[PATCH net v2 0/2] bareudp: fix several issues

[PATCH net v2 0/2] bareudp: fix several issues

[Taehee Yoo]

This patchset is to fix problems when bareudp is used nestedly.

1. If the NETIF_F_LLTX flag is not set, the lockdep warns about
a possible deadlock scenario when bareudp interfaces are used nestedly.
But, like other tunneling interfaces, bareudp doesn't need xmit lock.
So, it sets NETIF_F_LLTTX.
Lockdep no more warns about a possible deadlock scenario.

2. bareudp interface calculates min_headroom when it sends a packet.
When it calculates min_headroom, it uses the size of struct iphdr even
the ipv6 packet is going to be sent.
It would result in a lack of headroom so eventually occurs kernel panic.

Changelog
v1 -> v2:
 - Change reproducer script in 1/2 patch.
 - Fix reproducer script in 2/2 patch.

Taehee Yoo (2):
  bareudp: set NETIF_F_LLTX flag
  bareudp: Fix use of incorrect min_headroom size

 drivers/net/bareudp.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

-- 
2.17.1

Re: [PATCH net v2 0/2] bareudp: fix several issues

[Jakub Kicinski]

On Mon, 28 Dec 2020 15:21:21 +0000 Taehee Yoo wrote:
> This patchset is to fix problems when bareudp is used nestedly.
> 
> 1. If the NETIF_F_LLTX flag is not set, the lockdep warns about
> a possible deadlock scenario when bareudp interfaces are used nestedly.
> But, like other tunneling interfaces, bareudp doesn't need xmit lock.
> So, it sets NETIF_F_LLTTX.
> Lockdep no more warns about a possible deadlock scenario.
> 
> 2. bareudp interface calculates min_headroom when it sends a packet.
> When it calculates min_headroom, it uses the size of struct iphdr even
> the ipv6 packet is going to be sent.
> It would result in a lack of headroom so eventually occurs kernel panic.

Applied, thanks!