with you gmail address on the Reported-by?
>
> Sorry for the noise
> /Ilias
>
> On Thu, 8 Jul 2021 at 19:24, Ilias Apalodimas
> <[email protected]> wrote:
> >
> > As Alexander points out, when we are trying to recycle a cloned/expanded
> > SKB we might trigger a race. The recycling code relies on the
> > pp_recycle bit to trigger, which we carry that over to cloned SKBs.
> > When that cloned SKB gets expanded, we are creating 2 separate instances
> > accessing the page frags. Since the skb_release_data() will first try to
> > recycle the frags, there's a potential race between the original and
> > cloned SKB.
> >
> > Fix this by explicitly making the cloned/expanded SKB not recyclable.
> > If the original SKB is freed first the pages are released.
> > If it is released after the clone/expended skb then it can still be
> > recycled.
> >
> > Fixes: 6a5bcd84e886 ("page_pool: Allow drivers to hint on SKB recycling")
> > Reported-by: Alexander Duyck <[email protected]>
> > Signed-off-by: Ilias Apalodimas <[email protected]>
> > ---
> > net/core/skbuff.c | 7 +++++++
> > 1 file changed, 7 insertions(+)
> >
> > diff --git a/net/core/skbuff.c b/net/core/skbuff.c
> > index 12aabcda6db2..0cb53c05ed76 100644
> > --- a/net/core/skbuff.c
> > +++ b/net/core/skbuff.c
> > @@ -1718,6 +1718,13 @@ int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail,
> > }
> > off = (data + nhead) - skb->head;
> >
> > + /* If it's a cloned skb we expand with frags attached we must prohibit
> > + * the recycling code from running, otherwise we might trigger a race
> > + * while trying to recycle the fragments from the original and cloned
> > + * skb
> > + */
> > + if (skb_cloned(skb))
> > + skb->pp_recycle = 0;
> > skb->head = data;
> > skb->head_frag = 0;
> > skb->data += off;
Yeah, I would recommend a respin.
Also I would move this line up to the skb_cloned block just a few
lines before this spot just to avoid a second check.