Re: [PATCH net 1/2] net: tls: explicitly disallow disconnect

[Paolo Abeni <pabeni@redhat.com>]

On 4/15/25 5:16 AM, Ihor Solodrai wrote:
> On 4/4/25 11:03 AM, Jakub Kicinski wrote:
>> syzbot discovered that it can disconnect a TLS socket and then
>> run into all sort of unexpected corner cases. I have a vague
>> recollection of Eric pointing this out to us a long time ago.
>> Supporting disconnect is really hard, for one thing if offload
>> is enabled we'd need to wait for all packets to be _acked_.
>> Disconnect is not commonly used, disallow it.
>>
>> The immediate problem syzbot run into is the warning in the strp,
>> but that's just the easiest bug to trigger:
>>
>>   WARNING: CPU: 0 PID: 5834 at net/tls/tls_strp.c:486 tls_strp_msg_load+0x72e/0xa80 net/tls/tls_strp.c:486
>>   RIP: 0010:tls_strp_msg_load+0x72e/0xa80 net/tls/tls_strp.c:486
>>   Call Trace:
>>    <TASK>
>>    tls_rx_rec_wait+0x280/0xa60 net/tls/tls_sw.c:1363
>>    tls_sw_recvmsg+0x85c/0x1c30 net/tls/tls_sw.c:2043
>>    inet6_recvmsg+0x2c9/0x730 net/ipv6/af_inet6.c:678
>>    sock_recvmsg_nosec net/socket.c:1023 [inline]
>>    sock_recvmsg+0x109/0x280 net/socket.c:1045
>>    __sys_recvfrom+0x202/0x380 net/socket.c:2237
>>
>> Fixes: 3c4d7559159b ("tls: kernel TLS support")
>> Reported-by: syzbot+b4cd76826045a1eb93c1@syzkaller.appspotmail.com
>> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
> 
> Hi everyone.
> 
> This patch has broken a BPF selftest and as a result BPF CI:
> * https://github.com/kernel-patches/bpf/actions/runs/14458537639
> * https://github.com/kernel-patches/bpf/actions/runs/14457178732
> 
> The test in question is test_sockmap_ktls_disconnect_after_delete
> (tools/testing/selftests/bpf/prog_tests/sockmap_ktls.c) [1].
> 
> Since the test is about disconnect use-case, and the patch disallows
> it, I assume it's appropriate to simply remove the test?

Ideally, yes. disconnect() implementation by its own nature error and
race prone, I guess  TLS adds some more spice to it. Unless there is a
real end-user scenario behind it, removing the disconnect()
implementation is by far the best option.

Still the test presence hints at some possible use-case[???]. Was it
created using the plain tcp test cases as a template?

Thanks,

Paolo