xx-sources

primero.gentoo · Guru Joined: 23 Dec 2003 Posts: 402

Ok , recompiling right now.

I was wondering about "Intermezzo" file system support? where is it? has been dropped?

I was thinking about tryng it out to substitute NFS ... but i want to use you kernel patchset , so ... what about it?

Bye
_________________
"Linux, the choice of a GNU generation"
==Micro$oft - just say NO==
(L#USER 353039)

xiphux · Posted: Wed Jun 23, 2004 9:28 am Post subject:

Intermezzo's pretty old; it's almost obsolete at this point. It was dropped in 2.6.7.

I normally stick to the standard network filesystems - smbfs for hybrid microsoft/linux networks, and nfsv4 for linux only setups. If you are looking to try an alternate network filesystem, you might be interested in something like cifs or coda.

X-Ryl669 · Tux's lil' helper Joined: 16 Feb 2004 Posts: 135

I don't want to browse for the 8 pages thread. Could you explain what is inside your sources that is not inside dev-sources (in the 2.6.7 version) ?

Thanks.
_________________
Think it twice, think it thrice, you never think enough...

Markus_T · n00b Joined: 09 Feb 2004 Posts: 49 Location: Berlin

Two variables are declared twice in drivers/block/cfq-iosched.c
at line 162-162 (after the comment 'tunables').
Simply deleting the duplicates solves the problem...

xiphux · Posted: Wed Jun 23, 2004 10:33 am Post subject:

*sigh* Laziness at its finest...

The -mm branch
Several process schedulers selectable during configuration:
- Nicksched v30g
- Staircase 7.3
- Single Priority Array (SPA) 1.0
- Entitlement Based Scheduling (EBS) 1.1
- or default
Batch/Isochronous scheduling for Staircase
Locallocks (a better spinlock replacement) by Namesys
The latest Reiser4 snapshot and/or BK pull if necessary
Selectable IO scheduler setup:
- IO niceable CFQ scheduler
- Runtime selectable IO schedulers
- or just the default
Autoregulating swappiness
Configurable number of tty devices
Christopher Walken boot logo
Kernel name in menuconfig/gconfig
Configurable CFLAGS
Gcloop/ucl (compressloop)
BadRAM workaround
High resolution timers
Fast IRQ balancing
Option to use a custom ACPI DSDT from initrd
Layer 7 Netfilter
Supermount-ng
Vesafb refreshrate patch
Linux Infrared Controller (LIRC)
LuFS
Squashfs2
Bootsplash
Win4Lin
Swsusp2
ShFS
Systrace
JFS DMAPI
Morse code panics
Pramfs
CDFS
Delay Accounting
Relayfs
User-mode linux
UML skas mode
SuSE's backported twofish
Low latency
Option to choose from Low Latency / Preempt / Neither
Process aggregates (PAGG) and job containers
Option to hide a bunch of the boot messages (less verbose)
rICMP
Nmap Freak
Device mapper updates (Flakey target, snapshot target, etc)
Bad Block Relocation from EVMS
EVMS 2.6 fix (bd-claim)
Megaraid
m32r processor support
Julian Anastasov's networking stuff (loop, forward_shared, rp_filter_mask, hidden, routes, etc)
Alloc NUMA memory better (per-cpu)
Packet writing
Linux Intrusion Detection System (LIDS)
Submount (SubFS)
frandom (fast random number generators)
Better SpeedStep support
Stuff from -tiny, including
- Configurable alignment cflags
- Kmalloc accounting
- Bootmem auditing
- Configurable jiffies
Flexible mmap from Ingo Molnar
Expandable anonymous shared mappings
Proc entry to display the compiled process scheduler
And whatever fixes/new things may be floating around the LKML

Thanks Markus_T, I'll fix that.

primero.gentoo · Guru Joined: 23 Dec 2003 Posts: 402

xiphux · Posted: Wed Jun 23, 2004 10:50 am Post subject:

I haven't had issues with NFSv4.

The lockup of the client is a feature of NFS - 'soft' versus 'hard' mounting. Hard mode is default - when the server crashes, the process on the client accessing it will hang. It's a security measure. With soft mounting, if the server goes down, the process will just report an error message. This is a little more forgiving of crashes, but it also has a much larger risk of corruption and data loss and stuff; not many programs can handle the error gracefully. Since a lot of places use central nfs servers as actual system drives or import directories for a large amount of computers, hard is better that way. But if it's just relatively trivial filesharing - media files, etc, or if you have a server sharing some filesystem read-only - then it shouldn't be a big deal to use soft mounting.

... or did you have soft mounting on and it locked up anyway?

genstef · Posted: Wed Jun 23, 2004 10:57 am Post subject:

I seem to have some issues with xx:

swsusp2 did not compile
suspend-to-disk did not compile

Maybe you look at http://sourceforge.net/project/showfiles.php?group_id=25964
at the bootom of the page is the latest patch

And there is another problem i did not have with 2.6.7-ck1:

WARNING: /lib/modules/2.6.7-xx1/kernel/sound/pci/snd-intel8x0m.ko needs unknown symbol free

xiphux · Posted: Wed Jun 23, 2004 11:05 am Post subject:

Thanks for the swsusp2 headsup.

Did you try compiling intel8x0 directly into the kernel?

genstef · Posted: Wed Jun 23, 2004 11:43 am Post subject:

No it was as module
<M> Intel i8x0/MX440, SiS 7012; Ali 5455; NForce Audio; AMD768/8111
<M> Intel i8x0/MX440; AMD768/8111 modems (EXPERIMENTAL)

Now my runtime errors:

My sound does not seem to work although the modules and everything is loaded correctly:

xiphux · Posted: Wed Jun 23, 2004 11:57 am Post subject:

I meant does compiling the driver right into the kernel fix the issue.

