IRC channel logs

<damo22>i did not know apic works with NCPUS=1

<damo22>i thought it needs to enumerate the apics using acpi functions

<damo22>isnt that tied to smp?

<solid_black>anyone has any idea why libpager tries so hard to process incoming requests in the same order they were received, and not concurrently?

<solid_black>AFAICS the mechanism for this has been present since "initial revision" in 1994

<solid_black>guess I should just patch it out and see what breaks

<azert>solid_black: there was a dude at one point that analyzed all the shortcomings of the libpager in the Hurd, and went as far as to program a new one that is in this out of tree repo https://github.com/BrentBaccala/hurd

<solid_black>azert: I've seen that, yeah

<azert>You might be interested in checking that if you are improving the current lib-pager

<solid_black>indeed, libpager is the right layer of the stack to handle multiple clients (Mach kernels) at the same time, without making the hosting ext2fs (or whatever) care

<solid_black>it might be just that it wants to receive m_o_lock_completed() after all the m_o_data_return() calls

<solid_black>which makes perfect sense

<solid_black>but that's a sad reason to limit concurrency in the typical case when you're not locking anything, and the kernel is actively doing m_o_data_request() and sometimes m_o_data_return()

<AlmuHS>damo22.

<AlmuHS>damo22: in a patch apic was allowed to enable without smp

<AlmuHS>he told that it could be useful for graphics or other peripheral

<solid_black>aarch64 exception -> Mach exception -> Unix signal -> trampoline -> handler -> sigreturn works all the way

IRC channel logs

2024-03-28.log