05:57 <arnd> I think I have fully caught up on the pull requests now, with only a couple of minor changes I asked for. Five new SoC families this time, which is huge after we didn't really have any new ones for a couple of releases

05:58 <arnd> I have queued pxa1908 (not really a new family but still put it into the same branch) andes/QiLai, Cix/P1, Axiado AX3000, and SophGo SG2000 (arm64). The only one I see that didn't make is the Black Sesame chip that still needs more changes

06:11 <sven> weird, I can see that you pulled mine but I didn't get the usual patchwork emails this time. any idea why?

12:09 <arnd> sven: I never figured it out, it works for most people, but there are a few senders that always get ignored by patchwork. So these don't get an email and don't get the status updated

12:10 <arnd> Usually it's tagr, shawnguo and mmind00 that don't get the automatic updates

12:11 <sven> guess I’m part of that list now as well. the two things I changed are that I moved both my email and the tree to kernel.org

12:12 <mmind00> yep ... things just stopped for me, without any changes in my setup ... I guess one would need to look into the mailserver logs of the patchwork server to see what actually happens

12:39 <bjdooks> does anyone actually use mmc-spi any more?

12:42 <a3f> bjdooks, last time I did was for the HiFive RISC-V boards in QEMU

12:42 <bjdooks> yeah, exactly the fun i'm currently debugging

12:43 <broonie> I think one of the boards in my lab has it wired up.

12:43 <bjdooks> i think it is rare, which is probabyl why qemu's had bugs in it since v9.x

13:39 <bjdooks> and found the second bug

15:55 <milkylainen> anyone here familiar with net/dsa and dsa switches?

15:58 <cambrian_invader> milkylainen: what's your question

16:03 <milkylainen> cambrian_invader: Ack. If I have a DSA cascade, and try to delete a mac address down the cascade. A mac generated by the switchcore driver because of dynamic learning disabled (cpu based learning or whatever naming you prefer). The delete seems to terminate in the middle of the cascade because the current code expects to find the mac address on the DSA ports, which it does not.

16:04 <milkylainen> So if I just stop the termination, the event continues with more ports and eventually finds the switchcore blackhole on a user port down in the cascade.

16:06 <milkylainen> The blackhole needs to go, so I can get a new event. Sort of a restart of events to drive port-access thingies like whitelists, MAB, dot1x etc.

16:08 <milkylainen> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/net/dsa/switch.c#n329

16:09 <milkylainen> called from https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/net/dsa/switch.c#n471

16:10 <milkylainen> So the loop doesn't find the nodes downstream because the delete stops on a DSA port, expecting it to have this mac.

16:15 <milkylainen> I'll be back in a few.

16:38 <cambrian_invader> unfortunately, this exceeds my DSA knowledge

16:39 <bjdooks> right, qemu fixes for the mmc-spi sent

18:26 <milkylainen> cambrian_invader: It's a difficult question. Thanks anyway.