IRC channel logs

<almuhs>finally, i confirmed that the new smp kernel is very unstable. Now the system keeps freezed executing "update-grub". And only boots with some little combinations of -smp

<almuhs>**some few

<damo22>almuhs: smp was always unstable, but we are slowly getting rid of bugs that we know of

<damo22>netdde is known not to work with smp

<damo22>again, that is why i put so much effort on rumpnet

<almuhs>yes, rumpnet will solve the booting freeze. But the "update-grub" freeze is a different problem: probably from rumpdisk

<damo22>almuhs: its possible your qemu is restricted from writing to block 0

<damo22>check if you get WARNING: Image format was not specified for '/dev/sdd' and probing guessed raw.

<damo22> Automatically detecting the format is dangerous for raw images, write operations on block 0 will be restricted.

<damo22> Specify the 'raw' format explicitly to remove the restrictions.

<almuhs>with a single cpu i have no problem to make the "update-grub"

<almuhs>even with apic

<almuhs>i use a raw image

<damo22>ok

<damo22>you could run gdb on the outside of qemu and connect target remote :1234

<damo22>then check backtrace when the hang happens

<almuhs>yes, but i only knows how to connect to gnumach

<damo22>i assume gnumach is at fault, not update-grub

<damo22>(gdb) n

<damo22>336 simple_unlock_irq(s, &ioapic_lock);

<damo22>(gdb) n

<damo22>339 lapic_eoi ();

<damo22>(gdb) n

<damo22>interrupt () at ../i386/i386at/interrupt.S:105

<damo22>105 movl S_IPL,%eax

<damo22>not sure how that is possible if the IF is still clear

<youpi>again, I wouldn't be surprised that gdb+qemu support doesn't handle IF that well

<youpi>(while single-stepping)

<damo22>i think theres nothing wrong with qemus emulation

<damo22>i think the irq line is getting stuck on

<damo22>and due to it being a level triggered interrupt it keeps firing

<damo22>145 #endif /* (MACH_KDB || MACH_TTD) && !MACH_XEN */

<damo22>146 testl %edx,%edx /* spl0? */

<damo22>why are we testing edx against edx

<damo22>ok its testing for edx == 0

<gnucode>Did ya'll get an email about a chinese microkernel today ?

<gfleury>gnucode: yes i get it

<gnucode> do you think it's worth anything or just spam ?

<p4r4D0xum>gfleury: What's the name of it?

<gfleury>gnucode: i don't think it is a spam

<p4r4D0xum>gnucode: everything is possible, if CCP has your back, like with LoongArch

<gfleury>p4r4D0xum: they talk about HongMeng Kernel which is close source from Huawei but it describes technique that they use succefully in seL499,99. He gives also academic ressources from their research in microkernel

<youpi>everything is possible, it's a question of spending time on it :)

<youpi>and one has to put priorities on stuff

<p4r4D0xum>gfleury: Thank you.

IRC channel logs

2025-07-14.log