uio.9: Document uiomove_fromphys()

[0mp]

Summary

MFC after: 3 days

[royger]

share/man/man9/uio.9

165

I don't think this is fully accurate. uiomove_fromphys() is implemented on i386 that lacks a directmap using sf_bufs. I think whether the underlying implementation uses the direct map or transient mappings is not relevant in the context of the man page.

[kib]

share/man/man9/uio.9

165

Exactly, the point of uiomove_fromphys() is to provide machine-independent helper to copy to/from memory in the physical address space.

169

Described how? You should explain somehow that array provides the locations that form the place to target the copying. I do not know how to formulate it concise but understandable.

[royger]

share/man/man9/uio.9

169

Maybe use: "The source physical address are provided in the ma[] parameter."? I'm not a native speaker, so others might have better suggestions.

[kib]

share/man/man9/uio.9

169

'adresses' then, each page in the ma[] array provides the page-sized chunk of the phys space. It is not necessary contiguous.

[royger]

share/man/man9/uio.9

169

Ops, yes, sorry, that was a typo on my end.

[0mp]

• Improve the wording around the ma[] argument

[0mp]

share/man/man9/uio.9

165

It should be clearer now.

169

I added a note that ma[] contains physical page addresses and that those pages are not necessarily contiguous.

[kib]

share/man/man9/uio.9

164

165

169

Remove this line.

[0mp]

• Address kib's feedback

[0mp]

Thanks!

[markj]

share/man/man9/uio.9

172

It's the offset into the array, I believe it does not have to lie within the first page.

178

I'm not sure why this is worth noting...?