Reiser4 Gentoo FAQ [25Dec2016]

Message

dusanc · Post by **dusanc** » Mon Sep 01, 2008 6:25 pm

As I've noticed that some Reiser4 questions get asked all the time I think that this thread will be useful so that people should have one central place for answers.
All the Q and A will be in this first post.
_______________________________________________________________________
Q1: How to start using Reiser4?

You need:
1. kernel (>= 4.5.3) patched with latest patches from http://sourceforge.net/projects/reiser4 ... linux-4.x/
2. reiser4.progs-1.1.0 or higher http://sourceforge.net/projects/reiser4 ... er4-utils/
For HDDs:
3. Create reiser4 partition with:
Reiser4 with cryptocompression:

Code: Select all

mkfs.reiser4 /dev/xxx

Reiser4 without cryptocompression:

Code: Select all

mkfs.reiser4 -o create=reg40 /dev/xxx

4. Mount the new partition with these mount options:

Code: Select all

noatime

For SSDs:
3. Create reiser4 partition with:
Reiser4 with cryptocompression:

Code: Select all

mkfs.reiser4 -d /dev/xxx

Reiser4 without cryptocompression:

Code: Select all

mkfs.reiser4 -d -o create=reg40 /dev/xxx

4. Mount the new partition with these mount options (txmod=wa for COW, discard for Trim support):

Code: Select all

noatime,txmod=wa,discard

Q2: Why does du reports wrong size of data?

"du (1) statistics does not reflect (online) real space usage, as
i_bytes and i_blocks are unsupported by cryptcompress plugin
(supporting those fields "on-line" leads to performance drop).
However, their proper values can be set "offline" by reiser4.fsck." -Edward Shiskin
http://marc.info/?l=reiserfs-devel&m=11 ... 826615&w=2

OBSOLETE Q3: I get lots of errors about wrong size when fscking cryptocompress R4 partition. Am I in trouble?

No you're not. These are false positives when checking cryptocompress data. That's proper values mentioned in Q2. As of version of reiser4.progs-1.0.7 these issues are fixed. These warnings can be suppressed with -s.

OBSOLETE Q4: I'm having trouble with OpenRC and Reiser4, how can I fix it?

Set fsck_args="-a" in /etc/conf.d/fsck
Shouldn't be a problem with recent OpenRC

OBSOLETE Q5: Can I use loopback files on reiser4?

You can on plain Reiser4. On Reiser4+cryptocompress you have to create loopback file this way:

Code: Select all

dd if=incompressible_file of=loopfile bs=65537 count=1

Now using some incompressible file we created new file (of has to be new file) that R4 will remember as incompressible.
http://marc.info/?l=reiserfs-devel&m=12 ... 103900&w=2
This item should be fixed in recent kernels.

Q6: What are the R4 strong points?

Usage with small files, low fragmentation especially with cryptocompression, efficient space usage (usual / without /home is 40% smaller with R4+cc than with plain ext3), robustness wrt. power failures, very fast for certain workloads (eg. for /usr/portage).
When there are SSDs and HDDs in one system R4 can be optimized for both, to use COW allocation on SSD and Journal (or Hybrid-default) on HDD for maximum performance.

Q7: What are R4 weak points?

Currently some large multimedia files are falsely considered as compressable.
Loopback support still has some rough edges.
Performance with atime can be low on cryptocompressed partitions, and deletion of large files on cryptocompressed partitions can take more time than on plain partitions. This is a trade off for better overall performance.

Q8: Where can I find some benchmarks?

http://marc.info/?l=reiserfs-devel&m=12 ... 609180&w=2
http://openbenchmarking.org/result/1104 ... -110329637

Q9: How would I know what's the right fs for me?

Every fs has its strong points, so try to use those which suits you most. Test them not with some synthetic benchmarks, but with your usage patterns. Comparative tests have to be under same conditions (same partition etc.).

