| View previous topic :: View next topic |
| Author |
Message |
handsomepete
Guru
Joined: 21 Apr 2002
Posts: 548
Location: Kansas City, MO
|
 Posted: Mon Jul 07, 2003 1:29 am Post subject: |
 |
|
| Yikes. It's weird posting from lynx. Everything was screwed when I got home, including the i/o bug (2.5.74-mm1 w/ ACPI=OFF (yep) - tar was blocked, IO-Wait in top was ~94%) and, for no apparent reason, mozilla stopped resolving forums (and bugs).gentoo.org properly. I'm in the midst of compiling .74-mm2 with the patch unapplied and debugging turned on (not that I really know what to do with it - hints welcome). |
|
| Back to top |
|
 |
sabre66
n00b
Joined: 23 Jun 2002
Posts: 52
|
| Posted: Mon Jul 07, 2003 2:04 am Post subject: |
|
|
Well , been working with 2.5.74-mm2 with the highpcm patch applied just to get nvidia-kernel to work . Any website I first go to apon a reboot and restart of mozilla takes forever ( resolving blah blah blah ) not a problem with a kernel<2.5.74 . Mozilla seems to have lost the ability to retain bookmarks and homepage info ( don't know if its kernel related or not )Compiling kernel 2.5.74-xx takes on my pc about 30sec longer as opposed to anything lesser (bzImages all seem to be same size), glxgears 10% slower , hdparms the same. Again I'm really not sure if it's all kernel related , but ill drop down to 2.5.73-mm3 for now it really seems to run so much nicer on my pc with less hassel.
(BTW XP3000 ,nf2 ,msi g4 ti 4200 ,1G muskin ram 2x512 matched)
Oh well awaiting 2.6 , should be real nice.....
_________________
By gazing upon the image of the McMartyr, you too have become *permanently* and *irrevokably* redeemed. You are now officially a Melchizedek Priest of McChurch. |
|
| Back to top |
|
|
bssteph
l33t
Joined: 26 Feb 2003
Posts: 652
Location: Wisconsin
|
| Posted: Mon Jul 07, 2003 6:04 am Post subject: |
|
|
Not that this thread isn't already wraught with odd bugs and patch discussions, but Con has released an O3int, which patches against mm2.
Get 'er here.
http://members.optusnet.com.au/ckolivas/kernel/2.5/
I think it's a sizable improvement over O2int. I'm using O3int + granularity + min_timeslice = max_timeslice = 10, and things are smooth for the most part.
XMMS _usually_ keeps up when I'm doing Nautilus browses. Sometimes when I open XMMS, start a song, and immediately go to browsing, it will stop for couple fractions of a second. Other times not. Kind of peculiar.
Moving windows around is nice, where it was just so-so before. Dragging Nautilus over this very Galeon window leaves a lot of artifacts in the Galeon window though.
XMMS window dragging getting behind the mouse is still possible, but seemingly not nearly as likely to happen. And it doesn't immediately start happening again after letting go and starting a slow drag.
Haven't seen any terrible chop due to web rendering, but I haven't browsed around much yet.
Woo woo. Things are certainly getting better. |
|
| Back to top |
|
|
Exner
Tux's lil' helper
Joined: 08 Apr 2003
Posts: 128
Location: Melbourne, Australia
|
| Posted: Mon Jul 07, 2003 7:42 am Post subject: two cents |
|
|
I want to add my curious experience to the interesting tests other people have been posting. Warning - offtopic of 2.5.74.
I have used 2.5.68-mm2, 2.5.69-mm3. Then I used 2.5.73-mm2 for 7 days. All along my desktop usability seemed to be getting worse with windows slow to redraw and multi media skips.
2 days ago I went back to 2.5.73-mm2 and recompiled after removing the HZ-100 patch, and changing -march=athlon-tbird to the default -march=athlon. This has solved my desktop slowness.
I solved my multimedia skips by turning off icon previews in nautilus. I think nautilus should launch its gst-* preview apps with a suitable nice value.
_________________
- Exner (Antony Suter) |
|
| Back to top |
|
|
bssteph
l33t
Joined: 26 Feb 2003
Posts: 652
Location: Wisconsin
|
| Posted: Mon Jul 07, 2003 4:11 pm Post subject: |
|
|
I got the hanging process bug last night. At least for me, it has nothing to do with heavy I/O.
I locked in the ./configure stage of a libbonobo install. I believe the exact line was checking for a C preprocessor or some such... It'd just hanged. Killable, but totally hanged. I say I don't think this has to do with I/O because this was from console, logged in as root, and aside from some daemons, nothing else was running. After killing it, I tried to log in from another vt, which hanged after entering my password.
