[reiserfs-list] /etc/fstab
I do not want to put in /etc/fstab, because the partition is mounted at start up. /dev/hdd5 /mnt5 reiserfs notail,noatime0 0 How to mount it manually what options should I give to mount? Thanks.
Re: [reiserfs-list] /etc/fstab
Krasi Zlatev [EMAIL PROTECTED] writes: I do not want to put in /etc/fstab, because the partition is mounted at start up. /dev/hdd5 /mnt5 reiserfs notail,noatime0 0 How to mount it manually what options should I give to mount? The same :-) You can add noauto to the fstab options to prevent mount on boot. See man fstab. -- Matthias Andree
Re: [reiserfs-list] [patch] ignore forcefsck
Vladimir V. Saveliev wrote: Vladimir V. Saveliev wrote: For now we do not want it to be ran by boot scripts. So, 3.x.0k will just ignore -f and -a. Ooops, I meant: -f and -a will make reiserfsck to do nothing but print its version. whereas -r and -p are ignored. Thanks, vs Actually, I would like them to say: reiserfsck is being run with -f or -a, which usually means that your distro maintainer hasn't bothered to adjust the boot scripts to not run fsck on journaling filesystems that don't need or want it. Sigh. Doing nothing. We suggest fixing the boot script. Hans
Re: [reiserfs-list] filesystem interfaces
Ragnar Kjørstad [EMAIL PROTECTED] writes: 1. fsync The current fsync interface lacks a couple of interesting features. I believe lazy-fsync has been discussed, but another useful feature is Wait a minute. The major problem is that mail-servers, regardless of their performance efforts, need to be safe in that they *NEVER* lose any mail. Trading speed for lost mail over a crash is not a good deal. An architecture which defers fsync() for a finite, maximum amount of time to gather multiple fsync() requests into a batch of (possibly sorted, sent into a tagged SCSI command queue) writes and then acknowledging all at once may be a good idea, provided you don't run out of process table slots with all those waiting smtpds, and a mail server can not acknowledge receipt of a mail before all file and meta data is on disk (and not in a fast write cache or something). The caller should be able to wait for fsync to complete by using poll in the case of asyn fsync. It would require introducing a special syscall, I believe. However, in that case, fsync should also sync all meta data related to files. A different aspect is that many mailers expect BSD-type directory modifications are always synchronous semantics. Revisit the ReiserFS-and-qmail issue. On ext2, you can circumvent the problems with chattr +S, but on ReiserFS, there is no such thing. Now link and rename behaves differently with regards to replacing existing files, but what's the logic behind this? What's the problem with that? To clobber, use rename. To be careful, use link and if that succeeds, unlink. How would you like to establish the atomicity of either functionality (either rename or link (without unlink)) in your approach? 3. Ruby I just came across the ruby programming language today - the interesting thing is that this language has a concept of transactions! Does any other languages have this kind of features? Do anyone use them for real software? It would be really cool with a ruby-implementation that actually used filesystem-transactions to implement this instead of the library implementation that I assume ruby uses. That would probably be most useful in things like NFS where a link may succeed, but the success report fails. If this was transaction-oriented, the check if your file's st.n_link has increased to 2, if so, your link has succeeded could be avoided. Not sure if Coda or AFS have concepts like these. -- Matthias Andree
Re: [reiserfs-list] ftp.reiserfs.org down?
Dirk Haage wrote: On 18 May 2001 10:17:03 -0700, Hans Reiser wrote: Vladimir V. Saveliev wrote: You should use ftp.namesys.com. Thanks, vs It should be fixed by now, and reiserfs.org and reiserfs.net should work by now. Hmm, I just get some Infos from domaindiscover, that reiserfs.net/org is reserved :( /dirk reserved by me? Hans
Re: [reiserfs-list] [patch] ignore forcefsck
Dirk Mueller wrote: On Sam, 19 Mai 2001, Hans Reiser wrote: fsck should never be run automatically. That sysadmin might have a company waiting for the server to come up. Then they should have a sysadmin that knows how to disable it. This is not a policy, its just a reasonable default. I don't like much that reiserfs does not integrate into the standard fdisk after a while sheme the usual file systems do. Dirk usual file systems don't do it. netapp sure doesn't. hans
Re: [reiserfs-list] [patch] ignore forcefsck
Chris Mason wrote: On Saturday, May 19, 2001 06:40:48 PM +0200 Dirk Mueller [EMAIL PROTECTED] wrote: On Sam, 19 Mai 2001, Hans Reiser wrote: fsck should never be run automatically. That sysadmin might have a company waiting for the server to come up. Then they should have a sysadmin that knows how to disable it. This is not a policy, its just a reasonable default. I don't like much that reiserfs does not integrate into the standard fdisk after a while sheme the usual file systems do. There are two sides of this. fsck is only now getting strong enough that I would start to recommend that kind of thing. As fsck matures, we should at least have the option of ext2 style automatic runs after X number of mounts/crashes. -chris No, this is an absolutely hideous feature of ext2, if you want to have it print a reminder to the user that they have not run fsck recently, fine, but in the eyes of a suser trying to get their laptop to turn on for their presentation, or their server to not leave the whole company hanging, this feature of ext2 fsck is a bug.
