On Sun, Oct 27, 2024 at 11:46 AM Zhaoming Luo <zhaoming1...@qq.com> wrote: > Hi,
Hi Zhaoming, > I did the following things in a Debian/Hurd virtual machine. > > I was following this link > <https://darnassus.sceen.net/~hurd-web/hurd/translator/writing/example/> > to implement a rtc server. My idea is that I should implement a working > rtc server in user space before adding the source code into the repo > because I don't want to get stuck in Makefile system before I have a > working rtc server. Sure, that makes sense. And it's a great property of the Hurd that you can in fact develop out-of-tree translators with ease. Though in the case of rtc, we'd probably want it in-tree eventually. > I wrote the pioctl.defs (I paste it below), used mig > to resolve it, and mig told me 'tm_t' is not defined. Then I added the > following lines in /usr/include/i386-gnu/hurd/ioctl_types.h If it's MIG itself that complained about a missing type, adding a definition to a C header won't change anything. > It still didn't work. How can I use `struct tm` in a user space defs file? To use tm_t, it would have to be defined in some MIG .defs header (in a way that matches the C definition). See how timespec_t is defined in hurd_types.defs for example. But also, tm_t seems to be a weird type to be used in RPCs. I see that Linux RTC device ioctls actually operate on a 'struct rtc_time', which is only similar to a 'struct tm', but not the same thing. I would presume the RTC actually deals with broken-down tm-like time representation internally, rather than a timestamp? If so, it would make sense to me if you defined a type that represents RTC's time representation right in your .defs file, and used that in your RPCs. Perhaps like this: type rtc_time_t = struct { int tm_sec; ... }; with a matching definition in C. > /* TODO: adding other ioctl calls for /dev/rtc in mc146818rtc.h */ > skip; skip; skip; skip; /* 0 1 2 3 */ > skip; skip; skip; skip; /* 4 5 6 7 */ > skip; /* 8 */ > > /* 9 RTC_RD_TIME */ > routine pioctl_rtc_rd_time ( > reqport: io_t; > inout time: tm_t); Shouldn't this be 'out' rather than 'inout'? > /* 10 RTC_SET_TIME */ > routine pioctl_rtc_set_time ( > reqport: io_t; > time: tm_t); Do I understand it right that this would be more privileged than reading the time? Sergey
