gentoo sources?

rommel · Posted: Sun Jun 16, 2002 10:51 pm Post subject: gentoo sources?

i just rebuilt the kernel i had been running and i used the latest gentoo sources....my lsr lost like 15mb/s...its down to 68mb/s from the tranfer rates i was getting when i was running the vanilla kernel...whats the benefit of the gentoo kernel?...am i not configuring something right?

ciao

delta407 · Posted: Mon Jun 17, 2002 2:40 am Post subject:

The Gentoo kernel is a heavily patched kernel prepatch (i.e. a pre-lease of 2.4.19); it is often a lot less stable, though in many cases faster.

rommel · Posted: Mon Jun 17, 2002 6:27 am Post subject:

well i just went back to the vanilla kernel and my hdparm reads went right back up....just think its strange...i am using lsr raid 0 on a resierfs / partition...then i was setting up another system with an lsi scsi host adapter using a sym53c1010 chip...and scsi cheetah drives....they were scoring less then ide drives....havent tried them with the vanilla kernel yet though...possibly it could be similar but i am not sure why they would be performing like ata drives when their seek time is like 3.6ms and they run at 15k

DArtagnan · Posted: Mon Jun 17, 2002 6:36 am Post subject:

AutoBot · l33t Joined: 22 Apr 2002 Posts: 968 Location: Usually Out

delta407 · Posted: Tue Jun 18, 2002 6:09 am Post subject:

Well, of course it's not Gentoo-specific, but they are still issues. I'm running a gentoo-sources kernel right now, actually -- nothing has blown up on me yet, and I'm willing to babysit this computer in case something does. But, in a production environment (or anything else where it has to work, or else) I would not run anything other than a vanilla kernel.

Kernel releases get tested, Gentoo kernels don't (except for a couple developers and the Gentoo users as guinea pigs). vanilla-sources is a lot more conservative than gentoo-sources (read README.gentoo and see what you're getting first).

Behold, the beauty of *nix: the choice is yours.

AutoBot · l33t Joined: 22 Apr 2002 Posts: 968 Location: Usually Out

I wouldn't really refer to myself as a "guinea pig" :roll:

_________________
This message self destructed a long time ago.

c_kuzmanic · Posted: Tue Jun 18, 2002 11:40 pm Post subject:

Hmm, I think the term "Heavily patched Kernel" is a relative term, especially when applied to gentoo-sources. All the patches are listed in the gentoo-sources ebuild file, compared to red hat or mandrake kernels the patches are rather modest and few in number.

delta407 · Posted: Tue Jun 18, 2002 11:58 pm Post subject:

Diff it to 2.4.18 and you get... wait, let's see:

c_kuzmanic · Posted: Wed Jun 19, 2002 12:16 am Post subject:

From the gentoo-sources ebuild file:

#What's in this kernel?

#INCLUDED:
# from http://www.kernel.org (ac):
# 2.4.19-pre7-ac2
# removed the software suspend patch; it can be dangerous and
# conflicts with the new ACPI
# from http://oss.sgi.com/projects/xfs:
# SGI XFS 1.1 (Official release code -- the most thoroughly tested)
# from http://www.grsecurity.org:
# grsecurity-1.9.4 (with 2 updates/fixes and a fix for an NVIDIA driver compile problem)
# from http://www.zipworld.com.au/~akpm/linux/schedlat.html:
# 2.4.19-pre7-low-latency
# from http://luxik.cdi.cz/~devik/qos/htb/:
# htb2 (QoS support)
# from http://www.tech9.net/rml/linux:
# preempt-kernel-rml-2.4.19-pre7-ac2-1.patch
# preempt-stats-rml-2.4.19-pre5-ac3-1.patch
# from http://www.infolinux.de/jp10:
# 40_TIOCGDEV.bz2
# 51_loop-jari-2.4.16.0.bz2
# 98_tkparse-4096.bz2
# from http://www.kernel.org (aa):
# 00_3.5G-address-space-4
# from http://www.sourceforge.net/projects/acpi:
# acpi-20020503-2.4.18.diff.gz
# (This allows booting of Toshiba Satellite 5005-S507 "legacy free" laptops)
# Added tweak so that CONFIG_PM is defined only if CONFIG_ACPI or CONFIG_APM is also
# set. Removed CONFIG_PM toggle from configuration.
# from http://www.infolinux.de/jp13:
# 07_jiffies-for-i386 (tweaked so that 1000HZ is the default for x86)
# 54_mmx-init
# 55_p4-xeon
# 56_x86-fast-pte
# 58_acpi-lowerlatency-3
# 59_acpi-pciirq-18
# 60_acpi-y2k-1
# from Blue Lizard <webmaster at dofty.zzn.com>:
# A 2-line patch to enable compatibility with the new SiS 740/961 Athlon chipset
# from http://www.sourceforge.net/projects/evms:
# IBM Enterprise Volume Management System version 1.0.1
# from linux-2.4.18-wolk3.4-rc5-patchset (http://www.sourceforge.net/projects/wolk):
# 102_amd_lvcool.diff
# Allows slightly cooler running AMD Athlon CPUs in systems with VIA mobos
# Additional fixes:
# Marked ACPI option as "DANGEROUS" (because it is and can cause boot ooopses --
# use it only if you need it)
# Marked JFS filesystem option as "FOR TESTING ONLY" (current JFS code seems to
# have super-bad peformance and easily triggerable locks)
# from mjc's collection of patches:
# vfat symlink patch. Creating symlinks on vfat creates Windowsy .lnk files.
# CDDA dma patch -- enables DMA for CD devices#

No sorry, I don't not consider that a heavily patched kernel, and keep in mind that I am comparing it to a red hat or mandrake kernel. I'll say it again, compared to those kernels, gentoo-sources patches are few in number and moreover, their very economical and efficient. The only gentoo patch I ever had problems with was the low-latency patch and those problems would only manifest when I used low-latency and the pre-emptive kernel patches in conjunction. Also, the number of patches applied to a kernel says nothing about the quality/performance of that kernel, and from that perspective gentoo gets full marks.

delta407 · Posted: Wed Jun 19, 2002 12:31 am Post subject:

I don't disagree. Redhat and Mandrake do apply more patches, but they are tested on a very wide range of hardware. gentoo-sources is not. I, for one, am currently running gentoo-sources. An earlier revision was giving me some trouble, so I went to vanilla-sources; but the current one seems stable enough.

Nonetheless, I maintain my stance that gentoo-sources has caused many problems that do not exist with vanilla-sources. Kernel prepatches are historically less stable than kernel releases, and gentoo-sources (as a prepatch) inherits this reputation.

c_kuzmanic · Posted: Wed Jun 19, 2002 12:38 am Post subject: