4 kb stacks

[noselasd]

Fedora Core 2 (to be released in a month) will have only 4 kb stacks in kernel land.
(So will the vanilla 2.6 kernel soon)
This causes the nVidia module not to work, presumably it uses more than 4kb
on the stack.
Will/can we see an updated driver soon which handles this ?

[pzgren]

Originally Posted by noselasd

This causes the nVidia module not to work, presumably it uses more than 4kb
on the stack.

[marcus@redtuxi marcus]$ cat /etc/fedora-release
Fedora Core release 1.91 (FC2)

[marcus@redtuxi marcus]$ uname -ro
2.6.6-rc1 GNU/Linux

[marcus@redtuxi marcus]$ cat /proc/driver/nvidia/version
NVRM version: NVIDIA Linux x86 NVIDIA Kernel Module 1.0-5341...
GCC version: gcc-Version 3.3.3 20040412 (Red Hat Linux 3.3.3-7)

[marcus@redtuxi marcus]$ cat /usr/src/linux-2.6.6-rc1/.config | grep 4K
CONFIG_4KSTACKS=y

And if there is someone with problems with that feature outta space: You have the source, Luke.

Marcus

[noselasd]

Originally Posted by pzgren

[marcus@redtuxi marcus]$ cat /etc/fedora-release
Fedora Core release 1.91 (FC2)

[marcus@redtuxi marcus]$ uname -ro
2.6.6-rc1 GNU/Linux

[marcus@redtuxi marcus]$ cat /proc/driver/nvidia/version
NVRM version: NVIDIA Linux x86 NVIDIA Kernel Module 1.0-5341...
GCC version: gcc-Version 3.3.3 20040412 (Red Hat Linux 3.3.3-7)

[marcus@redtuxi marcus]$ cat /usr/src/linux-2.6.6-rc1/.config | grep 4K
CONFIG_4KSTACKS=y

And if there is someone with problems with that feature outta space: You have the source, Luke.

Marcus

No I only have some source for the nVidia driver
Where did you get 5341 btw ? I only have 1.0-5336 , and havn't seen any newer
anywhere..

[pzgren]

http://www.nforcershq.com/forum/viewtopic.php?t=44256

Marcus

[ckloiber]

Originally Posted by pzgren

http://www.nforcershq.com/forum/viewtopic.php?t=44256

Marcus

Fedora Core 2, test2 + latest rawhide. kernel-2.6.5-1-327.i686.rpm:

Apr 19 23:54:46 galileo kernel: Unable to handle kernel NULL pointer dereference at virtual address 00000270
Apr 19 23:54:46 galileo kernel: printing eip:
Apr 19 23:54:46 galileo kernel: 42f122c2
Apr 19 23:54:46 galileo kernel: *pde = 00000000
Apr 19 23:54:46 galileo kernel: Oops: 0002 [#1]
Apr 19 23:54:46 galileo kernel: CPU: 0
Apr 19 23:54:46 galileo kernel: EIP: 0060:[<42f122c2>] Tainted: P
Apr 19 23:54:46 galileo kernel: EFLAGS: 00010216 (2.6.5-1.327)
Apr 19 23:54:46 galileo kernel: EIP is at os_alloc_mem+0x63/0x81 [nvidia]
Apr 19 23:54:46 galileo kernel: eax: 00000000 ebx: 00000270 ecx: 003a8000 edx: 00000039
Apr 19 23:54:46 galileo kernel: esi: 34ce5f68 edi: 42f5f1ac ebp: 34ce5f3c esp: 34ce5f28
Apr 19 23:54:46 galileo kernel: ds: 007b es: 007b ss: 0068
Apr 19 23:54:46 galileo kernel: Process insmod (pid: 6826, threadinfo=34ce5000 task=356eccd0)
Apr 19 23:54:46 galileo kernel: Stack: 00000270 00000000 42da6b60 34ce5f68 00000270 34ce5f70 42d9c2be 00000000
Apr 19 23:54:46 galileo kernel: 34ce5f68 00000270 00000000 00000000 34ce5f6c 00000001 42f87580 42f88360
Apr 19 23:54:46 galileo kernel: 41d00900 00000000 34ce5f88 42d9baa9 42f5f1ac 00000000 00000270 00000001
Apr 19 23:54:46 galileo kernel: Call Trace:
Apr 19 23:54:46 galileo kernel: [<42da6b60>] _nv001309rm+0x10/0x28 [nvidia]
Apr 19 23:54:46 galileo kernel: [<42d9c2be>] _nv001355rm+0x32/0x9c [nvidia]
Apr 19 23:54:46 galileo kernel: [<42d9baa9>] _nv000905rm+0x15/0x50 [nvidia]
Apr 19 23:54:46 galileo kernel: [<42daae58>] rm_init_rm+0x8/0xc [nvidia]
Apr 19 23:54:46 galileo kernel: [<42858155>] nvidia_init_module+0x155/0x19d [nvidia]
Apr 19 23:54:46 galileo kernel: [<02136583>] sys_init_module+0x1ef/0x2c6
Apr 19 23:54:46 galileo kernel:
Apr 19 23:54:46 galileo kernel: Code: 89 03 83 3b 00 75 07 b8 06 01 00 01 eb 05 b8 00 00 00 00 8b