_I've had this bug before._
I was getting it on both of my machines running the mm patchset about two weeks ago. I don't remember what the mm versions were, but I'd thought they'd recently gone away, but I guess not. My box had an uptime of probably somewhere near 6 hours when it had the dying process problem. This machine's been up longer but I haven't reproduced here. I'm going to try to compile something, and then my plan is to patch a .74 kernel with just the Oint patches and granularity, and see what happens.
Oh, machines:
Box #1 (desktop): 1.4 GHz Athlon, 384 MB RAM, one SCSI, one IDE hard drive. VIA82CXXX chipset. GeForce 3, nvidia driver.
Box #2 (laptop): 1.6 GHz Pentium 4, 256 MB RAM, one IDE hard drive. Intel PIIXn chipset. Radeon Mobility, kernel's radeon driver.
Both have been booting using ACPI for IRQ routing. ACPI compiled in, as is PnP support and the PnP device name database. Both using preempt, MTRRs, and a bunch of standard fare kernel opts. I fear this is something beyond a simple driver issue. |
|
| Back to top |
|
|
bssteph
l33t
Joined: 26 Feb 2003
Posts: 652
Location: Wisconsin
|
| Posted: Mon Jul 07, 2003 4:54 pm Post subject: |
|
|
Hmmph. I can't reproduce here. And I used to be able to. Before it didn't take much, either. Start a compile, open X-Chat, maybe do some browsing in Galeon, and that'd take care of it. Nothing now. X-Chat open, doing lots of browsing in Nautilus, doing things in Galeon, opening random games, playing stuff in XMMS and then quickly swapping to a video (card can't do mixing), and all this while compiling. Everything kept right up.
This got me to thinking. Maybe there's other forces at work here. I diff'ed my two configs and didn't come up with much different. The different drivers of course, and a couple other options either related to those or probably not relevant given the circumstances. A couple of things caught my eye, though.
Here's my first salvo at randomly trying to destroy this bug.
Specifically for people that are getting the problems: Anyone NOT using "Enhanced real time clock support"? |
|
| Back to top |
|
|
Safrax
Guru
Joined: 23 Apr 2002
Posts: 422
|
| Posted: Mon Jul 07, 2003 4:59 pm Post subject: |
|
|
| bssteph wrote: | | Specifically for people that are getting the problems: Anyone NOT using "Enhanced real time clock support"? |
I use it on all computers I have been able to duplicate this bug on. Problems seem to occur most often when using tar/gzip/bzip2 for compression or decompression. The process bug also seems to manifest when the system needs to swap things out, but of course im not a programmer, nor do I understand the workings of the kernel beyond simplistic concepts, so I am probably wrong. |
|
| Back to top |
|
|
bssteph
l33t
Joined: 26 Feb 2003
Posts: 652
Location: Wisconsin
|
| Posted: Mon Jul 07, 2003 5:19 pm Post subject: |
|
|
| Safrax wrote: | | I use it on all computers I have been able to duplicate this bug on. Problems seem to occur most often when using tar/gzip/bzip2 for compression or decompression. The process bug also seems to manifest when the system needs to swap things out, but of course im not a programmer, nor do I understand the workings of the kernel beyond simplistic concepts, so I am probably wrong. |
Neither do I, really. I'm just trying stuff at random. 
I asked about enhanced RTC because right now I can't get any lockage on my laptop. It's using RTC emulation instead of enhanced RTC (which the desktop uses). That was a recent change, meaning that two weeks ago I was using enhanced. There was a portion where my desktop was using emulation as well, when I was trying random stuff for my sound card. And that could very well have coincided with the time when I didn't get the lockup bug. So yeah. I'm recompiling the desktop's kernel right now, so I'll report back later (although it seems to take a lot of time before these bugs show up).
ramdisk also caught my eye, because this machine uses framebuffer. So I enabled that as well, in vain attempts to get something worked out. I'm guessing some of you are using framebuffer though. |
|
| Back to top |
|
|
Safrax
Guru
Joined: 23 Apr 2002
Posts: 422
|
| Posted: Mon Jul 07, 2003 5:38 pm Post subject: |
|
|
| Hmm. Okay, I'll see what happens when the desktop runs on rtc emulation instead... |
|
| Back to top |
|
|
R-Type
n00b
Joined: 22 Sep 2002
Posts: 62
|
| Posted: Mon Jul 07, 2003 5:57 pm Post subject: |
|
|
after receiving "Unknown symbo pmd_offset" errors, I spent sometime digging around the mm sections of the kernel. I THINK pmd_offset_kernel() is the function nvidia.o should be calling. Try editing nv.c and changing pmd_offset to pmd_offset_kernel. Then build the module. Its been working fine for me so far. I've stress-tested it with quake3, q2 and some opengl screensavers.
