Closed brycekahle closed 9 months ago
Updated
Wouldn't it make sense to look into passing all BTF objects (vmlinux and modules) in one step, and generate a single minimised BTF object from that? Or does it defeat what you're trying to achieve? (Sorry for failing to raise this on the previous review.)
There may be multiple kernel modules in play for a single eBPF object file, since it can contain multiple programs and they each may probe different modules. Since a kernel module BTF has conflicting type IDs (they all start at max(vmlinux) + 1) with other kernel module BTF, I'm not sure there is a way with libbpf to load them all at the same time.
Updated
Welp, first ML attempt was a failure. Gmail borked the formatting. I'm looking into my options, but my work email is pretty locked down so I cannot even get an app password.
Yeah GMail interface is known to break patch formatting. I'm using git send-email
, but yes, it needs an app password to use the SMTP.
Have you looked into using a local client that supports 0Auth2 to connect to GMail (I think Thunderbird at least can)? I don't think you'd need an app password to do that?
If you don't find a solution for your work email, you can always use a personal email to send your patch. git send-email
should preserve the From:
header and the Signed-off-by
, so only your work email would appear in the git logs (your personal email would be on https://lore.kernel.org/bpf/ though).
If none of these work or is convenient, I'm happy to post the patch on your behalf.
Closing because it is on ML now
Updated, let me know if it looks good for me to submit to the mailing list.