[Buildroot] [PATCH] configs/qemu_xtensa_lx60_nommu: use busybox minimal config
Arnout Vandecappelle
arnout at mind.be
Thu Apr 7 16:39:53 UTC 2022
On 06/04/2022 22:00, Vincent Stehlé wrote:
> On Mon, Apr 04, 2022 at 10:28:12PM +0200, Arnout Vandecappelle wrote:
> ..
>> But that actually causes a problem: if udhcpc doesn't keep running the
>> background any more, it's not going to refresh its lease, but will keep on
>> using its IP address indefinitely. Although this seems to work at first
>> hand, it's going to cause some problem somewhere down the line...
>
> Hi Arnout,
>
> After your comment at first I thought all the no-MMU buildroot defconfigs were
> subtly broken :-S
>
> I did some investigations and now my understanding is that udhcpc does indeed
> daemonize by default in the no-MMU case. See [1]:
>
> #if !BB_MMU
> /* on NOMMU reexec (i.e., background) early */
> if (!(opt & OPT_f)) {
> bb_daemonize_or_rexec(0 /* flags */, argv);
> logmode = LOGMODE_NONE;
> }
> #endif
>
> That behaviour was added in busybox udhcpc while removing the -b option (in
> commit [2]).
>
> Also, this is confirmed on qemu where we can see the udhcpc process running in
> the background:
>
> ~ # ps
> PID USER VSZ STAT COMMAND
> 1 root 616 S init
> ...
> 62 root 612 S udhcpc -R -p /var/run/udhcpc.eth0.pid -i eth0 -x hos
>
>> So we should find a different solution, and we should probably use that in
>> the MMU case as well (just to keep things consistent.
>
> Now I think that we are safe. What is your opinion on this, please?
I'll add a summary of this entire explanation and commit it.
Regards,
Arnout
>
> Best regards,
> Vincent.
>
> [1]: https://git.busybox.net/busybox/tree/networking/udhcp/dhcpc.c#n1352
> [2]: https://git.busybox.net/busybox/commit/?id=21765fa063830923d13426ec6989c16da9210e49
More information about the buildroot
mailing list