Q10: I've formated my / with r4 and now can't boot. What should I do?

You'll have to use a separate /boot partition (50-100MB ext2).

Q11: When will Reiser4 be included in mainline Linux kernel?

Ask $DEITY.

I don't see any technical obstacles for Reiser4 inclusion. There are only organisation ones: I don't think it will be accepted without support from some vendor.
Thanks,
Edward.

http://www.phoronix.com/scan.php?page=n ... &px=OTA2Mw

Q12: Where can I find more informations?

There's a mailing list reiserfs-devel@vger.kernel.org with archives and IRC channel #reiser4 at irc.oftc.net
There's a wiki too at http://reiser4.wiki.kernel.org/index.php/Main_Page

Q13: Where can I find a LiveCD with Reiser4 support?

www.sysresccd.org supports Reiser4 from version 3.1.1: http://sourceforge.net/projects/systemr ... esccd-x86/
www.elivecd.org supports Reiser4 from version 2.4.0: http://www.elivecd.org/download/beta/

Q14: Any bugs that I should know of?

All of these bugs are reported as fixed as of reiser4-for-3.7 http://marc.info/?l=reiserfs-devel&m=13 ... 421045&w=2
There is a bug in patches for kernels 3.17 to 4.5 because kernel dropped barriers support but R4 still relied on it. If you use reiser4 patch <4.5.3 add "no_write_barrier" mount option.
It is fixed as of reiser4-for-4.5.3.patch

Q15: Are there any tuning options?

There is a dedicated Reiser4 tuning thread at http://forums.gentoo.org/viewtopic-t-707465.html

Q16: Why is there no patch for just released linux kernel?

VFS layer in linux kernel changes from time to time, and fixing the R4/VFS interaction is usually a nontrivial task because bugs can be easily introduced and because we like our data. All we can do is wait.

Q17: Does it support encryption?

Edward wrote:Actually it is implemented on a low level. Moreover,
long time ago I got a reiser4 partition transparently
compressed and ciphered by aes ecb. The problem
was that Linux didn't have a good key manager. Now
I believe that the situation is improved, so in order to
announce encryption we need:
1. a fast iv-generator stable to watermark attack for
cbc encryption mode;
2. to adjust any convenient key manager to perform
encryption per file/directory/volume.

I estimate the complexity of this task as 2 of 5.

Thanks,
Edward.

http://marc.info/?l=reiserfs-devel&m=12 ... 120315&w=2

Q18: Are there any special options when using Reiser4 on a SSD? What about TRIM support?

Edward wrote:When formatting your SSD partition by mkfs.reiser4 use the option -d: it will issue discard request for the whole partition before creating reiser4 structure on it. This option is available in reiser4progs-1.0.9

It's advisable to not use Reiser4 compression with SSD controllers that compress data (sandforce ones) as you'll include overhead for compressing the data twice.
As of reiser4-for-3.14.1.patch there is an option for different transaction model per fs so you can activate COW mode for your SSD partitions with a mount option: "txmod=wa"
http://forums.gentoo.org/viewtopic-t-986096.html

Also as of reiser4-for-3.16.2.patch reiser4 supports Trim feature http://en.wikipedia.org/wiki/Trim_%28computing%29 http://marc.info/?l=reiserfs-devel&m=13 ... 207357&w=2 which enable automatic discarding of deleted blocks with mount option:"discard". If you use controller/SSD drive <SATA3.1 you could induce a performance penalty on other FSes because Trim is slow on those but Reiser4 solves this problem by queuing discard requests together and by dispatching them at the end of queue..
http://forums.gentoo.org/viewtopic-t-990364.html

dusanc wrote:If you align your FS, use -noatime as you should do with any FS on SSD, thing is that good SSD controllers won't deteriorate easily over time (like Intel or sandforce ones). Not using your SSD for portage, tmp, or swap helps too

http://forums.gentoo.org/viewtopic-p-65 ... ml#6533722

Q19: I have very long mount times with my large Reiser4 partition. What should I do?

