reiser4 in 2.6.15?

Message

codergeek42 · Post by **codergeek42** » Wed Dec 21, 2005 5:52 pm

Matteo Azzali wrote:Imho is not tha opinion of "fairly decent" but the "time to broke" that will determine the time to cut off 486 stuffs. 486 are no more sold by about 11 years,
and at least in 5-6 years will be fair to suppose that even the last one is broken...

For consumer PCs and whatnot yes, but what about for Embedded stuff that requires minimal computing power?

Matteo Azzali · Post by **Matteo Azzali** » Wed Dec 21, 2005 5:52 pm

dsd wrote:
epic wrote:one thing though, i got really surprised as i read gentoo devs and forum mods stating
that ReiserFS was "left-behind"/unmaintained, this is _wholly_ untrue, and you should know this
and certainly not made statements about it if your not sure about what you are saying(!)
i've had trouble getting anyone to look at reiser3 bugs reported by gentoo users in the past, nobody is interested. fortunately there havent been any serious bugs reported for a while, so i cant comment on the current situation.

Till now I found a lots of bug here and there, and reported. I'm used to stress any software just as it enters my machine. Still in a ~x86 environment
(since 2 months with kernel compiled for AMD64 machines, but my new processor is a sempron 64 revision E) I could not find any bug
(still have to try for files larger than 4 Gb) other than the resize/move issue (don't resize/move reiserfs partitions whitout a backup of everything).

Just my 2 cents, but we should do a poll: "are you using reiserFS or reiser4?" . Someone could be surprised by the results....

Matteo Azzali · Post by **Matteo Azzali** » Wed Dec 21, 2005 5:55 pm

codergeek42 wrote:
Matteo Azzali wrote:Imho is not tha opinion of "fairly decent" but the "time to broke" that will determine the time to cut off 486 stuffs. 486 are no more sold by about 11 years,
and at least in 5-6 years will be fair to suppose that even the last one is broken...
For consumer PCs and whatnot yes, but what about for Embedded stuff that requires minimal computing power?

Are these 486-based embedded stuffs on sale? If are you talking of industry hardware, is that running linux?

codergeek42 · Post by **codergeek42** » Wed Dec 21, 2005 5:59 pm

Good points. I'm not really sure of embedded things, actually.

Matteo Azzali · Post by **Matteo Azzali** » Wed Dec 21, 2005 11:43 pm

codergeek42 wrote:Good points. I'm not really sure of embedded things, actually.

Nothing bad, I just asked cause in my Industry Electronics exam (a pairs of years ago) we studied
just microcontrollers machines and not microcomputer ones (they said this was cause of the exactly
predictable timings of ASM instructions in microcontrollers.....but in facts our exercise boards were using
the old commodore transformers...)

petrjanda · Post by **petrjanda** » Thu Dec 22, 2005 1:25 am

Sun's ZFS should be ported to Linux and ditch ricer4 altogether.

Matteo Azzali · Post by **Matteo Azzali** » Thu Dec 22, 2005 9:31 am

Should? Is that allowed,a started project or what? Couldn't find anything googling,
just references to Solaris....

DarkFoon · Post by **DarkFoon** » Thu Dec 22, 2005 11:26 am

@petrjanda
I would disagree with the last part.
I believe that Reiser4 has potential to become something really interesting.
I am pleased with reiser3 (ReiserFS) and I look forward to when Reiser4 is as stable as reiser3.
A short story:
When I first came to the land of the penguin, from the land of broken shards of glass, I knew nothing, and was cocky.
I broke my install in 1 reboot. How? I used ext3, and I managed to freeze my desktop (I used redhat 9, I think)
Because I knew that CTRL+ALT+DEL wouldn't work, and I didn't know of anything else to do, I hit the little reset button.
On the reboot, I found that most of my filesystem was damaged and I couldn't even reach a prompt.
Once I learned more(and got a different distro), I tried this "ReiserFS". I again managed to freeze my computer at some point, and with great hesitation, I pressed the reset button. On reboot everything came back just fine. I was shocked.
I got curious, and I decided to delibrately try to break my install. (this computer wasn't all that important) So, I pressed the reset button at random times; during disk writes, idle time, boot. Each and everytime it was fine.
Because of my experiences, I assume (perhaps foolishly; time will tell) that reiser4, when it is mature, will be as amazing to me as ReiserFS was then.

But I disagree with 1U as well. The kernel maintainers aren't "neophobic". Reiser4 isn't mature, and they know it. Thanks to people like himself, the patches are used, and reiser4 gets some testing done. Once it is mature, and stable, and enough people have used it and proved it is stable, then it will be included in the kernel, and I look forward to that day.

