IRC channel logs


back to list of logs

<damo22>clarity_: "The whole issues with multi-core processors makes me think that IPC is better in the kernel instead of the microkernel" there is only one kernel, the microkernel
<damo22>mach is the microkernel, it provides memory management, facilitates IPC and some authentication i think
<damo22>the problem with Hurd in general is the lack of drivers for hardware
<damo22>rump is supposed to address that problem
<damo22>we're getting there, slowly
<damo22>mach had multiprocessor support for years but it was turned off
<clarity_>oh, interesting
<damo22>but there were missing bits for the x86 specific hardware
<damo22>to turn on SMP you need ACPI and APIC
<damo22>this stuff was only written for mach recently
<damo22>in fact you can compile mach with SMP turned on, but disk wont work yet
<damo22>we're hoping to upgrade rump to latest upstream code and see if it fixes the memory bloat
<damo22>so the disk can run without hogging >600MB ram
<clarity_>that's really neat
<clarity_>Yeah, I was almost thinking packaging a stripped down Linux w/ KVM for hurd to run on top of, to fix the hardware support issues, but ramp looks like a better solution
<damo22>its not about packaging, microkernel design means all the drivers run as separate processes in userspace
<damo22>the kernel has no drivers in it
<Pellescours>for rump we are so close
<Pellescours>thanks to damo :)
<clarity_>Also, is thread scheduler is outside of mach?
<clarity_>yeah, looks like thread scheduler is inside of the microkernel
<clarity_>I'm a little bit off today. I just left a job
<clarity_>Yeah, this make sense after looking at some diagrams
<clarity_>I need to do some reading, and then, I'll be able to ask better questions :-)
<clarity_>what I've gathered is rump is a generic interface for implementing drivers, and with that being supported in hurd, you can use all of the netbsd drivers ?
<mepy>hi (lol stux|away and ThinkT510)
***077AAB7C8 is now known as valeriusN