Use dont_load_bitmap mount option.

+dont_load_bitmap
+ Don't load all bitmap blocks at mount time, it is useful for
+ machines with tiny RAM and large disks.

Q20: I need $FEATUREX. What should I do?

$FEATURE1=Encryption
Use LUKS

$FEATURE2=Snapshots
Use LVM or Rsnapshot

$FEATURE3=Subvolumes
WIP so use LVM

$FEATURE4=RAID
WIP so use MD RAID

Q21: Is there a Git repository?

https://github.com/edward6/reiser4

Q22: What do these mount options mean?

Active mount options can be shown with:

Code: Select all

cat /proc/mounts

tmgr.atom_max_size=N - Atoms containing more than N blocks will be forced to commit. N is decimal. Default=TotalRAM/4
tmgr.atom_max_age=N - Atoms older than N seconds will be forced to commit. N is decimal. Default=600s
tmgr.atom_min_size=N - In committing an atom to free dirty pages, force the atom less than N blocks in size to fuse with another one. Default=256blocks
tmgr.atom_max_flushers=N - limit of concurrent flushers for one atom. 0 means no limit. Default=1
tree.cbk_cache.nr_slots=N - Number of slots in the cbk cache.
flush.relocate_threshold=N - If flush finds more than FLUSH_RELOCATE_THRESHOLD adjacent dirty leaf-level blocks it will force them to be relocated.
flush.relocate_distance=N - If flush finds can find a block allocation closer than at most FLUSH_RELOCATE_DISTANCE from the preceder it will relocate to that position.
flush.written_threshold=N - If we have written this much or more blocks before encountering busy jnode in flush list - abort flushing hoping that next time we get called this jnode will be clean already, and we will save some seeks.
flush.scan_maxnodes=N - The maximum number of nodes to scan left on a level during flush.
optimal_io_size=N - The preferred IO size in bytes for programs to write to Reiser4.
32bittimes - turn on 32 bit times.
dont_load_bitmap - Don't load all bitmap blocks at mount time, it is useful for faster boot time.
atomic_write - disable transaction commits during write().
discard - enable issuing of discard requests, useful for SSDs.
dont_punch_holes - disable hole punching at flush time
onerror= - what to do in case of error, available values remount-ro, panic, NULL. Default=NULL
txmod= - which transaction model [journal (for HDDs), cow (for SSDs), hybrid (default)] is used to commit transactions.

_____________________________________________________________________

I hope that this will help people and I would be very grateful for every addition/correction to this FAQ.

Have fun
Dushan

Edit:
25Dec2016: Updated Q20 and added Q22.
25Sep2016: Added Q21.
29May2016: Updated Q1, added info about removed barriers bug to Q14.
10Dec2014: Rearranged Q1, Q3-Q5 now obsolete, added more info to Q6, updated Q13 with new liveCD, added "discard" mount option to Q18, added Q20.
30Oct2014: Added Q19.
25Oct2014: Q18: R4 now supports trim.
06May2014: Q18: Added info about new features for SSD drives (COW and Trim support per volume).
01Dec2013: Q18:Added comment about not using compression with SSD controllers that compress data (sandforce).
17Jan2013: Updated Q1 and Q14 as there are no more open bug reports.
08Dec2012: Q13: R4 support is back in SystemRescueCD. Q14: Added some current bug info.
10Sep2012: Updated links in Q1 and Q13, and updated Q14.
15Nov2011: Updated to new location of files since kernel.org was compromised. Updated Q14.
11Jun2011: Updated Q11.
30May2011: Added new bug info to Q14.
04Apr2011: Added more benchmarks and updated Q10 (there's no grub support for R4).
08Feb2011: Updated Q14.
04Jan2011: Added more info to Q18.
30Dec2010: Updated Q1, Q7, Q14, added Q18.
13Jan2010: Added Q17, bugs info to Q14, wiki link to Q12, and inclusion info to Q11.
11Jan2010: Added Q16.
25Jul2009: Added info about hitchhiker sources.
07Apr2009: Added info about GCC build bug, and about tuning thread.
27Feb2009: Added info about reiser4.progs-1.0.7, and brought some issues up to date.
08Sep2008: Added minimum kernel version to use and location of R4 LiveCD, thanks to gentree and neuron.

