LinuxKPI: add simplified vesion of page_frag_cache

[bz]

Summary

For the moment and the currently only consumer (mt76) add a simplified
version of the page_frag_cache. We will only accept fragement sizes up
to 1 PAGE_SIZE (KASSERT) and we will always return a full page.
Should we add more consumers or small (or large) objects would become a
problem we can always add a more elaborate version.

Discussed with: markj
MFC after: 3 days

[• hselasky]

sys/compat/linuxkpi/common/src/linux_page.c

453

Doesn't fls() take an integer. While fragsz is size_t ? Not that it matters, but maybe check the valid range?

[bz]

sys/compat/linuxkpi/common/src/linux_page.c

453

Well with the KASSERT above and the 0 check I think the input is pretty limited into fls.

[• hselasky]

sys/compat/linuxkpi/common/src/linux_page.c

453

Can you change that KASSERT() into if (fragsz == 0 || fragsz > PAGE_SIZE) return (NULL); ?

I mean try to avoid panicing?

[markj]

sys/compat/linuxkpi/common/src/linux_page.c

453

I'd still argue for using flsl() here. Or just don't bother with it at all, i.e., just write pages = alloc_pages(gfp, 1);, since it will take a lot more work to support fragsz > PAGE_SIZE.

[bz]

sys/compat/linuxkpi/common/src/linux_page.c

453

The panic is on purpose as if we "hide" the implementation limitation we might have spurious errors one day which may take hours to debug only to find this.

453

It's alloc_pages(gfp, 0); I now understand what @hselasky meant above. Would changing it to flsl() work for both of you?

[• hselasky]

sys/compat/linuxkpi/common/src/linux_page.c

453

Well, only panic if gfp & M_WAITOK && fragsz > PAGE_SIZE ? Else clients will handle and print failures on NULL allocations?

I'm worried that user-space input may be passed directly to fragsz, which is not good.

--HPS

[markj]

sys/compat/linuxkpi/common/src/linux_page.c

453

That's fine with me.

[bz]

Update to flsl() as suggested during review.
Keep the panic in the cases KASSERTs is not a NOP.