Page MenuHomeFreeBSD

safexel manuals: improve apropos and HW relnotes
Needs ReviewPublic

Authored by concussious.bugzilla_runbox.com on Nov 22 2024, 8:44 PM.
Tags
None
Referenced Files
Unknown Object (File)
Fri, Dec 20, 8:59 PM
Unknown Object (File)
Sun, Dec 15, 7:46 PM
Unknown Object (File)
Wed, Dec 4, 3:47 PM
Unknown Object (File)
Tue, Nov 26, 3:28 PM
Unknown Object (File)
Tue, Nov 26, 3:28 PM
Subscribers

Details

Summary

safe.4:
Add model numbers to document description for apropos.
Fix synopsis sysctls for apropos (Nm -> Cd, as in vt(4))
Convert the HARDWARE tagged list to a column list for
better rendering in the hardware release notes.
Tested in man(1) at MANWIDTH 59 and 80.

safexcel.4:
Add crypto to document description for apropos.
Create a HARDWARE section for inclusion in the HW relnotes.

Sponsored by: Google
SPDX: both tagged
MFC after: 3 days
MFC to: 13.5

Diff Detail

Repository
rG FreeBSD src repository
Lint
Lint Skipped
Unit
Tests Skipped

Event Timeline

Turns out, actually if the column specification is the way
that the begin list macro wants, the lines can be folded
however, so... fold them better in the src.

Tested on MANWIDTH 59 and 80

+ add crypto to safexcel document description for apropos
+ attempt to improve commit message

share/man/man4/safexcel.4
32
78

This effectively duplicates the wording from the DESCRIPTION section. How exactly does this section get used?

This revision is now accepted and ready to land.Nov 25 2024, 4:33 PM

I am interested in waiting for Mark's feedback upon the HARDWARE section. Better would be to list which devices are supported, if anyone knows anything.

Also, as for his comment to s/security packet/cryptographic offload/, I think those are much more specific/appropriate search keywords, can we do that consistently? There's also hifn(4) https://man.freebsd.org/cgi/man.cgi?query=hifn&apropos=0&sektion=0&manpath=FreeBSD+14.1-RELEASE+and+Ports&arch=default&format=html and ccr(4) https://man.freebsd.org/cgi/man.cgi?query=ccr&apropos=0&sektion=0&manpath=FreeBSD+14.1-RELEASE+and+Ports&arch=default&format=html

Adding jhb for that discussion.

share/man/man4/safe.4
126

Rewrapping the existing text here seems gratuitous?

safe(4) and hifi(4) are mostly obsolete and the kernel no longer supports many of these algorithms (DES, 3DES, MD-5) for IPsec or other crypto(9) consumers. In terms of manpage language, "crypto accelerator" seems to be the common pattern in .Nd descriptions and is what I used in ccr(4). I don't have any strong opinions, though "security packet engine" seems a bit more obtuse. The typical language I see is something like "crypto co-processor" or "crypto accelerator". Mark's suggestion of "offload engine" is also better than "packet engine". OpenSSL has in the past used the term "engine" for accelerators/co-processors though I think those are now supposed to be custom providers in modern OpenSSL jargon rather than engines.

In D47706#1088743, @jhb wrote:

safe(4) and hifi(4) are mostly obsolete and the kernel no longer supports many of these algorithms (DES, 3DES, MD-5) for IPsec or other crypto(9) consumers. In terms of manpage language, "crypto accelerator" seems to be the common pattern in .Nd descriptions and is what I used in ccr(4). I don't have any strong opinions, though "security packet engine" seems a bit more obtuse. The typical language I see is something like "crypto co-processor" or "crypto accelerator". Mark's suggestion of "offload engine" is also better than "packet engine". OpenSSL has in the past used the term "engine" for accelerators/co-processors though I think those are now supposed to be custom providers in modern OpenSSL jargon rather than engines.

"Security packet engine" is the terminology from the datasheets I used to write the driver. The device has a lot of functionality to support offloading of IPSec, TLS, various wifi protocols, ...

share/man/man4/safexcel.4
77
concussious.bugzilla_runbox.com marked 2 inline comments as done.
concussious.bugzilla_runbox.com edited the summary of this revision. (Show Details)

While testing, I discovered the Nm's in SYNOPSIS were causing this page to get pulled into results for apropos sysctl. I rewrote it the way it's written in vt(4).

I do not know if it is accurate though that the sysctls can be either loader.conf or sysctl.conf.

add Sponsored trailer because I'm doing this at work.

Wrap some text back the way it was.

This revision now requires review to proceed.Mon, Dec 2, 3:29 PM

If you mail me git format-patch output or submit it on github, I'll apply it.

While testing, I discovered the Nm's in SYNOPSIS were causing this page to get pulled into results for apropos sysctl. I rewrote it the way it's written in vt(4).

I do not know if it is accurate though that the sysctls can be either loader.conf or sysctl.conf.

Looking at the definitions in safe.c, they can only be set in sysctl.conf, so this part should be corrected.

add Sponsored trailer because I'm doing this at work.

Wrap some text back the way it was.