Mr. Tao · Post by **Mr. Tao** » Mon Sep 01, 2008 7:31 pm

Nice summary. I just would like to add that after experimenting with loop files on R4 a lot (different kernels and machines) I wouldn't recommend this for serious use (like portage or ccache). It always led to problems, sometimes fatal, at least for me. Typical symptoms are long syncs (in order of tenths of minutes) ocassionaly ending with on-loop fs corruption as fatal that it couldn't be recovered by fsck. Since I gave up and use separate partitions for used-to-be-in-loop-file content I'm very satisfied with Reiser4

DigitalCorpus · Post by **DigitalCorpus** » Tue Sep 02, 2008 7:40 pm

Since there are two types of compression, do you want to include how to enabled one over the other and a simple speed/performance trade off? SPeaking of which, has anyone identified how much more compression one would get on average and how much of a performance impact there is?

Code: Select all

create=ccreg40,compress=gzip1

and

Code: Select all

create=ccreg40,compress=lzo1

Gentree · Post by **Gentree** » Sun Sep 07, 2008 11:38 pm

It always led to problems, sometimes fatal, at least for me.

Oh, I'm sorry to hear that. I hope your death was quick and painless.

Personally, I've found R4 rock solid despite all too frequent power outs here, (though I don't use compression, the rare times I've used loopback has been fine).

Nice overview. You could add which kernel version is needed for compression plugin.

regards.

neunon · Post by **neunon** » Mon Sep 08, 2008 12:20 am

Might be good to add that SystemRescueCD can be used if you want to set up a Gentoo install with the root FS being Reiser4.

MostAwesomeDude · Post by **MostAwesomeDude** » Tue Sep 09, 2008 7:08 am

Gentree wrote:
It always led to problems, sometimes fatal, at least for me.
Oh, I'm sorry to hear that. I hope your death was quick and painless.

Well, one for two, I guess.

Personally, I've found R4 rock solid despite all too frequent power outs here, (though I don't use compression, the rare times I've used loopback has been fine).

The reports are solid, although I've found it to be a bit verbose while debugging.

Nice overview. You could add which kernel version is needed for compression plugin.

Personally, I prefer compression by pieces.

minor_prophets · Post by **minor_prophets** » Fri Jan 16, 2009 1:52 am

Feel like sharing some /etc/fstab examples? R4, of course.

Dont Panic · Post by **Dont Panic** » Fri Jan 16, 2009 3:00 pm

This is a simple example I have for my Reiser4 root partition

Code: Select all

/dev/sdb6    /          reiser4                 noatime,nodiratime        1 1

With the latest versions of Mozilla Firefox, I've found that 'noatime' is fairly important.

There's another Reiser4 tuning thread going on here that has more examples:

http://forums.gentoo.org/viewtopic-t-707465.html

minor_prophets · Post by **minor_prophets** » Fri Jan 16, 2009 7:59 pm

Simple as that, huh.

I saw an interesting fstab on root part. over on the "Filesystems Choice thread" by KernelofTruth(?) I think. I'll post it when I reach my place of peace...

minor_prophets · Post by **minor_prophets** » Sat Jan 17, 2009 1:07 am

taking a look at that thread in a minute.

