05:17 <HdkR> So I found a curious edge case in memory mapping and using the GROW feature. If the VMA grows, how is the userspace supposed to get notified about the new range to safely munmap that vma? As far as I can tell, the only safe thing to do is parse `/proc/self/maps`, find the intersecting range, and free what was found.

12:43 <kos_tom> headless: unfortunately, I no longer have this SH7786 hardware :/

12:46 <headless> kos_tom: that's a pity? what board did you have?

12:46 <headless> s/?/.../

12:59 <broonie> ukleinek: The issue was with applying your patch on the base commit you'd advertised, there's an issue with cs35l41_hda_spi.c on that commit.

13:24 <ukleinek> broonie: is this an apply or a build issue?

13:25 <broonie> apply

13:25 <arnd> headless: did you see my email reply? I think the IRQ0 ethernet on that board has been broken for 13 years, so I would stress out about breaking it more

13:25 <arnd> s/would/would not/

13:26 <ukleinek> broonie: hmm, cannot reproduce your problem, looks all fine on my end. https://paste.debian.net/1230122/

13:28 <broonie> well, it's applied now so I'm not going to spend much more time trying to figure it out.

13:32 <headless> arnd: I've read it but haven't had the time to check the cloiks yet

13:33 <headless> *s/cloicks/links/ 8-)

13:34 <headless> arnd: thank for you comments

15:11 <ukleinek> broonie: \o/

15:20 <ukleinek> broonie: FTR: I double checked and the patches look fine

16:23 <headless> arnd: so my old manual came handy at last :-)

16:24 <headless> the kernel/cpu/ code is really convoluted...

16:24 <headless> (arch/sh/ I mean)

19:12 <headless> back again

19:12 <headless> kos_tom: have you replied, by chance?

19:13 <headless> arnd: I'm thinking of a stupid hack: don't subtract 0x10 from EVT after shift

19:13 <headless> makes hwirqs mapped at 0x10, just like good old x86

19:13 <headless> :-)

19:16 <headless> well, actually x86/mips/ppc had the 1st 16 irqs from 8259 mapped at 0

19:17 <arnd> headless: arm also has special NR_IRQS_LEGACY=16, with normal interrupt controllers allocating their numbers above that

19:18 <arnd> I like the idea, that should totally work for arch/sh

19:18 <headless> arnd: thanx :-)

19:21 <arnd> of course getting anything into arch/sh/ may be a challenge, with ysato being completely out of the loop and dalias rarely being active. Volunteering geertu could work, but he's already stuck with maintaining arch/m68k ;-)

19:33 <geertu> arnd: I'm so lucky the other m68k people are more helpful than the sh maintainers ;-)

20:07 <kos_tom> headless: I had two boards

20:08 <kos_tom> the AP-SH4-AD and a customer-specific board

20:13 <headless> kos_tom: ah, the AP is what's going to be broken...

20:13 <headless> kos_tom: have you made sh_eth work there by chance?

21:05 <kos_tom> headless: yes, Ethernet was working IIRC

21:05 <kos_tom> headless: but I had some patches

21:05 <kos_tom> which I sent upstream, but I'm not sure they were merged

21:07 <headless> kos_tom: you mean sh_eth patches?

21:08 <headless> I have v2 of your patch with my modifications which I offered you to test -- but you told me it doesn't quite work

21:09 <headless> that's my recollection :-)

21:09 <headless> if you have s/th working please send it, we'll merge it at last...

21:52 <kos_tom> headless: I no longer have HW

21:52 <kos_tom> headless: so I can't test anything.

21:56 <headless> kos_tom: OK, got it

21:58 <headless> theoretically, I can merge what I have...

21:59 <headless> though nobody cares anylonger prolly

22:01 <headless> g'ntie