This ZFS you are talking about will have to go through the same thing (once it is ported).

sirdilznik · Post by **sirdilznik** » Thu Dec 22, 2005 3:39 pm

DarkFoon wrote: Once I learned more(and got a different distro), I tried this "ReiserFS". I again managed to freeze my computer at some point, and with great hesitation, I pressed the reset button. On reboot everything came back just fine. I was shocked.
I got curious, and I decided to delibrately try to break my install. (this computer wasn't all that important) So, I pressed the reset button at random times; during disk writes, idle time, boot. Each and everytime it was fine.
Because of my experiences, I assume (perhaps foolishly; time will tell) that reiser4, when it is mature, will be as amazing to me as ReiserFS was then.

Reiser4 writes data to the disk "atomically". It either writes the file completely, or not at all. If I understand correctly, this means that data loss/corruption due to power outages/lock-ups is impossible in reiser4. My own experiances support this theory

plasmagunman · Post by **plasmagunman** » Thu Dec 22, 2005 5:43 pm

1U wrote:My data is as safe as yours, you're not the ones that use it so why keep making up false statements about how it works on MY machine? This is NOT ricing because I don't sacrifice any performance, stability, or anything. I simply get a BETTER system, not much to it.

i don't see anyone making statements about how reiser4 works on YOUR machine, actually 5.999.999.999 people on this world don't give anything on your machine. torvalds tries to provide a kernel, which works best on MOST systems. that means: if there are 10 people saying "it works" and ten saying "it doesn't", better wait until the issues are solved. that way you get a stable kernel for "normal" people. for the "bleeding edge"-customer there are mm-sources with all nifty new features, to be used at your own risk. you do have the choice.

What's wrong with Hans cutting off ReiserFS? If he wants to move onto newer projects then good for him. Things evolve, and there's no point of supporting old systems. It's not hard to update a partition.

actually he dropped maintaining reiserfs long before reiser4 got released, so he let us for some months without any filesystem. he did never say: "don't fix that in reiserfs, better upgrade to reiser4", he said: "don't fix that in reiserfs, better wait some months for reiser4".

in the end the stability of reiser4 is not really part of the discussion, there are other reasons it doesn't get included in vanilla. check the most recent discussion (at least it's the most recent i know...). imho it's hans reiser's own fault, and i would like to say: many people are waiting for reiser4 (me too) and hans reiser should do his work and prepare his patch for the vanilla kernel. he said, kernel-inclusion is top-priority, but it really doesn't look like he is very interested in that.

off topic: is "Location: where whites aren't the minority yet" meant to be funny?

EDIT: this post seems quite missplaced, actually i just recently realized that this thread is 5 pages long, i just had read and answered to the first page

anyway: happy hanukkah!

Gentree · Post by **Gentree** » Tue Jan 10, 2006 4:47 pm

[WARNING]
Just a warning for those using or thinking of using R4 on recent (>=2.6.12 kernels)

The current R4 (1.0.5) is a patch for mm kernels where the effort is to include it eventually.

It includes some changes (additional plugins) that make it INCOMPATIBLE with earlier partitions based on 1.0.4
SIMPLY mounting the partition seems to modify it.

Once an older partition is mounted with a 1.0.5 capable kernel it will no longer mount with and earlier kernel: ie your previous stable kernel and many R4 aware live CDs for example.

There is a way back but it's a bit of hassle http://forums.gentoo.org/viewtopic.php?p=3019057

Basically you need to make a choice between <2.6.12 with 1.0.4 or the later kernels and stick with one or the other.

Knowing this in advance helps in making the right choice and taking suitable precautions.

Backwards compatability has never been high on the R4 feature list but this should have been clearly signalled.

Maybe this is one area where improvement could be enforced as a condition of kernel inclusion.

[/WARNING]

I agree with earlier posts , the atomic write is the key feature for me.

I got in a mess recently when a careless error filled my root partition with several linux ISO images. The result was that the system could not even handle 'ls' with crapping itself. I had to hit the reset button. Thanks to atomic writes I was able to boot to a CD delete one file and I was back in business without any corruption.

I have been in power-off situations with ext3 in the past that have ended a lot less joyfully.

I'm not saying ext3 is crap because of that, it's a case of tools for the job. I've never had to waste day untangling the mess after a change in the kernel support for ext3 either. That's why there's a delay in getting R4 into the kernel and why it will be great once it gets accepted. This sort of thing will not happen.