I remember the opts not.

noatime,nodiratime,flush,scan_maxnodes=15000

I'm just trying to find a link to reiser4 mount options and also reiser4 patches for grub

Dont Panic · Post by **Dont Panic** » Sat Jan 17, 2009 1:34 am

With respect to grub, I took the easy way out, and copy my boot files to a regular reiserfs partition, and boot from that partition initially.

minor_prophets · Post by **minor_prophets** » Sat Jan 17, 2009 1:39 am

I was just about to post a "nevermind" regarding the reiser4 grub patch as I realized I probably don't need it.

/boot is ext2 and I have the kernel sources patched and reiser4 support activated there.

This is correct, no?

neuron · Post by **neuron** » Sat Jan 17, 2009 12:09 pm

minor_prophets wrote:I was just about to post a "nevermind" regarding the reiser4 grub patch as I realized I probably don't need it.

/boot is ext2 and I have the kernel sources patched and reiser4 support activated there.

This is correct, no?

Correct

Hell-Razor · Post by **Hell-Razor** » Sat Jan 17, 2009 12:47 pm

Hey guys I finally found some people that seem to like reiser4

Anyway I have reiser4 on my /home partition (and love it) and reiserfs on my / partition. Anyway the problem is ever sense I installed reiser4 it seems during boot localmount takes FOREVER to mount everything. It takes about three-five minutes and I just keep on getting an error saying "waiting on local mount" and it just repeats until it is finally done. Any ideas of what I can do to see what is wrong / speed it up?

Thanks in advanced.

kernelOfTruth · Post by **kernelOfTruth** » Sat Jan 17, 2009 2:40 pm

reiser4progs-1.0.7 is OUT !

kudos to Edward Shiskin and Mathieu Bélanger;

reiserfsprogs 3.6.21 is out, too, btw

copy over the ebuild into your local overlay, rename the ebuild in the portage-tree, comment out any lines for additional patches (e.g. the preen patch) and digest the ebuild

Dont Panic · Post by **Dont Panic** » Sat Jan 17, 2009 4:50 pm

It's good to see some activity and releases in reiserfs and reiser4.

I know a few people who like reiser4, but they're just expecting it to disappear as a project someday.

Some signs of active development, even if it is just mostly bug patches, will hopefully change the perception that it's a near-dead project.

dusanc · Post by **dusanc** » Sat Jan 17, 2009 8:13 pm

Hell-Razor wrote:Hey guys I finally found some people that seem to like reiser4

Anyway I have reiser4 on my /home partition (and love it) and reiserfs on my / partition. Anyway the problem is ever sense I installed reiser4 it seems during boot localmount takes FOREVER to mount everything. It takes about three-five minutes and I just keep on getting an error saying "waiting on local mount" and it just repeats until it is finally done. Any ideas of what I can do to see what is wrong / speed it up?

Thanks in advanced.

Well 3 things.
fsck your partitions
noatime in fstab
If you /home is pretty large (for me 200GB takes 13sec) you could add dont_load_bitmap (I hope it's correct, look in docs) in kernel parameters, so that r4 doesn't load pretty large bitmap to ram on boot time, but on load.

minor_prophets · Post by **minor_prophets** » Mon Jan 19, 2009 3:17 am

Good tips.

I'm using one from kernelofTruth. scan_maxnodes=15000.

My question is rather noobish, though, as I'm questioning its appropriateness for all reiser4 partitions or just the root part.

Hell-Razor · Post by **Hell-Razor** » Mon Jan 19, 2009 3:45 am

dusanc wrote:If you /home is pretty large (for me 200GB takes 13sec) you could add dont_load_bitmap (I hope it's correct, look in docs) in kernel parameters, so that r4 doesn't load pretty large bitmap to ram on boot time, but on load.

My /home dir is huge, im assuming ~700 gigs.
I will check the bitmap thing out when I get home. What is it used for and why would it want to be loaded? and are there any risks with not loading it?

HecHacker1 · Post by **HecHacker1** » Mon Jan 19, 2009 4:37 am

For reference, from:
http://lists.pld-linux.org/mailman/pipe ... 63501.html

Probably not the newest.