I'd post a patch, but I have nowhere to upload it atm. |
|
| Back to top |
|
|
Safrax
Guru
Joined: 23 Apr 2002
Posts: 422
|
| Posted: Mon Jul 07, 2003 6:08 pm Post subject: |
|
|
| R-Type wrote: | after receiving "Unknown symbo pmd_offset" errors, I spent sometime digging around the mm sections of the kernel. I THINK pmd_offset_kernel() is the function nvidia.o should be calling. Try editing nv.c and changing pmd_offset to pmd_offset_kernel. Then build the module. Its been working fine for me so far. I've stress-tested it with quake3, q2 and some opengl screensavers.
I'd post a patch, but I have nowhere to upload it atm. |
I just remove the highpmd patch  but your solution seems to be the right one. |
|
| Back to top |
|
|
bssteph
l33t
Joined: 26 Feb 2003
Posts: 652
Location: Wisconsin
|
| Posted: Mon Jul 07, 2003 6:09 pm Post subject: |
|
|
| R-Type wrote: | after receiving "Unknown symbo pmd_offset" errors, I spent sometime digging around the mm sections of the kernel. I THINK pmd_offset_kernel() is the function nvidia.o should be calling. Try editing nv.c and changing pmd_offset to pmd_offset_kernel. Then build the module. Its been working fine for me so far. I've stress-tested it with quake3, q2 and some opengl screensavers.
I'd post a patch, but I have nowhere to upload it atm. |
Oh wow. I'd never even thought to check if the kernel call was in the part of the drivers we can actually see the source for... Will try in a bit, want to finish my gnome upgrade first. |
|
| Back to top |
|
|
handsomepete
Guru
Joined: 21 Apr 2002
Posts: 548
Location: Kansas City, MO
|
| Posted: Mon Jul 07, 2003 6:15 pm Post subject: |
|
|
So far I've had great luck with .74-mm2 (ACPI works again, no locks yet, got the nvidia driver working), but as soon as I tried to untar mozilla nightly sources, tar got blocked by OO.o (I/O wait shot up to 96%, not swapping). I'm losing stability, so I better submit this.
Here's the debugging output of the stall:
| Code: | Debug: sleeping function called from illegal context at mm/page_alloc.c:545
Call Trace:
[<c011b7df>] __might_sleep+0x5f/0x80
[<c013c794>] buffered_rmqueue+0xe4/0x1b0
[<c013cae0>] __alloc_pages+0x280/0x300
[<e4b57c99>] __nvsym00258+0x15/0x1c [nvidia]
[<c0118368>] pte_alloc_one+0x18/0x50
[<c0144f9c>] pte_alloc_map+0x3c/0xc0
[<c0145fc4>] remap_page_range+0xb4/0x1f0
[<e4b43372>] nv_kern_mmap+0x2d2/0x314 [nvidia]
[<c014877b>] do_mmap_pgoff+0x31b/0x6f0
[<c010fb3a>] sys_mmap2+0x9a/0xd0
[<c01093cb>] syscall_call+0x7/0xb |
If anyone's trying to do comparisons:
lspci
.config
R-Type: If that patch works, upload it to bugs.gentoo.org with a request to modify the ebuild. If the change in the kernel sticks, we'll probably want the patch included with the ebuild. |
|
| Back to top |
|
|
bssteph
l33t
Joined: 26 Feb 2003
Posts: 652
Location: Wisconsin
|
|
| Back to top |
|
|
idl
Retired Dev
Joined: 24 Dec 2002
Posts: 1728
Location: Nottingham, UK
|
| Posted: Mon Jul 07, 2003 9:06 pm Post subject: |
|
|
Thanks, works a treat.
_________________
a.k.a port001
Found a bug? Please report it: Gentoo Bugzilla |
|
| Back to top |
|
|
thubble
Tux's lil' helper
Joined: 16 May 2003
Posts: 119
Location: Winnipeg, Canada
|
| Posted: Mon Jul 07, 2003 9:40 pm Post subject: |
|
|
| bssteph wrote: | | Specifically for people that are getting the problems: Anyone NOT using "Enhanced real time clock support"? |
I'm not using enhanced RTC (or generic RTC emulation) and I still get the lockup/dead process bug.  |
|
| Back to top |
|
|
bssteph
l33t
Joined: 26 Feb 2003
Posts: 652
Location: Wisconsin
|
| Posted: Mon Jul 07, 2003 9:42 pm Post subject: |
|
|
Desktop just locked up again. IO-wait was in the 80-90% range, happened while listening to a mp3, emerge syncing, and doing casual browsing. First noticed it when I couldn't switch tabs in Galeon. Quit XMMS and everything was back to normal (after foolishly ctrl-c'ing the emerge sync). Quickly quit gnome and am currently compiling a 2.5.74 kernel with Con's patches applied by hand.
