reiser4 and 2.6.31

[kernelOfTruth]

Hi guys,

Edward has released the reiser4 for 2.6.31 patch:

kernel.org directory

kernel.org directory (EU mirror)


anyone else is suffering from the fact that the latest reiser4-incarnation doesn't support additional mount-options besides noatime,nodiratime anymore ?

[ 688.454026] WARNING: Unrecognized option: "tree.cbk_cache_slots=48"
[ 719.333757] reiser4[mount(21267)]: parse_options (fs/reiser4/init_super.c:253)[nikita-2307]:
[ 719.333760] WARNING: Unrecognized option: "tree.cbk_cache.nr_slots=32"
[ 974.412858] reiser4[mount(21270)]: parse_options (fs/reiser4/init_super.c:253)[nikita-2307]:
[ 974.412859] WARNING: Unrecognized option: "tree.cbk_cache.nr_slots=32"
[ 985.749312] reiser4[mount(21272)]: parse_options (fs/reiser4/init_super.c:253)[nikita-2307]:
[ 985.749314] WARNING: Unrecognized option: "tree.cbk_cache_slots=48"
[ 1031.902314] reiser4: sdb6: found disk format 4.0.0.
[ 1263.843430] reiser4[mount(21312)]: parse_options (fs/reiser4/init_super.c:253)[nikita-2307]:
[ 1263.843432] WARNING: Unrecognized option: "flush.scan_maxnodes=3000"
[ 1799.561675] reiser4[mount(21353)]: parse_options (fs/reiser4/init_super.c:253)[nikita-2307]:
[ 1799.561677] WARNING: Unrecognized option: "atom_min_size=0x100"
[ 1832.505679] reiser4[mount(21356)]: parse_options (fs/reiser4/init_super.c:253)[nikita-2307]:
[ 1832.505681] WARNING: Unrecognized option: "atom_max_flushers=0x1"
[ 1872.971929] reiser4[mount(21358)]: parse_options (fs/reiser4/init_super.c:253)[nikita-2307]:
[ 1872.971931] WARNING: Unrecognized option: "optimal_io_size=65

-> it's in the code but fs/reiser4/init_super.c has changed somewhat besides that I removed reiser4-support from my zen-sources tree and patched it in so that might be the culprit

any comments and experiences with it ?


interesting NEWS:

- it seems to sync data much faster than before <-- great news since that IMO was one of its achilles heels (besides bad delete performance)
- delete performance still needs to be evaluated

opinion:

reiser4 is getting ready for prime time

[costel78]

I used, until now, gentoo-sources-2.6.31-r1 patched with 2.6.30-reiser4 patch
I just reinstall it with 2.6.31-reiser4 patch now and yes, tree.cbk_cache_slots option is missing.
Also removed

Code: Select all

for i in `pidof pdflush` ; do ionice -c1 -p $i ; done > /dev/null
for i in `pidof pdflush` ; do renice -20 $i ; done  > /dev/null

from /etc/conf.d/local.start and I didn't notice any slowdown.
Due to it's big size openoffice it's the only package compiled from HDD

Code: Select all

genlop -t openoffice
 * app-office/openoffice

     Tue Oct  6 10:39:07 2009 >>> app-office/openoffice-3.1.1
       merge time: 1 hour, 4 minutes, 28 seconds.

     Tue Oct  6 14:27:11 2009 >>> app-office/openoffice-3.1.1
       merge time: 1 hour, 2 minutes, 8 seconds.

I know that openoffice compile time it's not proper for reiser4 performance test, but it give me a good impression of reiser4 slowdown when was the case, especially during install/merge phase.

[kernelOfTruth]

thanks for the confirmation costel78 !

is that compilation-time slower than usual ?


1) you can find a fix at http://thread.gmane.org/gmane.comp.file ... ocus=22626

or

2) revert reiser4-for-2.6.31 patch; re-download reiser4-for-2.6.31 patch and apply the new one

kudos to Edward

[costel78]

You're welcome!
Actually the second compile from genlop is with new reiser4 patch, and it's not slower.
The kernel it's full new - unmerge old one, remove /usr/src/linux*, modified my reiser4-gentoo-sources ebuild to pick reiser4-for-2.6.31.patch.bz2, emerge it, compiled, rebooted and used. After that I started openoffice build.