Code: Select all

++Mount options
++=============
++tmgr.atom_max_size=N
++	Atoms containing more than N blocks will be forced to commit.
++	N is decimal.
++	Default is nr_free_pagecache_pages() / 2 at mount time.
++
++tmgr.atom_max_age=N
++	Atoms older than N seconds will be forced to commit. N is decimal.
++	Default is 600.
++
++tmgr.atom_max_flushers=N
++	Limit of concurrent flushers for one atom. 0 means no limit.
++	Default is 0.
++
++tree.cbk_cache.nr_slots=N
++	Number of slots in the cbk cache.
++
++flush.relocate_threshold=N
++	If flush finds more than N adjacent dirty leaf-level blocks it
++	will force them to be relocated.
++	Default is 64.
++
++flush.relocate_distance=N
++	If flush finds can find a block allocation closer than at most
++	N from the preceder it will relocate to that position.
++	Default is 64.
++
++flush.scan_maxnodes=N
++	The maximum number of nodes to scan left on a level during
++	flush.
++	Default is 10000.
++
++optimal_io_size=N
++	Preferred IO size. This value is used to set st_blksize of
++	struct stat.
++	Default is 65536.
++
++bsdgroups
++	Turn on BSD-style gid assignment.
++
++32bittimes
++	By default file in reiser4 have 64 bit timestamps. Files
++	created when filesystem is mounted with 32bittimes mount
++	option will get 32 bit timestamps.
++
++mtflush
++	Turn off concurrent flushing.
++
++nopseudo
++	Disable pseudo files support. See
++	http://namesys.com/v4/pseudo.html for more about pseudo files.
++
++dont_load_bitmap
++	Don't load all bitmap blocks at mount time, it is useful for
++	machines with tiny RAM and large disks.

Hell-Razor · Post by **Hell-Razor** » Mon Jan 19, 2009 6:35 am

Sweet thanks HecHacker1. Where do I add this line? just in my grub.conf kernel line ? I would assume so so here is a clip of what I added --

Code: Select all

title Gentoo Linux 2.6.27-r2
root (hd0,0)
kernel /boot/gentoo-sources-2.6.27-r2 root=/dev/sda3 vga=0x317 dont_load_bitmap

kernelOfTruth · Post by **kernelOfTruth** » Mon Jan 19, 2009 9:16 am

Hell-Razor wrote:Sweet thanks HecHacker1. Where do I add this line? just in my grub.conf kernel line ? I would assume so so here is a clip of what I added --
Code: Select all
title Gentoo Linux 2.6.27-r2
root (hd0,0)
kernel /boot/gentoo-sources-2.6.27-r2 root=/dev/sda3 vga=0x317 dont_load_bitmap

those are MOUNT options, so just add them to /etc/fstab

there you might have "defaults", or "noatime,nodiratime" add that thingy there

dusanc · Post by **dusanc** » Mon Jan 19, 2009 10:06 am

Nope, thing like don't load bitmap have to go to kernel options, Edward told me

kernelOfTruth · Post by **kernelOfTruth** » Mon Jan 19, 2009 10:21 am

dusanc wrote:Nope, thing like don't load bitmap have to go to kernel options, Edward told me

ok, well then the documentation is really outdated:

++Mount options

someone should really create a wiki where we can write down all of the important tid bits

Hell-Razor · Post by **Hell-Razor** » Mon Jan 19, 2009 12:03 pm

Alright lets see here, my fstab -

Code: Select all

/dev/sda1		/boot		reiserfs			noauto,noatime,notail				0 2
/dev/sda3		/			reiserfs			noatime,notail						0 1
/dev/sda2		none		swap			sw								0 0
/dev/sdb1		/home/		reiser4			noatime,dont_load_bitmap  			0 2
/dev/cdrom	/mnt/cdrom	iso9660			noauto,ro,user					0 0

how does that look?