Skip to contentSkip to conversation

net: Fix potential deadloop in skb_copy_ubufs()

PATCH linmiaohe 2 participants 2 messages Aug 14, 2020

lore
Share & tools
lore

linmiaohe

author
#0 lore raw
Quoted 17 lines · Willem de Bruijn
Quoted 6 lines · linmiaohe

From: Miaohe Lin <[email protected]>

We could be trapped in deadloop when we try to copy userspace skb frags buffers to kernel with a cloned skb:
Reproduce code snippet:

Quoted 7 lines
        skb = alloc_skb(UBUF_DATA_LEN, GFP_ATOMIC);
        clone = skb_clone(skb, GFP_ATOMIC);
        skb_zcopy_set_nouarg(clone, NULL);
        pskb_expand_head(skb, 0, 0, GFP_ATOMIC);

Catch this unexpected case and return -EINVAL in skb_orphan_frags() before we call skb_copy_ubufs() to fix it.

Is this a hypothetical codepath?

Quoted 8 lines

skb zerocopy carefully tracks clone calls where necessary. See the call to skb_orphan_frags in skb_clone, and the implementation of that callee.

The only caller of skb zerocopy with nouarg is tpacket_fill_skb, as of commit 5cd8d46ea156 ("packet: copy user buffers before orphan or clone").

As the commit subject indicates, this sets skb_zcopy_set_nouarg exactly to be sure that any clone will trigger a copy of "zerocopy" user data to private kernel memory.

No clone must happen between alloc_skb and skb_zcopy_set_nouarg, indeed. But AFAIK, none exists.

Since we always call skb_orphan_frags in skb_clone, is it unnecessary to call skb_orphan_frags in pskb_expand_head when skb is cloned ?

Thanks.

#1 lore raw
Quoted 21 lines · linmiaohe
Quoted 19 lines · Willem de Bruijn
Quoted 6 lines · linmiaohe

From: Miaohe Lin <[email protected]>

We could be trapped in deadloop when we try to copy userspace skb frags buffers to kernel with a cloned skb:
Reproduce code snippet:

Quoted 7 lines
        skb = alloc_skb(UBUF_DATA_LEN, GFP_ATOMIC);
        clone = skb_clone(skb, GFP_ATOMIC);
        skb_zcopy_set_nouarg(clone, NULL);
        pskb_expand_head(skb, 0, 0, GFP_ATOMIC);

Catch this unexpected case and return -EINVAL in skb_orphan_frags() before we call skb_copy_ubufs() to fix it.

Is this a hypothetical codepath?

skb zerocopy carefully tracks clone calls where necessary. See the call to skb_orphan_frags in skb_clone, and the implementation of that callee.

Quoted 8 lines
Quoted 6 lines

The only caller of skb zerocopy with nouarg is tpacket_fill_skb, as of commit 5cd8d46ea156 ("packet: copy user buffers before orphan or clone").

As the commit subject indicates, this sets skb_zcopy_set_nouarg exactly to be sure that any clone will trigger a copy of "zerocopy" user data to private kernel memory.

No clone must happen between alloc_skb and skb_zcopy_set_nouarg, indeed. But AFAIK, none exists.

Since we always call skb_orphan_frags in skb_clone, is it unnecessary to call skb_orphan_frags in pskb_expand_head when skb is cloned ?

Please give us a real case.

I fear that your patches are coming directly from some kind of automated tool, that really misses how the code is really used from _current_ code base, not _hypothetical_ one.

This is very time consuming. Please provide evidence first.

Thank you.