Interestingly, it's a "non fatal" OOPS, in that the machine doesn't lock up completely.
Well, I , and will wait for a real update. Sucks not being able to play my UT2004.

[pzgren]

Originally Posted by ckloiber

Well, I , and will wait for a real update. Sucks not being able to play my UT2004.

Are my postings in Japanese? As I said cowboy,

CONFIG_REGPARM=y

It breaks it. And not...

Look in your /boot/config-2.6.5-1.327

Pull the source and disable it, 10 min and you can play UT all night long!



Marcus

[LordMorgul]

Originally Posted by pzgren

CONFIG_REGPARM=y
<snip>
Pull the source and disable it, 10 min and you can play UT all night long!

Not quite so, removal of REGPARM does not correct the problem for all users, myself included -- and beside that Zander's patch to 5341 should workaround REGPARM to leave it enabled.

I am intrigued by your system working with 4KSTACKS as it has never worked for one of my builds (and I have built somewhere around 24 separate versions / configs since 4KSTACKS showed up in FC2 Rawhide kernels). REGPARM has worked (with said patch). Clearly this is not a cut and dry situation on either side of these options. Honestly, I would like to understand more.

What chipsets do you have, what video card / bios?

$-> cat /proc/driver/nvidia/cards/0
Model: GeForce4 Ti 4400
IRQ: 7
Video BIOS: 04.25.00.22.50
Card Type: AGP

01:00.0 VGA compatible controller: nVidia Corporation NV25 [GeForce4 Ti 4400] (rev a2) (prog-if 00 [VGA])
Subsystem: VISIONTEK: Unknown device 003c
Flags: bus master, VGA palette snoop, 66Mhz, medium devsel, latency 248, IRQ 7

00:01.0 PCI bridge: Intel Corp. 82850 850 (Tehama) Chipset AGP Bridge (rev 04) (prog-if 00 [Normal decode])
Flags: bus master, 66Mhz, fast devsel, latency 64
Bus: primary=00, secondary=01, subordinate=01, sec-latency=64

[pzgren]

ASUS 9570 FX 5700 256MB with an Abit NF7-S V2 nForce2.

[marcus@redtuxi marcus]$ cat /proc/driver/nvidia/cards/0
Model: GeForce FX 5700
IRQ: 12
Video BIOS: 04.36.20.21.00
Card Type: AGP

Ive straight played ET for ~2hours. All stable and fine. And I think many have CONFIG_REGPARM compiled in and do not know that...
Have you IRQ sharing with IRQ7 for your GFX card? parport, USB?

Regards,
Marcus

[LordMorgul]

Originally Posted by pzgren

Have you IRQ sharing with IRQ7 for your GFX card? parport, USB?

No, AGP is isolated on IRQ 7 (lspci -v), bios IRQ 11.