/g/ - Thread 108270573

Anonymous
syscall of the day: msg(get|snd|rcv|ctl) 03/01/26(Sun)15:47:37 No.108270573

syscall of the day: msg(get|snd|rcv|ctl) Anonymous 03/01/26(Sun)15:47:37 No.108270573 [Reply]▶

File: Tumblr_l_215661021174968.png (1.2 MB)

previous: >>108262843
#define __NR_msgget                68
#define __NR_msgsnd                69
#define __NR_msgrcv                70
#define __NR_msgctl                71
https://man7.org/linux/man-pages/man2/msgget.2.html
https://man7.org/linux/man-pages/man2/msgop.2.html
https://man7.org/linux/man-pages/man2/msgctl.2.html

tl;dr:
message queues, but worse!
https://man7.org/linux/man-pages/man7/mq_overview.7.html
to be clear, these syscalls are NOT posix message queues. but the above link mentions them in the notes, and it should explain a bit more

one of about infinity different implementations of IPC. from the before times...
what IPC do you use? do you have a favorite subset of IPC? personally, i'm quite fond of RPC.

relevant resources:
man man
man syscalls
https://man7.org/linux/man-pages/
https://linux.die.net/man/
https://elixir.bootlin.com/linux/
https://elixir.bootlin.com/musl/
https://elixir.bootlin.com/glibc/

16 RepliesView Thread

Showing all 16 replies.

Anonymous
03/01/26(Sun)16:12:36 No.108270723

Anonymous 03/01/26(Sun)16:12:36 No.108270723▶

I love message queues, the only drawback is that eventually in the development lifecycle you notice that you actually sometimes need two way communication and then you are kind of fucked and have to bolt on something else anyways

Anonymous
03/01/26(Sun)16:47:46 No.108270928

Anonymous 03/01/26(Sun)16:47:46 No.108270928▶

What's the point of these when you can use a socket instead?

Anonymous
03/01/26(Sun)16:59:31 No.108270987

Anonymous 03/01/26(Sun)16:59:31 No.108270987▶

>>108270573
>one of about infinity different implementations of IPC. from the before times...
This is because of Unix brain damage. They can't just put things in a library. It has to be in the kernel for all time and you can't ever remove it. And they can't just extend features. They have to make a completely different feature that does the same thing and keep both around forever.

Anonymous
03/01/26(Sun)17:39:59 No.108271270

Anonymous 03/01/26(Sun)17:39:59 No.108271270▶

>>108270573
>what IPC do you use?
AF_UNIX SOCK_SEQPACKET/SOCK_DGRAM, with socketpair() whenever possible.

>>108270928
This guy knows.

Anonymous
03/01/26(Sun)19:45:36 No.108272011

Anonymous 03/01/26(Sun)19:45:36 No.108272011▶

bampu

Anonymous
03/01/26(Sun)19:46:54 No.108272022

Anonymous 03/01/26(Sun)19:46:54 No.108272022▶

>>108270573
How did this ever get implemented by a real, breathing human? How do you miss the point so hard?

Anonymous
03/01/26(Sun)20:43:06 No.108272288

Anonymous 03/01/26(Sun)20:43:06 No.108272288▶

>>108272022
what's your big problem with them?
>>108270987
certain primitives simply cannot be implemented in user space. and even if some of them could be abstracted out of kernel parts, doing it all in-kernel is much more performant

Anonymous
03/01/26(Sun)21:07:15 No.108272404

Anonymous 03/01/26(Sun)21:07:15 No.108272404▶

File: 1762998239037806.png (102.8 KB)