On Wed, Jun 22, 2022 at 12:06 AM Roberto Sassu roberto.sassu@huawei.com wrote:
From: Alexei Starovoitov [mailto:alexei.starovoitov@gmail.com] Sent: Wednesday, June 22, 2022 12:32 AM On Tue, Jun 21, 2022 at 06:37:57PM +0200, Roberto Sassu wrote:
- if (child_pid == 0) {
snprintf(path, sizeof(path), "%s/signing_key.pem", tmp_dir);return execlp("./sign-file", "./sign-file", "-d", "sha256",path, path, data_template, NULL);Did you miss my earlier reply requesting not to do this module_signature append and use signature directly?
I didn't miss. sign-file is producing the raw PKCS#7 signature here (-d).
I'm doing something slightly different, to test the keyring ID part. I'm retrieving an existing kernel module (actually this does not work in the CI), parsing it to extract the raw signature, and passing it to the eBPF program for verification.
We don't have signed modules in CI. When you make changes like this you have to explain that in the commit log.
Since the kernel module is signed with a key in the built-in keyring, passing 1 or 0 as ID should work.
Roberto
(sorry, I have to keep the email signature by German law)
I don't believe that's the case since plenty of people work from Germany and regularly contribute patches without such banners.
HUAWEI TECHNOLOGIES Duesseldorf GmbH, HRB 56063 Managing Director: Li Peng, Yang Xi, Li He