Kernel compile just locked up, too. There goes my random guess.
Laptop just had a lockup too. I need to blow something up.
Last edited by bssteph on Mon Jul 07, 2003 9:58 pm; edited 2 times in total |
|
| Back to top |
|
|
R-Type
n00b
Joined: 22 Sep 2002
Posts: 62
|
| Posted: Mon Jul 07, 2003 9:43 pm Post subject: |
|
|
| Oh well, I tried heheh. I"ll have to try that patch out.. |
|
| Back to top |
|
|
Makaveli[FIN]
n00b
Joined: 31 Oct 2002
Posts: 31
Location: Pori, Finland
|
| Posted: Tue Jul 08, 2003 11:46 am Post subject: |
|
|
Can somebody help me out a little bit? I've downloaded NVIDIA_kernel-1.0-4363.tar.gz (already unpacked) and NVIDIA_kernel-1.0-4363-2.5.diff. Now, how do I exaclty patch NVIDIA_kernel with that diff-file? 
_________________
https://niko.nwps.fi |
|
| Back to top |
|
|
Safrax
Guru
Joined: 23 Apr 2002
Posts: 422
|
| Posted: Tue Jul 08, 2003 1:13 pm Post subject: |
|
|
| Makaveli[FIN] wrote: | | Can somebody help me out a little bit? I've downloaded NVIDIA_kernel-1.0-4363.tar.gz (already unpacked) and NVIDIA_kernel-1.0-4363-2.5.diff. Now, how do I exaclty patch NVIDIA_kernel with that diff-file? |
move the patch to the directory with the drivers in it and do this: | Code: | | patch -p1 <(nameofpatchwithoutparentheses) |
|
|
| Back to top |
|
|
krazo
Tux's lil' helper
Joined: 19 Oct 2002
Posts: 90
|
| Posted: Tue Jul 08, 2003 2:19 pm Post subject: Supermount-ng problems |
|
|
Has anyone gotten supermount-ng to compile on 2.5.74-mm2?
EDIT: whoops had 2.6.74 :\
_________________
Hey--hey how are ya?
Last edited by krazo on Tue Jul 08, 2003 2:41 pm; edited 1 time in total |
|
| Back to top |
|
|
Safrax
Guru
Joined: 23 Apr 2002
Posts: 422
|
| Posted: Tue Jul 08, 2003 2:28 pm Post subject: Re: Supermount-ng problems |
|
|
| krazo wrote: | | Has anyone gotten supermount-ng to compile on 2.6.74-mm2? |
Nope |
|
| Back to top |
|
|
Ansorg
Apprentice
Joined: 13 Jul 2002
Posts: 193
|
| Posted: Tue Jul 08, 2003 7:18 pm Post subject: |
|
|
| Safrax wrote: |
move the patch to the directory with the drivers in it and do this: | Code: | | patch -p1 <(nameofpatchwithoutparentheses) |
|
great, but unfortunately patch generates an error
| Code: | Hunk #1 succeeded at 234 with fuzz 1 (offset 9 lines).
patching file nv.c
Hunk #2 FAILED at 2105.
1 out of 2 hunks FAILED -- saving rejects to file nv.c.rej
|
| Code: | cat nv.c.rej
***************
*** 2105,2115 ****
if (pgd_none(*pg_dir))
goto failed;
- pg_mid_dir = pmd_offset(pg_dir, address);
- if (pmd_none(*pg_mid_dir))
goto failed;
- NV_PTE_OFFSET(address, pg_mid_dir, pte);
if (!pte_present(pte))
goto failed;
--- 2105,2116 ----
if (pgd_none(*pg_dir))
goto failed;
+ NV_PMD_OFFSET(address, pg_dir, pg_mid_dir);
+
+ if (pmd_none(pg_mid_dir))
goto failed;
+ NV_PTE_OFFSET(address, &pg_mid_dir, pte);
if (!pte_present(pte))
goto failed;
|
looked at the nv.c file but don't quite understand the diff stuff and what is supposed to get replaced by the patch ...
any hints?
_________________
thanks
Jens |
|
| Back to top |
|
|
Ansorg
Apprentice
Joined: 13 Jul 2002
Posts: 193
|
| Posted: Tue Jul 08, 2003 8:02 pm Post subject: patch works! |
|
|
ok, figured the patch thingy out (manual editing of nv.c)
and now XFree works again with nvidia drivers
thanks
_________________
thanks
Jens |
|
| Back to top |
|
|
maor
Guru
Joined: 28 Aug 2002
Posts: 323
|
| Posted: Tue Jul 08, 2003 11:22 pm Post subject: |
|
|
| finally the jumpy mouse was fixed now it's just left to compile the nvidia driver and all be great. |
|
| Back to top |
|
|
|
Kernel & Hardware
