[<prev] [next>] [day] [month] [year] [list]
Message-ID: <CAM0EoM=UPTp2MKZbGk0U0rHb6srfVQ9djDOAPGtXj=Papw=TPA@mail.gmail.com>
Date: Thu, 19 Oct 2023 15:32:02 -0400
From: Jamal Hadi Salim <jhs@...atatu.com>
To: Daniel Borkmann <daniel@...earbox.net>
Cc: bpf <bpf@...r.kernel.org>,
Linux Kernel Network Developers <netdev@...r.kernel.org>, Jakub Kicinski <kuba@...nel.org>,
Florian Westphal <fw@...len.de>
Subject: challenges with DEBUG_INFO_BTF
Since patchworks does make allmodconfig it caught a small challenge
with our setup.
The makefile looks as follows:
obj-y := p4tc_types.o p4tc_tmpl_api.o p4tc_pipeline.o \
p4tc_action.o p4tc_table.o p4tc_tbl_entry.o \
p4tc_runtime_api.o
obj-$(CONFIG_DEBUG_INFO_BTF) += p4tc_bpf.o
This compiles fine with the exception the bpf program that references
the kfuncs will fail to load. Bad for usability of course.
make allmodconfig does not enable CONFIG_DEBUG_INFO_BTF which means
CONFIG_DEBUG_INFO_BTF_MODULES is not turned on given dependency;
This leaves us with two options:
1) Issue a warning, as such
+ifndef CONFIG_DEBUG_INFO_BTF
+$(warning WARNING: DEBUG_INFO_BTF is off. No P4TC kfuncs will be compiled.)
+endif
At least the user will get a compile warning..
2) so another approach is we can force enable CONFIG_DEBUG_INFO_BTF.
Sadly this would require many other options in addition be to force
enabled such as CONFIG_DEBUG_INFO, etc (which is complex).
My current thinking is to leave it as #1 or the variant we have right
now (without the warning).
Thoughts?
+cc Florian (since netfilter may have had similar challenges)
cheers,
jamal
Powered by blists - more mailing lists