On 9/15/25 12:14, Breno Leitao wrote:
On Sun, Sep 14, 2025 at 06:05:17AM +0800, kernel test robot wrote:
tree/branch:
https://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue.git dev-queue
branch HEAD: a6c7254e995a80be8c3239fb631d9d65c9e75248 idpf: enable XSk
features and ndo_xsk_wakeup
Error/Warning ids grouped by kconfigs:
recent_errors
`-- loongarch-loongson3_defconfig
|-- ld.lld:error:undefined-symbol:libie_fwlog_deinit
|-- ld.lld:error:undefined-symbol:libie_fwlog_init
`-- ld.lld:error:undefined-symbol:libie_get_fwlog_data
I am having a similar issue on net-next, where the kernel is failing to
compile due to the undefined symbols above.
This happens when CONFIG_ICE is set.
thank you for a human report!
I were analyzing this in the past, (after all fixes by Michal), and
code looked fine, despite lkp complains.
Now, when an actual problem reproduces, I tried it too, and indeed there
was problem - but, it was not a full build (I made a full build just
before fwlog PR, then just added the patches)
during my attempts at understanding what happens, I have fixed the
problem on my machine, so not sure what was needed.
I have moved the below line prior to #includes in fwlog.c:
#define DEFAULT_SYMBOL_NAMESPACE "LIBIE_FWLOG"
that itself didn't helped, and with that reverted the fwlog.o an .ko
files are still building
I have invoked
make nsdeps # in the root of the tree, interrupted in a few minutes
then later:
make M=drivers/net/ethernet/intel/libie -j nsdeps
what got my attention as the very first time that I got fwlog.o
sorry, I have no time for proper digging if there is some incremental
build problem (which I have no idea to classify as a "wrong build-user
operations" vs "buildsystem issue"), at least today
