IRC channel logs

2021-10-24.log

back to list of logs

***Forza_ is now known as Forza
<civodul>o/
<civodul>by stopping/starting pfinet, i seem to easily trigger a kernel page fault and get thrown at the kdb prompt
<civodul>that's with rtl8139 NIC emulation and the corresponding in-kernel driver
<civodul>does that ring a bell?
<jpoiret>hello everyone! it's my first time reading hurd source (and also haven't tried it yet but I kind of understand the basic mechanisms). I want to implement swap support in Guix, so I was looking at sutils/swapon.c, but I can't find the included header sys/mmap.h (which I thought must contain the default_pager_paging_storage declaration)
<civodul>jpoiret: hi! <sys/mman.h> is in glibc
<youpi>jpoiret: default_pager_paging_storage is in default_pager_U.h, generated during the hurd build
<youpi>./sutils/default_pager_U.h:kern_return_t default_pager_paging_storage
<youpi>that's built through the default_pagerUser.o dependency
<jpoiret>oh alright, I guess this is built by MIG and that mach-defpager/default_pager.c implements the server part, right?
<youpi>yes
<jpoiret>I see that swapon.c supports the old rpc call to set paging, because you're supposed to be able to use a newer version of swapon with an older Hurd? Trying to determine if we should try both RPCs on GNU/Hurd Guix
<jpoiret>i'll look if the one we package has that protocol enabled, if so we'll just ignore the old one i guess
<youpi>I don't think you actually need to care?
<youpi>I mean, both are passed the same way, so from a packaging point of view that shouldn't matter
<jpoiret>oh but I mean, I want to implement swap support directly inside guix rather than relying on the swapon utility (just like we use the syscall ourselves for Linux rather than use util-linux's swapon)
<youpi>ah ok
<youpi>I'd say you can probably assume the newer protocol
<youpi>mach-defpager has it since 2001