User: Password:
Subscribe / Log in / New account

Quote of the week

Quote of the week

Posted Jun 19, 2004 18:58 UTC (Sat) by khim (subscriber, #9252)
In reply to: Quote of the week by wawjohn
Parent article: Quote of the week

Hans have good understanding of technical problems but when supposedly stable filesystem will happily eat your data for breakfast... it really does not matter if it's other kernel developers immaturity or Hans's arrogance - you are forced to switch to other filesystem.

Now facts...

1. Reisersfs 3.x with 2.6 kernel will eat your data in default configuration (4K stacks!) - it happened to me when I've tried to switch to 2.6 without throwing reiserfs away.

2. Hans does not care since his perfect filesystem is "stable" and so everyone just need to switch to 8K stacks and stop whining.

Sorry, guys. Filesystem is important part of kernel but not the most important part. 4K stacks are important for heavy load and good filesystem is important for heavy load as well. So I'm just forced to use working version - XFS. I do not like XFS design, I think it's unnecessary huge and ugly but... XFS does work for me and reiserfs does not. I'm not really inclined to hear guy who says that XFS does work and reiserfs does not since kernel developers are "bunch of young kids who lack professional experience in release management". Especially when his supposed maturity makes fixing of broken fs more difficult.

(Log in to post comments)

Quote of the week

Posted Jun 21, 2004 18:40 UTC (Mon) by chad.netzer (subscriber, #4257) [Link]

4K stacks are not default for the generic kernel, and if your distribution turns them on by default, it should have a method for dealing with possible Reiser problems (if they indeed exist, which has actually been up for some debate; I'll take your word for it).

As he says, Reiser had been developed for and debugged for 8K stacks. Changing to 4K stacks is a drastic change in programing assumptions, and changing Reiser to accomodate them results in new, "unstable" code that hasn't been tested. He simply advocates not relying on new, untested code, during a "stable" (ie. 2.6 series) kernel regime.

BTW - Incendiary things are said on lkml all the time; this one is tame by comparison.

reiser3 corruption

Posted Jun 25, 2004 12:40 UTC (Fri) by joern (subscriber, #22392) [Link]

According to Arjan, who I fully trust, you effectively had 4k stacks with 2.4 already. The 8k are shared with a struct task_struct (slightly over 1k) and all interrupts. If 4k are not enough, 2.4 will occasionally break as well. Better to always break, as it doesn't create false hope.

That said, I am quite surprised to see reports of Reiser3 and 4k stacks causing any form of corruption. My checker didn't find any call trace over 3.2k and only one recursion that looks quite harmless. Can anyone back that up or even reproduce the corruption?

Quote of the week

Posted Jun 24, 2004 8:59 UTC (Thu) by job (guest, #670) [Link]

These "facts" are not true. I've used most release kernels in 2.6, both
Linus' and Andrew's trees, and none have had any problems with my
reiserfs. I'm not sure I've used 4k stacks yet, since that is not the
default, contrary to what you say.

Copyright © 2017, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds