acpi problem with nforce motherboards and ethernet
(sorry with my sucking english)
ok, let me know if you know for me to test something with this newer mm kernels.
by the way, yes, I'm experiencing the lockups. Not with heavy io, but almost when I boot and enter X. The system gets completly frozzened and the HD led keps on. When I reboot with a sane kernel, I found several files from my /home directory are deleted or filled with garbage.
Now, I patch each kernel I use with the two patches Andrew sent to me and I'm having no problems.
Actually i'm running 2.6.2-ck1 + nforce-patches and the temperature of the system is 55º while idle, and 631 while compiling (only cpu fan, no case fans). I don't know if it's high (some people reported high temperatures with this patches) but it runs very well this way.
If you want for me to test some patches or something, please drop me a note.
Thanks a lot...
Luis Miguel García
P.S.: by the way, why am I getting strage "arabesque" characters when I reply to your emails? Perhaps something with wrong encoding?
Quoted 14 lines
This is interesting, I will test it myself later on. At one point Len admitted to owning "foreign hardware :p" so maybe this could get resolved.
Personally I haven't tried kernels with newer forcedeth drivers, because I can no longer explicitely set the power state of the NIC to D3. The machine complains about irqs (new debugging code since forcedeth v.20 or so) and will not powerdown.
On another note, have you noticed lockups of your system with heavy io? Think of fsck'ing, burning cdroms, du on large dirs etc? Maybe it is helpful to set up a list of boards and document what works or doesn't work with which kernel. For instance the lockups with heavy io seem to be resolved here if I leave APIC from my kernel.
Quoted 158 lines
Arjen
Quoted 148 lines · On Thu, 5 Feb 2004, [ISO-8859-15] Luis Miguel Garc?a wrote:
Quoted 146 lines
Hi:
Since Andrew Morton picked up latest acpi bk updates, nforce motherboards have problems, mainly with ethernet adapters. Reporters say that with acpi=off, the problm gets fixed, so we think the problem could be acpi. Some more useful info:
Quoted 5 lines · On Tue, 3 Feb 2004, [ISO-8859-1] Luis Miguel Garc?a wrote:
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
When I try to boot with latest mm series (such as actual rc3-mm1 or rc2-mm2), my nforce ethernet device doesn't works. It worked in the past with the forcedeth reverse engineered driver but now it keeps for 30 or more seconds halted (at boot) and then the network device dosn't run.
Quoted 7 lines
Quoted 7 lines
Quoted 3 lines
Quoted 3 lines
Here is the dmesg of rc3-mm1. Do you want for me to test something? Thanks!
P.S.: The ACPI related messages are larger that in rc3.My e100 on an nforce2 won't work in rc3-mm1.
The "acpi=off" boot parameter makes it go.And for the record, I can boot with that kernel and save one dmesg for you if you want. Only send me a request and I'll send it to you.
P.S.: Sent any messages you want directly to me as i'm not subscribed to acpi-devel.Thanks,
Luis Miguel Garc?a
Quoted 73 lines
Quoted 73 lines
Which part of nforce support are you talking about luis?
Quoted 60 lines · Andrew Morton
Quoted 60 lines
Quoted 29 lines · Luis Miguel Garc?a
Quoted 29 lines
Quoted 29 lines
Quoted 29 lines
Quoted 15 lines · Andrew Morton
Quoted 15 lines
ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.2/2.6.2-mm1/
- Merged some page reclaim fixes from Nick and Nikita. These yield some
performance improvements in low memory and heavy paging situations.Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Andrew, do you know if this acpi pull down has nforce support fixed?
It doesn't appear that way.
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Or perhaps it's even unnotified to the acpi team?
I do not know. Sending them a bugzilla ID would help, if such a thing exists.
Quoted 15 lines
Quoted 15 lines
Quoted 11 lines
Quoted 11 lines
-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
Acpi-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/acpi-devel-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
Acpi-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/acpi-devel
-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
I have the same problem. I "solved" it a while ago by mucking with the AGP stuff. IIRC, it was turning off AGP fast writes or 8x or something similar in cmos. Went from incredibly broken to stable instantly. I'll check my cmos settings in a bit and refresh my memory.
What patches are you using?
p.s. reiserfs is pretty useless against crashes. hundreds of files around the harddrive get messed up everytime the machine locks up...files that haven't been used in weeks, as well as files opened read-only, i.e. libraries.
Quoted 210 lines · Luis Miguel García
(sorry with my sucking english)
ok, let me know if you know for me to test something with this newer mm kernels.
by the way, yes, I'm experiencing the lockups. Not with heavy io, but almost when I boot and enter X. The system gets completly frozzened and the HD led keps on. When I reboot with a sane kernel, I found several files from my /home directory are deleted or filled with garbage.
Now, I patch each kernel I use with the two patches Andrew sent to me and I'm having no problems.
Actually i'm running 2.6.2-ck1 + nforce-patches and the temperature of the system is 55º while idle, and 631 while compiling (only cpu fan, no case fans). I don't know if it's high (some people reported high temperatures with this patches) but it runs very well this way.
If you want for me to test some patches or something, please drop me a note.
Thanks a lot...
Luis Miguel García
P.S.: by the way, why am I getting strage "arabesque" characters when I reply to your emails? Perhaps something with wrong encoding?Quoted 14 lines
This is interesting, I will test it myself later on. At one point Len admitted to owning "foreign hardware :p" so maybe this could get resolved.
Personally I haven't tried kernels with newer forcedeth drivers, because I can no longer explicitely set the power state of the NIC to D3. The machine complains about irqs (new debugging code since forcedeth v.20 or so) and will not powerdown.
On another note, have you noticed lockups of your system with heavy io? Think of fsck'ing, burning cdroms, du on large dirs etc? Maybe it is helpful to set up a list of boards and document what works or doesn't work with which kernel. For instance the lockups with heavy io seem to be resolved here if I leave APIC from my kernel.
Quoted 178 lines
Arjen
Quoted 167 lines · On Thu, 5 Feb 2004, [ISO-8859-15] Luis Miguel Garc?a wrote:
Quoted 166 lines
Hi:
Since Andrew Morton picked up latest acpi bk updates, nforce motherboards have problems, mainly with ethernet adapters. Reporters say that with acpi=off, the problm gets fixed, so we think the problem could be acpi. Some more useful info:
Quoted 9 lines · On Tue, 3 Feb 2004, [ISO-8859-1] Luis Miguel Garc?a wrote:
Quoted 9 lines
Quoted 1 line
Quoted 1 line
When I try to boot with latest mm series (such as actual
rc3-mm1 or
Quoted 1 line
Quoted 1 line
rc2-mm2), my nforce ethernet device doesn't works. It worked
in the past
Quoted 1 line
Quoted 1 line
with the forcedeth reverse engineered driver but now it keeps
for 30 or
Quoted 1 line
Quoted 1 line
more seconds halted (at boot) and then the network device
dosn't run.
Quoted 1 line
Quoted 1 line
Quoted 8 lines
Quoted 8 lines
Quoted 1 line
Quoted 1 line
Here is the dmesg of rc3-mm1. Do you want for me to test
something? Thanks!
Quoted 2 lines
Quoted 2 lines
P.S.: The ACPI related messages are larger that in rc3.My e100 on an nforce2 won't work in rc3-mm1.
The "acpi=off" boot parameter makes it go.And for the record, I can boot with that kernel and save one dmesg for you if you want. Only send me a request and I'll send it to you.
P.S.: Sent any messages you want directly to me as i'm not subscribed to acpi-devel.Thanks,
Luis Miguel Garc?a
Quoted 88 lines
Quoted 88 lines
Which part of nforce support are you talking about luis?
Quoted 63 lines · Andrew Morton
Quoted 56 lines
Quoted 38 lines · Luis Miguel Garc?a
Quoted 38 lines
Quoted 38 lines
Quoted 38 lines
Quoted 10 lines · Andrew Morton
ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.2/2.6.2-mm1/
Quoted 4 lines
- Merged some page reclaim fixes from Nick and
Nikita. These yield some
Quoted 1 line
performance improvements in low memory and
heavy paging situations.
Quoted 2 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Andrew, do you know if this acpi pull down has nforce
support fixed?
It doesn't appear that way.
Quoted 8 lines
Quoted 8 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Or perhaps it's even unnotified to the acpi team?
I do not know. Sending them a bugzilla ID would help, if such
a thing exists.
Quoted 3 lines
Quoted 3 lines
Quoted 15 lines
Quoted 15 lines
Quoted 11 lines
Quoted 11 lines
-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
Acpi-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/acpi-devel-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
Acpi-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/acpi-devel
Quoted 5 lines
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
Quoted 3 lines · David Ford
I have the same problem. I "solved" it a while ago by mucking with the AGP stuff. IIRC, it was turning off AGP fast writes or 8x or something similar in cmos. Went from incredibly broken to stable instantly. I'll check my cmos settings in a bit and refresh my memory.
What patches are you using?
I'm using nforce2-apic.patch and nforce2-disconnect-quirk.patch that Andrew Morton have sent to me. I think they have been included in previous mm kernels but now are droped because they caused some temperature problems for some people with no nforce motherboards.
If you want it, I can send it to you.
By the way, is anyone involved in solving the IO-APIC thing in nforce motherboards? Anyone trying a different approach? Anyone contacting nvidia about this problem?
Quoted 1 line
p.s. reiserfs is pretty useless against crashes. hundreds of files around the harddrive get messed up everytime the machine locks up...files that haven't been used in weeks, as well as files opened read-only, i.e. libraries.
I don't think this could be a reiserfs problem, because all the computer gets frezzed and locked, but perhaps someone at Namesys can realize why files gets deleted all over the hard disc.
Thanks.
Luis Miguel Garcia
Quoted 131 lines
p.s. reiserfs is pretty useless against crashes. hundreds of files around the harddrive get messed up everytime the machine locks up...files that haven't been used in weeks, as well as files opened read-only, i.e. libraries.
Quoted 126 lines · Luis Miguel García
(sorry with my sucking english)
ok, let me know if you know for me to test something with this newer mm kernels.
by the way, yes, I'm experiencing the lockups. Not with heavy io, but almost when I boot and enter X. The system gets completly frozzened and the HD led keps on. When I reboot with a sane kernel, I found several files from my /home directory are deleted or filled with garbage.
Now, I patch each kernel I use with the two patches Andrew sent to me and I'm having no problems.
Actually i'm running 2.6.2-ck1 + nforce-patches and the temperature of the system is 55º while idle, and 631 while compiling (only cpu fan, no case fans). I don't know if it's high (some people reported high temperatures with this patches) but it runs very well this way.
If you want for me to test some patches or something, please drop me a note.
Thanks a lot...
Luis Miguel García
P.S.: by the way, why am I getting strage "arabesque" characters when I reply to your emails? Perhaps something with wrong encoding?Quoted 108 lines
This is interesting, I will test it myself later on. At one point Len admitted to owning "foreign hardware :p" so maybe this could get resolved.
Personally I haven't tried kernels with newer forcedeth drivers, because I can no longer explicitely set the power state of the NIC to D3. The machine complains about irqs (new debugging code since forcedeth v.20 or so) and will not powerdown.
On another note, have you noticed lockups of your system with heavy io? Think of fsck'ing, burning cdroms, du on large dirs etc? Maybe it is helpful to set up a list of boards and document what works or doesn't work with which kernel. For instance the lockups with heavy io seem to be resolved here if I leave APIC from my kernel.
Arjen
Quoted 88 lines · On Thu, 5 Feb 2004, [ISO-8859-15] Luis Miguel Garc?a wrote:
Quoted 88 lines
Hi:
Since Andrew Morton picked up latest acpi bk updates, nforce motherboards have problems, mainly with ethernet adapters. Reporters say that with acpi=off, the problm gets fixed, so we think the problem could be acpi. Some more useful info:
Quoted 5 lines · On Tue, 3 Feb 2004, [ISO-8859-1] Luis Miguel Garc?a wrote:
Quoted 5 lines
Quoted 1 line
Quoted 1 line
When I try to boot with latest mm series (such as actual
rc3-mm1 or
Quoted 1 line
Quoted 1 line
rc2-mm2), my nforce ethernet device doesn't works. It worked
in the past
Quoted 1 line
Quoted 1 line
with the forcedeth reverse engineered driver but now it keeps
Quoted 10 lines
Quoted 8 lines
for 30 or
Quoted 1 line
Quoted 1 line
more seconds halted (at boot) and then the network device
dosn't run.
Quoted 2 lines
Quoted 2 lines
Here is the dmesg of rc3-mm1. Do you want for me to test
something? Thanks!
Quoted 2 lines
Quoted 2 lines
P.S.: The ACPI related messages are larger that in rc3.My e100 on an nforce2 won't work in rc3-mm1.
The "acpi=off" boot parameter makes it go.And for the record, I can boot with that kernel and save one dmesg for you if you want. Only send me a request and I'll send it to you.
P.S.: Sent any messages you want directly to me as i'm not subscribed to acpi-devel.Thanks,
Luis Miguel Garc?a
Quoted 46 lines
Quoted 45 lines
Which part of nforce support are you talking about luis?
Quoted 8 lines · Andrew Morton
Quoted 8 lines
Quoted 5 lines · Luis Miguel Garc?a
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Andrew Morton wrote:
ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.2/2.6.2-mm1/
Quoted 21 lines
Quoted 18 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
- Merged some page reclaim fixes from Nick and
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Nikita. These yield some
performance improvements in low memory and
heavy paging situations.Andrew, do you know if this acpi pull down has nforce
support fixed?
It doesn't appear that way.
Quoted 5 lines
Quoted 5 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Or perhaps it's even unnotified to the acpi team?
I do not know. Sending them a bugzilla ID would help, if
such a thing exists.
-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
Quoted 8 lines · Luis Miguel García
Quoted 5 lines · David Ford
I have the same problem. I "solved" it a while ago by mucking with the AGP stuff. IIRC, it was turning off AGP fast writes or 8x or something similar in cmos. Went from incredibly broken to stable instantly. I'll check my cmos settings in a bit and refresh my memory.
Quoted 7 lines
Quoted 1 line
What patches are you using?
I'm using nforce2-apic.patch and nforce2-disconnect-quirk.patch that Andrew Morton have sent to me. I think they have been included in previous mm kernels but now are droped because they caused some temperature problems for some people with no nforce motherboards.
Yes, the patch which disables "Halt Disconnect and Stop Grant Disconnect" apparently causes the CPU to run hot.
Quoted 3 lines
By the way, is anyone involved in solving the IO-APIC thing in nforce motherboards? Anyone trying a different approach? Anyone contacting nvidia about this problem?
As far as I know, we're dead in the water on these problems.
Here's one:
[x86] do not wrongly override mp_ExtINT IRQ
From: Mathieu <[email protected]>.
With this patch timer IRQ0 is correctly set to IO-APIC-edge (not XT-PIC) on nForce2 boards when using APIC and ACPI.
arch/i386/kernel/mpparse.c | 3 ++- 1 files changed, 2 insertions(+), 1 deletion(-)
arch/i386/kernel/mpparse.c renamed from arch/i386/kernel/mpparse.c~nforce2-apic +2 -1
@@ -962,7 +962,8 @@ void __init mp_override_legacy_irq (*/ for (i =0 ; i < mp_irq_entries; i++) {if ((mp_irqs[i].mpc_dstapic == intsrc.mpc_dstapic) - && (mp_irqs[i].mpc_srcbusirq == intsrc.mpc_srcbusirq)) { + && (mp_irqs[i].mpc_srcbusirq == intsrc.mpc_srcbusirq) + && (mp_irqs[i].mpc_irqtype == intsrc.mpc_irqtype)) { mp_irqs[i] = intsrc; found =1 ;break ;
_
Here's the other:
From: Bartlomiej Zolnierkiewicz <[email protected]>
[PATCH] fix lockups with APIC support on nForce2
Add PCI quirk to disable Halt Disconnect and Stop Grant Disconnect (based on athcool program by Osamu Kayasono).
(Mark McPherson <[email protected]> reports that this patch causes his CPU temperature to skyrocket).
25-akpm/arch/i386/pci/fixup.c | 17 +++++++++++++++++ 1 files changed, 17 insertions(+)
arch/i386/pci/fixup.c renamed from arch/i386/pci/fixup.c~nforce2-disconnect-quirk +17 -0
@@ -187,6 +187,22 @@ static void __devinit pci_fixup_transpar dev->transparent =1 ; } +/* +* Halt Disconnect and Stop Grant Disconnect (bit 4 at offset 0x6F) +* must be disabled when APIC is used (or lockups will happen). +*/ +static void __devinit pci_fixup_nforce2_disconnect (struct pci_dev *d) +{ +u8 t; + +pci_read_config_byte (d,0x6F , &t); +if (t &0x10 ) { +printk (KERN_INFO "PCI: disabling nForce2 Halt Disconnect" +" and Stop Grant Disconnect\n" ); +pci_write_config_byte (d,0x6F , (t &0xef )); + } +} +struct pci_fixup pcibios_fixups[] = { {PCI_FIXUP_HEADER ,PCI_VENDOR_ID_INTEL ,PCI_DEVICE_ID_INTEL_82451NX , pci_fixup_i450nx }, {PCI_FIXUP_HEADER ,PCI_VENDOR_ID_INTEL ,PCI_DEVICE_ID_INTEL_82454GX , pci_fixup_i450gx },
@@ -205,5 +221,6 @@ struct pci_fixup pcibios_fixups[] = { { PCI_FIXUP_HEADER, PCI_VENDOR_ID_VIA, PCI_DEVICE_ID_VIA_8367_0, pci_fixup_via_northbridge_bug }, { PCI_FIXUP_HEADER, PCI_VENDOR_ID_NCR, PCI_DEVICE_ID_NCR_53C810, pci_fixup_ncr53c810 }, { PCI_FIXUP_HEADER, PCI_VENDOR_ID_INTEL, PCI_ANY_ID, pci_fixup_transparent_bridge }, + { PCI_FIXUP_HEADER, PCI_VENDOR_ID_NVIDIA, PCI_DEVICE_ID_NVIDIA_NFORCE2, pci_fixup_nforce2_disconnect }, { 0 } };
_
Quoted 8 lines
Quoted 4 lines
Quoted 1 line
What patches are you using?
I'm using nforce2-apic.patch and nforce2-disconnect-quirk.patch that Andrew Morton have sent to me. I think they have been included in previous mm kernels but now are droped because they caused some temperature problems for some people with no nforce motherboards.
Yes, the patch which disables "Halt Disconnect and Stop Grant Disconnect" apparently causes the CPU to run hot.
Well, my temp is about 6-9°C hotter in idle mode, but with current kernels APIC (and no CPU diconnect) seems to be more stable than PIC and CPU Disconnect with my nforce. I at least don't have problems with forcedeth driver and current acpi.
Quoted 4 lines
Quoted 1 line
By the way, is anyone involved in solving the IO-APIC thing in nforce motherboards? Anyone trying a different approach? Anyone contacting nvidia about this problem?
As far as I know, we're dead in the water on these problems.
I remember Jesse Allan sending some lspci outputs to Bart to investigate the BIOS fix. I wonder what has happened to it. I tried to do a comparison with my values, but due to lack of some smart tools to show what is common and what is not (diff is not very suitable) it is very hard, esp if you have no knowledge about the actual meanings of those parameters.
Prakash
-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
Quoted 17 lines · Andrew Morton
Quoted 8 lines · Luis Miguel García
Quoted 5 lines · David Ford
I have the same problem. I "solved" it a while ago by mucking with the AGP stuff. IIRC, it was turning off AGP fast writes or 8x or something similar in cmos. Went from incredibly broken to stable instantly. I'll check my cmos settings in a bit and refresh my memory.