I put a broken-out tarball up.

genstef · Posted: Wed Jun 23, 2004 12:12 pm Post subject:

I do not like compiling modules into the kernel. It makes unloading on suspend impossible. But reverting your alsa-stuff does the trick for both, what was the use of it?

Btw, when i rebooted that the hd was shut down.

primero.gentoo · Guru Joined: 23 Dec 2003 Posts: 402

scaba · Posted: Wed Jun 23, 2004 1:32 pm Post subject:

yngwin · Posted: Wed Jun 23, 2004 4:17 pm Post subject:

ebuild for 2.6.7-xx1 available at http://www.stijlstek.nl/os/linux/gentoo/ebuilds/sys-kernel/xx-sources/

xiphux, you're doing such a good job, I don't see a need to continue with lokean-sources, xx has everything and more that I would put in...

AstralStorm · Posted: Wed Jun 23, 2004 6:36 pm Post subject:

Hello, this patchset looks very nice... I'm going to try it instead of love-sources. Right now I have a bug to report:
CFQ IOnice is broken: it creates duplicate class members, very easy to fix.

yngwin · Posted: Wed Jun 23, 2004 7:06 pm Post subject:

It suffers from the double-clockspeed bug. I'm going to try the patches mentioned in https://forums.gentoo.org/viewtopic.php?p=1266818#1266818

scaba · Posted: Wed Jun 23, 2004 8:23 pm Post subject:

DaMouse · Apprentice Joined: 18 Jul 2003 Posts: 233

Or you could use the second of those two patches

-DaMouse

[EDIT] Could someone do some benchmarks of the cpu schedulers please [/EDIT]
_________________
I am the dark lord DaMouse I eat souls and wear the dressing gown of evil.

neenee · Veteran Joined: 20 Jul 2003 Posts: 1786

http://cos.evilforums.com/bin/mm1_to_mm1-fixed.diff.bz2

get that to fix 2.6.7-mm1 (timers + rcu compile problem)

PLum · Posted: Thu Jun 24, 2004 9:50 am Post subject:

2.6.7-xx1

kernel/power/pagedir.c: In function `mark_pages_for_pageset2':
kernel/power/pagedir.c:298: error: structure has no member named `nr_active_mapped'
kernel/power/pagedir.c:298: error: structure has no member named `active_mapped_list'
kernel/power/pagedir.c:298: error: structure has no member named `active_mapped_list'
kernel/power/pagedir.c:299: error: structure has no member named `nr_active_unmapped'
kernel/power/pagedir.c:299: error: structure has no member named `active_unmapped_list'
kernel/power/pagedir.c:299: error: structure has no member named `active_unmapped_list'
make[2]: *** [kernel/power/pagedir.o] Error 1
make[1]: *** [kernel/power] Error 2
make: *** [kernel] Error 2

if i disable all the power managment
then the error is

mm/page_alloc.c: In function `free_hot_cold_page':
mm/page_alloc.c:544: error: `software_suspend_state' undeclared (first use in this function)
mm/page_alloc.c:544: error: (Each undeclared identifier is reported only once
mm/page_alloc.c:544: error: for each function it appears in.)
mm/page_alloc.c:544: error: `SOFTWARE_SUSPEND_RUNNING' undeclared (first use in this function)
mm/page_alloc.c: In function `buffered_rmqueue':
mm/page_alloc.c:683: error: `software_suspend_state' undeclared (first use in this function)
mm/page_alloc.c:683: error: `SOFTWARE_SUSPEND_RUNNING' undeclared (first use in this function)
make[1]: *** [mm/page_alloc.o] Error 1
make: *** [mm] Error 2

?

yngwin · Posted: Thu Jun 24, 2004 11:03 am Post subject:

Now that -mm2 is out and a new reiser4 autosnapshot, can we expect a -xx2? Would be nice...

m_el_k · n00b Joined: 20 Dec 2003 Posts: 11

what happened with include/linux/802_11.h ?

BTW, i used the ebuild submitted by yngwin

xiphux · Posted: Thu Jun 24, 2004 12:45 pm Post subject:

Yes you can. I'm doing a couple test compiles of -xx2, but it'll be up soon.

They rearranged a lot of the includes, I think. I know the scsi include arrangement was completely redone. In -mm, they took out 802_11.h. I'm not quite sure what for or where it's now implemented, though...

xiphux · Posted: Thu Jun 24, 2004 2:58 pm Post subject:

Ok, here it is:

http://www.thelightofthesoul.com/linux/patches/xx-sources-2.6.7-xx2/

Not a whole lot of differences here. Updated to -mm2 and the newest reiser4 snapshot (2004.06.23-19.36). I omitted the same rcu patches from -mm that break reiser4. I also traced the patch that was conflicting with nick's vm stuff - there's a patch in -mm that rearranges a bunch of the stuff in mm/vmscan.c, but it doesn't really change any functions. I figured it wasn't worth it to rearrange the code when I could include something that actually did make things better (nickvm). So that's back in. The swsusp2 issue from -xx1 was related to that too; I omitted nick's vm stuff but forgot to revert the changes to swsusp2 that were required to make it work with nickvm in the first place. Swsusp2 is also updated to 2.0.0.88, the latest.

Fixed the duplicate variable issue in cfqionice. Dropped the alsa malloc patches - they were mostly for cleanliness (it fixed alsa to use the generic systemwide alloc functions as opposed to its own). But since they ended up causing issues, it wasn't really a big deal to drop them. I dropped flexible mmap since it's in -mm2 now. I also added out-of-order net packet scheduling.

As always, goes on top of vanilla 2.6.7.