I recently celebrated a link to Habré, about HW MTU and IP MTU. Partially like everything is understandable, but I can't see the whole picture.

  1. Why do we generally have a HW MTU, the only thing that sees the meaning is that it will not pass with tags in L2?

  2. in Linux, as I understood, implemented only IP MTU, that is, the dimensions of the L2 Header, MPLS and different tags are discarded?

  3. If, so, are there any opportunities in creating HW MTU on Linux?

Well, where is the link?

Qwertiy♦2021-09-13 15:39:09

HW MTU is something that can iron. Closely consider that this is the size of the receiving buffer in the external interface chip. If he is a half kilobyte, then two in it do not flush purely physically (less -easily).

Akina2021-09-13 15:39:09

@Akina Do not confuse with MRU

eri2021-09-13 15:39:09

@RI MTU -the size of the transmission frame, MRU -the size of the reception frame. With regard to the topic discussed, there is no significant difference IMHO. Well, except for the fact that at each level of the model they are their own.

Akina2021-09-13 15:39:09

HW MTU (MAXIMUM TRANSMISSION UNIT) is usually set by the standard, and above it does not jump. IP MTU depends on it obviously, because The IP package is encapsulated in the underlying level frame. If, for example, you have a gigabit ethernet network, and Jumbo Frames support is enabled everywhere on it, then you can play with the MTU installation of about 8 kB.

Alexander Prokoshev2021-09-13 15:39:09