Description of Problem: Creating a large ext2 filesystem on Smart Array 5300 may take an extraordinarily long time to complete. The process starts out normally, but slows during the creation of the inode tables. Creating a filesytem of about 175gB on an ADG volume took about 8 to 10 minutes to complete. Creating the same filesystem when using the 2.4.3 tree downloaded fron kernel.org completed within a couple of minutes. Current configuration: Compaq IA64 System pilot-4 w/2 C0 processors & 8gB RAM Smart Array 5300 in slot 2 w/2 logical volumes Smart Array 5300 in slot3 w/2 logical volumes Boot controller: Smart Array 5300 Controller
You filed this under i386, but reported observing the problem on an IA-64. Have you observed this on i386 as well? Please remember to file Itanium defects under ia64 in the future.
What beta version have you tried this with? What is the kernel version? Thanks, Florian La Roche
This was tried with Beta 3 and kernel 2.4.3.
Re-assigned to the kernel.
Al, is this the O(N^2) buffer algo ?
This defect considered SHOULD-FIX for Fairfax gold-release.
It would be very interesting to know if the 2.4.7-0.4 kernel still has this behavior; the relevant kernelcode is totally rewritten to get rid of bad algorithms and such.
Comments from Compaq Engineer The comments from RedHat indicate they have their own 2.4.7 (version 2.4.7- 0.4). When will we see that? Or do they want to know if it's fixed in the kernel.org tree?
2.4.7-0.3 was released to the betatime about half an hour ago, that's the same as 2.4.7-0.4 which will have a rpm depenency fix in it.