It seems that Herr Raiser has created a very impressive fs with minimal outside help. If kernel inclusion can oblige some more stability and the addition of some other minds with a wider outlook I think this will become a major force in Linux in the not too distant future.

I continue to use R4 on several partitions, as always with due care and attention.

joey_knisch · Post by **joey_knisch** » Tue Jan 10, 2006 6:02 pm

My 2 cents...

cent 1:
Reiser4 does need to settle down before getting vanilla. However someone should make a simple reiser4kernel ebuild and get it in portage for those that demand that sort of thing. That would most likely shut everybody up (no offense).

cent2:
EXT3 fanboys need to stop posting whenever reiser4 is brought up. Nobody should care that much about a stupid fs. I guarantee I can screw up any fs including your precious ext3. The trick is hard resets during a large group of writes to multiple files. You will mess up data no matter what fs you are using. Same goes for XFS, JFS...

sirdilznik · Post by **sirdilznik** » Tue Jan 10, 2006 6:18 pm

joey_knisch wrote:My 2 cents...

cent 1:
Reiser4 does need to settle down before getting vanilla. However someone should make a simple reiser4kernel ebuild and get it in portage for those that demand that sort of thing. That would most likely shut everybody up (no offense).

cent2:
EXT3 fanboys need to stop posting whenever reiser4 is brought up. Nobody should care that much about a stupid fs. I guarantee I can screw up any fs including your precious ext3. The trick is hard resets during a large group of writes to multiple files. You will mess up data no matter what fs you are using. Same goes for XFS, JFS...

cent1: ebuilds for reiser4kernels exist and have existed for a long time(mm,nitro, archck, etc...). MM is in portage and the others just need to be put in an overlay.

cent2: I've had power-outages and hard-locks during all types of heavy multiple file-writes. Never lost/corrupted a single piece of data with R4. I agree ext3 fanboys need to stop bashing. I can (and have) broken ext3.

Gentree · Post by **Gentree** » Tue Jan 10, 2006 6:57 pm

cent2:
EXT3 fanboys need to stop posting whenever reiser4 is brought up.

well I agree in principal but let's not be provocative

everyone has their favourites whether it be FS, NS, OS or GODESS.

What is interesting is the way we get so hot under the collar and feel so threatened and defensive everytime s.o. mentions "the Other One".

I can hear the cries of "burn the heretics" just around the corner.

CKirocZ · Post by **CKirocZ** » Mon Jan 23, 2006 7:47 pm

I have been using Reiser 4 as long as I have been using Gentoo (about 6 months), and I have had problems with portage, xorg, nvidia-kernel, gnome, glibc, gcc, mozilla, mozilla-firefox, cairo..... Well, you get the idea. I have had no problems with Reiser 4 at all. I have 3 different partitions that are Reiser 4 (mainly to control the disk space usage of portage and ccache and their tmp directories because I have "keepwork" in make.conf FEATURES="") No filesystem problems at all !!!!

I am neither a linux newbie nor am I a grisled veteran but I picked Gentoo to learn more about linux. The support in these forums is second to none, but stop bashing Reiser 4 until you have used it ! The constant shameless bashing of Reiser 4 is totally uncalled for and only serves to expose the weaknesses and inexperience of those doing the bashing. It also downgrades others opinions of this forum.

No one has said that Reiser 4 is perfectly stable and ready to be used on the largest servers or the most sensitive data.

Be smart when trying new things, back up all your important data (even Ext 2, Ext 3 users !!), because all things break, whether it be from user error or hardware failure It will happen to you. I hope you're prepared !!

joey_knisch · Post by **joey_knisch** » Mon Jan 23, 2006 8:46 pm

sirdilznik wrote: cent1: ebuilds for reiser4kernels exist and have existed for a long time(mm,nitro, archck, etc...). MM is in portage and the others just need to be put in an overlay.

I believe you misunderstood what I meant. I wasn't clear. While standing on the sidelines for all of these discussions I have come to realize how many r4 people just want it in the Gentoo Sources. The Gentoo Sources people have taken a bit of a hardline and are saying it won't be there until vanilla. I think there just needs to be a compromise. Why not make a reiser4 use flag for the ~x86 Gentoo Sources so that people can deal with it where they want it. In my experience people are too lazy to keep an overly up to date.

sirdilznik wrote:cent2: I've had power-outages and hard-locks during all types of heavy multiple file-writes. Never lost/corrupted a single piece of data with R4. I agree ext3 fanboys need to stop bashing. I can (and have) broken ext3.