Re: [reiserfs-list] [patch] ignore forcefsck
On Saturday 19 May 2001 12:28, Hans Reiser wrote: fsck should never be run automatically. That sysadmin might have a company waiting for the server to come up. Conversly that sysadmin might be hours away from the box that needs to be fscked before restarting... In this case, which _does_ happen, an automatic fsck can be a good thing. I administer a SAP systems. Ed Tomlinson [EMAIL PROTECTED]
Re: [reiserfs-list] [patch] ignore forcefsck
On Saturday, May 19, 2001 11:31:29 AM -0700 Hans Reiser [EMAIL PROTECTED] wrote: Chris Mason wrote: On Saturday, May 19, 2001 06:40:48 PM +0200 Dirk Mueller [EMAIL PROTECTED] wrote: On Sam, 19 Mai 2001, Hans Reiser wrote: fsck should never be run automatically. That sysadmin might have a company waiting for the server to come up. Then they should have a sysadmin that knows how to disable it. This is not a policy, its just a reasonable default. I don't like much that reiserfs does not integrate into the standard fdisk after a while sheme the usual file systems do. There are two sides of this. fsck is only now getting strong enough that I would start to recommend that kind of thing. As fsck matures, we should at least have the option of ext2 style automatic runs after X number of mounts/crashes. -chris No, this is an absolutely hideous feature of ext2, if you want to have it print a reminder to the user that they have not run fsck recently, fine, but in the eyes of a suser trying to get their laptop to turn on for their presentation, or their server to not leave the whole company hanging, this feature of ext2 fsck is a bug. No. There are many, many different kinds of users. The home user, where fsck should only take a few minutes, should have an (readonly) fsck every once and a while to make sure the FS is correct, especially since these users tend to not have backups. It is a very good feature, that not everyone needs. Huge installations don't want periodic fscks, but huge installations are the exception, not the rule. -chris
Re: [reiserfs-list] [patch] ignore forcefsck
On Sat, May 19, 2001 at 11:31:29AM -0700, Hans Reiser wrote: There are two sides of this. fsck is only now getting strong enough that I would start to recommend that kind of thing. As fsck matures, we should at least have the option of ext2 style automatic runs after X number of mounts/crashes. No, this is an absolutely hideous feature of ext2, if you want to have it print a reminder to the user that they have not run fsck recently, fine, but in the eyes of a suser trying to get their laptop to turn on for their presentation, or their server to not leave the whole company hanging, this feature of ext2 fsck is a bug. Sorry. It's never a bug. You can control the behavior quite easily by tune2fs. Mount count: 3 Maximum mount count: 20 Last checked: Thu May 3 04:16:07 2001 Check interval: 15552000 (6 months) So it's clearly a feature that makes sense at least in some environments. -- ciao - Stefan Man gebe jedem Niedersachsen - seinen eigenen Castor-Kasten. Stefan TrabyLinux/ia32 fax: +43-3133-6107-9 Mitterlasznitzstr. 13 Linux/alphaphone: +43-699-10157505 8302 Nestelbach Linux/sparc http://www.hello-penguin.com Austria mailto:[EMAIL PROTECTED] Europe mailto:[EMAIL PROTECTED]
Re: [reiserfs-list] What about this one liner in fix_node.c?
On Sunday, May 20, 2001 12:07:33 AM +0200 Dieter Nützel [EMAIL PROTECTED] wrote: Hello Chris, is this one usefull or obsolete? diff -ur linux/fs/reiserfs/fix_node.c linux/fs/reiserfs/fix_node.c --- linux/fs/reiserfs/fix_node.c Mon Jan 15 18:31:19 2001 +++ linux/fs/reiserfs/fix_node.cFri Feb 2 15:40:54 2001 @@ -936,6 +936,7 @@ if (p_s_tb-FEB[p_s_tb-cur_blknum]) BUG(); +mark_buffer_journal_new(p_s_new_bh) ; p_s_tb-FEB[p_s_tb-cur_blknum++] = p_s_new_bh; } Still useful, but not strictly a bug fix so it hasn't gone in yet. Without the patch, very balance instensive ops are a little slower (but not much). -chris