On 4/9/25 2:17 AM, Geliang Tang wrote:
Hi Yonghong,
On Sun, 2024-08-25 at 21:05 -0700, Yonghong Song wrote:
On 8/25/24 8:29 PM, Gang Yan wrote:
Hi Alexei:
It's my honor to recieve your reply. The response to your concerns
is attached below
for your review.
On Mon, Aug 26, 2024 at 10:57:12AM +0800, Gang Yan wrote:
On Thu, Aug 22, 2024 at 8:33 AM Jakub Kicinski wrote:
On Thu, 22 Aug 2024 14:08:57 +0800 Gang Yan wrote:
diff --git a/net/socket.c b/net/socket.c
index fcbdd5bc47ac..63ce1caf75eb 100644
--- a/net/socket.c
+++ b/net/socket.c
@@ -1695,6 +1695,7 @@ __weak noinline int
update_socket_protocol(int family, int type, int protocol)
{
return protocol;
}
+ALLOW_ERROR_INJECTION(update_socket_protocol, ERRNO);
IDK if this falls under BPF or directly net, but could you
explain
what test will use this? I'd prefer not to add test hooks into
the
kernel unless they are exercised by in-tree tests.
This looks unnecessary.
update_socket_protocol is already registered as fmodret.
There is even selftest that excises this feature:
tools/testing/selftests/bpf/progs/mptcpify.c
It doesn't need to be part of the error-inject.
The 'update_socket_protocol' is a BPF interface designed primarily
to
fix the socket protocol from TCP protocol to MPTCP protocol without
requiring modifications to user-space application codes. However,
when attempting to achieve this using the BCC tool in user-space,
the BCC tool doesn't support 'fmod_ret'. Therefore, this patch aims
to
further expand capabilities, enabling the 'kprobe' method can
overriding
the update_socket_protocol interface.
Gang Yan, could you explore to add fmod_ret support in bcc? It should
be
similar to kfunc/kretfunc support. I am happy to review your patches.
It took us some time to add this support in bcc, and we have recently
completed it in [1]. We would be grateful if you could help review
these patches.
Sounds good. I will take a look.
Thanks,
-Geliang
[1]
https://github.com/iovisor/bcc/pull/5274
Thanks,
Yonghong
As a Python developer, the BCC tool is a commonly utilized
instrument for
interacting with the kernel. If the kernel could permit the use of
an
error-inject method to modify the `update_socket_protocol`, it
would significantly
benefit the subsequent promotion and development of MPTCP
applications.
Thank you for considering this enhancement.
Best wishes!
Gang Yan