I am not saying the FS or data is corrupted in the normal sense. AFAIK the Atomic Writes prevent fs corruption but if the programming for an app is set up a certain way, ( ie if an app has to write to multiple files and only completes flushing one file before power is cut) that app will have problems.

@Gentree: That's what I was getting at. People take it so personally. It is like the Intel/AMD and nVidia/ATI hardware wars used to be.

codergeek42 · Post by **codergeek42** » Mon Jan 23, 2006 8:56 pm

joey_knisch wrote:Why not make a reiser4 use flag for the ~x86 Gentoo Sources so that people can deal with it where they want it. In my experience people are too lazy to keep an overly up to date.

If they're willing to use a deemed-experimental filesystem for their systems, they are assuredly not too lazy to keep their overlay up-to-date. The reason there is no 'reiser4' USE flag or similar would probably be that it would require that dsd and the other Gentoo kernel maintainers would then have the extra load of maintaining that patchset, which they are not willing to put effort into, as it currently does not play well with many things (such as Extended Attribute support, rigid POSIX file semantics, etc.). It has also been stated by dsd that one of the main goals of Gentoo's kernel tree is to deviate as little as possible from what is released upstream (excluding things like backporting security patches from an upstream release candidate or adding squashfs support for the LiveCD stuff, etc.). Then again, I'm not a Gentoo kernel developer, so I may very well be mistaken. Please speak with dsd or another person on the Gentoo kernel team who is, and they'll likely be happy to clarify this for you.

joey_knisch · Post by **joey_knisch** » Mon Jan 23, 2006 9:29 pm

codergeek42 wrote:
joey_knisch wrote:Why not make a reiser4 use flag for the ~x86 Gentoo Sources so that people can deal with it where they want it. In my experience people are too lazy to keep an overly up to date.
If they're willing to use a deemed-experimental filesystem for their systems, they are assuredly not too lazy to keep their overlay up-to-date. The reason there is no 'reiser4' USE flag or similar would probably be that it would require that dsd and the other Gentoo kernel maintainers would then have the extra load of maintaining that patchset, which they are not willing to put effort into, as it currently does not play well with many things (such as Extended Attribute support, rigid POSIX file semantics, etc.). It has also been stated by dsd that one of the main goals of Gentoo's kernel tree is to deviate as little as possible from what is released upstream (excluding things like backporting security patches from an upstream release candidate or adding squashfs support for the LiveCD stuff, etc.). Then again, I'm not a Gentoo kernel developer, so I may very well be mistaken. Please speak with dsd or another person on the Gentoo kernel team who is, and they'll likely be happy to clarify this for you.

Ok sure. I don't really know this stuff that well. I am just looking for a compromise between the two apparently extreme positions. I don't really have anything on the line since I won't be using r4 for quite some time/ever (except for portage part).

energyman76b · Post by **energyman76b** » Mon Jan 23, 2006 11:24 pm

Voltago wrote:
1U wrote:If people want to use it or not that's their choice to make, not the developers. The problem is they are creating inconveniences for everyone who wants to use it based on their own opinions and hatred of new things.
d) If reiser4 is included into mainstream and stuff breaks and data is lost (like it did when reiserfs3 was included, for example), people will come whining to the kernel devs, not to namesys.com. Including not-quite-mature software could damage the kernels reputation for stability.

the reiser3 breakage was mostly (all of it) caused by sudden and hidden vm changes, Linus Torvalds made in early 2.4 timeline.

So, Linus changed something, reiser broke because of the changes, people blamed reiserfs.

That is not very fair. The truth is:
everytime a low level subsystem changes, there is a risk of breakage. Reiserfs was the victim of this in its early days, today the devs are much more carefull about it. But it was not reiserfs cault.

Post by **dsd** » Tue Jan 24, 2006 12:12 am

indeed - the "extreme" approach from the gentoo-sources maintenance side is not any dislike of reiser4, its simply the fact that the cases where we patch the kernel with patches that *aren't* already in Linus' tree is becoming increasingly rare, and this is our aim: deviate from Linus as little as possible.

the USE-flag deal is documented in the FAQ:
http://dev.gentoo.org/~dsd/genpatches/faq.htm

and the more general "mission statement" can be found here:
http://dev.gentoo.org/~dsd/genpatches/about.htm

Gentree · Post by **Gentree** » Wed Jan 25, 2006 11:54 pm

Seems very fair and sensible. There's a bunch of patched kernels available that offer R4 .

I'm glad to hear there's no hardcore prejudice against R4 at the heart of the dev team.

Thanks for you efforts.