ru.unix.bsd

 
 - RU.UNIX.BSD ------------------------------------------------------------------
 From : Slawa Olhovchenkov                   2:5030/500     20 Nov 2006  14:41:42
 To : All
 Subject : Вести с полей
 -------------------------------------------------------------------------------- 
 

 
 20 Nov 06, Slawa Olhovchenkov writes to All:
 
  SO> Porting ZFS to FreeBSD
 
  SO>    URL:
  SO>    http://perforce.freebsd.org/depotTreeBrowser.cgi?FSPC=//depot/user/pjd
  SO>    /zfs
  SO>    URL: http://www.opensolaris.org/os/community/zfs/porting/
  SO>    URL: http://docs.freebsd.org/cgi/mid.cgi?20060822104516.GB16033
 
  SO>    Contact: Pawel Jakub Dawidek <pjd@FreeBSD.org>
 
  SO>    My work is moving slowly forward. ZVOL is, I believe, fully functional
  SO>    (I recently fixed snapshots and clones on zvols), which means you can
  SO>    put UFS on top of RAID-Z volume, take a snapshot of the volume, clone
  SO>    it if needed, etc. Very cool. The hardest part is the ZPL layer, I'm
  SO>    still working on it. Most file system methods work, but probably need
  SO>    detailed review and many fixes. Most of the time these days I'm
  SO>    spending on implementing mmap(2) correctly. It works more or less in
  SO>    simple tests but fails under fsx program. On the other hand, 'fsx -RW'
  SO>    works very stable and reliable. Other test programs (those that don't
  SO>    use mmap(2)) also work quite well. There is still a lot of work to do,
  SO>    mostly in ZPL area, many clean-ups, etc. Some functionality (like
  SO>    ACLs) I haven't even tried to touch yet.
 
 Hi.
 
 This is a first set of patches, which allows to use ZFS file system from
 OpenSolaris on FreeBSD.
 
 To apply the patch you need to have recent FreeBSD source (be sure you
 have rev. 1.284 of src/sys/kern/kern_synch.c).
 
 To try it out you need i386 machine (this is what I tested) and kernel
 without WITNESS compiled in (there are probably some warnings still).
 
 Currently it can only be compiled as a kernel module.
 
 To apply the patch you need the following steps:
 
         # cd /usr/src
         # mkdir -p cddl/lib/lib{avl,nvpair,umem,uutil,zfs,zpool}
         # mkdir -p cddl/usr.bin/ztest
         # mkdir -p cddl/usr.sbin/{zdb,zfs,zpool}
         # mkdir -p compat/opensolaris/{include,misc}
         # mkdir -p contrib/opensolaris/cmd/{zdb,zfs,zpool,ztest}
         # mkdir -p contrib/opensolaris/common/{acl,avl,nvpair,zfs}
         # mkdir -p contrib/opensolaris/head
         # mkdir -p contrib/opensolaris/lib/libnvpair
         # mkdir -p contrib/opensolaris/lib/lib{uutil,zfs}/common
         # mkdir -p contrib/opensolaris/lib/libzpool/common/sys
         # mkdir -p sys/compat/opensolaris/{kern,machine,rpc,sys}
         # mkdir -p sys/contrib/opensolaris/uts/common/fs/zfs/sys
         # mkdir -p sys/contrib/opensolaris/uts/common/{os,rpc}
         # mkdir -p sys/contrib/opensolaris/uts/common/sys/fm/fs
         # mkdir -p sys/contrib/opensolaris/uts/common/sys/fs
         # mkdir -p sys/modules/zfs
         # fetch http://people.freebsd.org/~pjd/patches/zfs_20061117.patch.bz2
         # bzip2 -d zfs_20061117.patch.bz2
         # patch < zfs_20061117.patch
         # make buildworld
         # make kernel
         # make installworld
         # kldload zfs.ko
         (zfs and zpool command should work now)
 
         Before reboot:
         # zfs export <your_pool>
 
         After reboot:
         # kldload zfs.ko
         # zfs import <your_pool>
 
         After a panic:
         # kldload zfs.ko
         # zfs mount -a
         # zfs volinit
 
 Most of the functionality should work, but there are exceptions.
 
 zfs share/unshare don't work yet, you also won't be able to export ZFS
 files systems via NFS.
 
 ACLs don't work yet.
 
 The ZFS file system is MPSAFE (it operates without the Giant lock), but
 performance isn't quite there yet. Please do not report that it is
 slower than UFS, etc. I know it is. On the other hand you should report
 if there are some huge differences in performance between UFS and ZFS,
 for example if ZFS is few times slower in some workloads.
 
 Under very heavy load (or maybe even under not that heavy load, but
 after a longer time) it may panic with
 "kmem_malloc(X): kmem_map too small: Y total allocated"
 message. The back-presure mechanism doesn't work well and SUN guys are
 helping me to figure out why. If you see such panic, please do not
 report it, just reboot your machine and continue (or not).
 
 Please do report any other strange panics or situations (like various
 commands not working as they should, you see strange file system
 behaviour, etc.), _but_ before reporting any issue, verify that it
 wasn't already reported on freebsd-fs@FreeBSD.org mailing list.
 
 If you have any questions or comments, I'd prefer if you send them to
 the mailing list instead of me privately, as it's quite possible others
 would like to know too.
 
 Good luck and enjoy!
 
 Big thanks to ZFS developers for great work and to SUN for opening ZFS
 source!
 ... Специалист подобен флюсу: полнота его одностороння.
 --- GoldED+/BSD 1.1.5
  * Origin:  (2:5030/500)
 
 

Вернуться к списку тем, сортированных по: возрастание даты  уменьшение даты  тема  автор 

Тема:    Автор:    Дата:    Вести с полей   Slawa Olhovchenkov   20 Nov 2006 14:08:58   Вести с полей   Slawa Olhovchenkov   20 Nov 2006 14:41:42