Page MenuHomeFreeBSD

Add hammer support for fstyp(8).
ClosedPublic

Authored by pfg on Dec 4 2017, 8:40 PM.
Tags
None
Referenced Files
Unknown Object (File)
Oct 2 2024, 4:07 PM
Unknown Object (File)
Oct 2 2024, 6:32 AM
Unknown Object (File)
Oct 1 2024, 1:27 AM
Unknown Object (File)
Sep 27 2024, 6:58 AM
Unknown Object (File)
Sep 27 2024, 5:17 AM
Unknown Object (File)
Sep 26 2024, 12:49 AM
Unknown Object (File)
Sep 24 2024, 5:58 PM
Unknown Object (File)
Sep 24 2024, 2:54 PM
Subscribers

Details

Summary

Dragonfly ported fstyp(8) and included support for the hammer filesystem
variants. Bringing back the support to FreeBSD seemed to be an easy task
so I gave it a try.

I did have to bring along two basic headers from Dragonfly and change
things a bit to get it to compile.

Test Plan

It remains untested: I don't really have Dragonfly installed and I can't
generate atest image either.

Diff Detail

Repository
rS FreeBSD src repository - subversion
Lint
Lint Not Applicable
Unit
Tests Not Applicable

Event Timeline

At the top of fstyp.8, you need to bump the .Dd to the date of the commit, since this is a content change.

In D13369#279103, @bcr wrote:

At the top of fstyp.8, you need to bump the .Dd to the date of the commit, since this is a content change.

Yes, I just currently don't know anyone that can test this so I have no idea when (or if) it will be committed.

Perhaps Eitan is interested ...

I did not review the header files which look like straight up copies from DragonFly. Most of comments are nits or style and can be ignored if I'm wrong.

Thanks!

usr.sbin/fstyp/hammer.c
38 ↗(On Diff #36205)

nit: I always forget style(9) but I think this is in the wrong place. also sort.

59 ↗(On Diff #36205)

I'd prefer globals to static variables in functions unless we have good reason

61 ↗(On Diff #36205)

where is 64 coming from?

65 ↗(On Diff #36205)

can we name these constants?

149 ↗(On Diff #36205)

pretty sure you can replace most of this with strtok

usr.sbin/fstyp/hammer2.c
36 ↗(On Diff #36205)

nit: sort

70 ↗(On Diff #36205)

nit: I feel like FreeBSD style frowns on mixing declarations and init like this, though I'm not sure that's a good thing

Interesting ...
I took the C files from DragonfLyBSD as well: I only changed them so that they would compile with the headers placed locally.
The style issues should probably be reported upstream, it is not something we should be fixing..
The logical question is if it works at all. The next question is ... do we want this?

usr.sbin/fstyp/hammer.c
38 ↗(On Diff #36205)

This is the only part I wrote ;).

In D13369#280745, @pfg wrote:

The logical question is if it works at all. The next question is ... do we want this?

This is the main thing I was wondering about when I saw this DR. Do we ever plan to have this Hammer thing support, like being able to mount their volumes? What is possible usage scenario of this code in FreeBSD right now, without actual mount support? I fear that this code won't be maintained well and might start to bitrot since day one.

P.S. On an unrelated note: why fstyp(8) is not called fstype(8)?

In D13369#280745, @pfg wrote:

The logical question is if it works at all. The next question is ... do we want this?

This is the main thing I was wondering about when I saw this DR. Do we ever plan to have this Hammer thing support, like being able to mount their volumes? What is possible usage scenario of this code in FreeBSD right now, without actual mount support? I fear that this code won't be maintained well and might start to bitrot since day one.

I think it would eventually make sense to have support for hammer2, and then this would make perfect sense. It doesn't hurt to know more of what is installed in the hard disk.

This said, we can keep it in code review until the time comes when its more useful.

P.S. On an unrelated note: why fstyp(8) is not called fstype(8)?

Solaris and linux use fstyp.

In D13369#280745, @pfg wrote:

Interesting ...
I took the C files from DragonfLyBSD as well: I only changed them so that they would compile with the headers placed locally.
The style issues should probably be reported upstream, it is not something we should be fixing..

Alright.

The logical question is if it works at all.

I could test this out, but it'll take some time. My current dev setup is all VMs and my fbsd time is limited.

The next question is ... do we want this?

IMHO yes.

In D13369#280745, @pfg wrote:

The next question is ... do we want this?

IMHO yes.

I still think that bringing in such amount of code for an otherwise unsupported filesystem is premature. As it was said, we can (and should) keep it in code review until the time comes when it's more useful.

In D13369#280745, @pfg wrote:

The next question is ... do we want this?

IMHO yes.

I still think that bringing in such amount of code for an otherwise unsupported filesystem is premature. As it was said, we can (and should) keep it in code review until the time comes when it's more useful.

In all honesty ... it is not huge and it is self contained. Furthermore it is maintained in DragonFlyBSD. But there is no hurry. let's wait to see if it works first ;).

Since this is an import from DragonflyBSD, any changes to the man page should be sent upstream. So far, I'm fine with it and won't block the integration from the doc side of things.

This revision is now accepted and ready to land.Dec 20 2018, 8:24 PM

cem@ has been adding some interesting support to fstyp so perhaps he may want to review or even take over.

Also upstream has made some small changes, but I have no time to catch up:

https://gitweb.dragonflybsd.org/dragonfly.git/history/HEAD:/usr.sbin/fstyp

Regarding it not being mountable at the moment: that’s what the ‘-u’ option (and the corresponding flag in the file systems list) is for.

No objection. Most of the imported code is headers, which won't result in a huge binary. The small C program is a handful of tiny functions, seems unobjectionable to me. I am not super familiar with hammer and didn't read this code thoroughly, but given most of it comes verbatim from dfly I'm willing to trust their ability to recognize their own FS.

usr.sbin/fstyp/fstyp.c
65–67 ↗(On Diff #36205)

This'll need rebasing now (sorry). I think you want true rather than false as well.

rebase the patches to match modern FreeBSD.

Also, the hammer2 format has changed a bit since the last time I
worked on this so I now brought the changes to about 2019-09-13.

On 2019-10-06 DragonFly added "HAMMER2 PFS label and @name syntax" which
seems specific to their OS.

This revision now requires review to proceed.Dec 24 2019, 3:39 AM
This revision was not accepted when it landed; it landed in state Needs Review.Dec 24 2019, 7:00 PM
This revision was automatically updated to reflect the committed changes.