Quoted 7 lines
Quoted 1 line
What patches are you using?
I'm using nforce2-apic.patch and nforce2-disconnect-quirk.patch that Andrew Morton have sent to me. I think they have been included in previous mm kernels but now are droped because they caused some temperature problems for some people with no nforce motherboards.
Quoted 8 lines
Yes, the patch which disables "Halt Disconnect and Stop Grant Disconnect" apparently causes the CPU to run hot.
Quoted 3 lines
By the way, is anyone involved in solving the IO-APIC thing in nforce motherboards? Anyone trying a different approach? Anyone contacting nvidia about this problem?
As far as I know, we're dead in the water on these problems.
One day hopefully this will be sorted in the BIOSes and in mainline. I keep having to patch for every release (although as thats the only patch I have to do I'm sure there are many worse off than me). I use the 3com n/w on my A7N8X Deluxe v2 BIOS 1007 so no need for nforcedeth.
Best patches are at:
http://lkml.org/lkml/2003/12/21/7
Ive applied them to 2.6.0 and 2.6.1 and give no crashes and no heat issues.
(XP2600+ runs at 31/32C normal use and 38C compiling with Zalman cooler
+exhaust fans in box)
Craig
Quoted 39 lines · Craig Bradney
Quoted 20 lines · Andrew Morton
Quoted 4 lines · Luis Miguel García
David Ford wrote:
Quoted 8 lines
Quoted 5 lines
I have the same problem. I "solved" it a while ago by mucking with the AGP stuff. IIRC, it was turning off AGP fast writes or 8x or something similar in cmos. Went from incredibly broken to stable instantly. I'll check my cmos settings in a bit and refresh my memory.
What patches are you using?
I'm using nforce2-apic.patch and nforce2-disconnect-quirk.patch that Andrew Morton have sent to me. I think they have been included in previous mm kernels but now are droped because they caused some temperature problems for some people with no nforce motherboards.
Yes, the patch which disables "Halt Disconnect and Stop Grant Disconnect" apparently causes the CPU to run hot.
Quoted 6 lines
Quoted 3 lines
By the way, is anyone involved in solving the IO-APIC thing in nforce motherboards? Anyone trying a different approach? Anyone contacting nvidia about this problem?
As far as I know, we're dead in the water on these problems.
One day hopefully this will be sorted in the BIOSes and in mainline. I keep having to patch for every release (although as thats the only patch I have to do I'm sure there are many worse off than me). I use the 3com n/w on my A7N8X Deluxe v2 BIOS 1007 so no need for nforcedeth.
Best patches are at:
http://lkml.org/lkml/2003/12/21/7
Quoted 9 lines
Ive applied them to 2.6.0 and 2.6.1 and give no crashes and no heat issues.
(XP2600+ runs at 31/32C normal use and 38C compiling with Zalman cooler +exhaust fans in box)
Craig
you mean 31 - 38 C readed from /proc/acpi/temp[........]????
I'm having readings of 53 in idle and even 64 while compiling!! I have no case fan, but I don't think it's so important for this bug difference.
by the way, has anyone tried to contact nvidia with detailed information of this bug? Perhaps they can tell us something, not to?
-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
Quoted 10 lines · Luis Miguel García
Quoted 8 lines · Craig Bradney
One day hopefully this will be sorted in the BIOSes and in mainline. I keep having to patch for every release (although as thats the only patch I have to do I'm sure there are many worse off than me). I use the 3com n/w on my A7N8X Deluxe v2 BIOS 1007 so no need for nforcedeth.
Quoted 5 lines
Quoted 5 lines
Best patches are at:
http://lkml.org/lkml/2003/12/21/7Ive applied them to 2.6.0 and 2.6.1 and give no crashes and no heat issues.
Unfortunately that patch doesn't work for me. Still locks if I try APIC +CPU DIsc.
Quoted 10 lines
Quoted 7 lines
(XP2600+ runs at 31/32C normal use and 38C compiling with Zalman cooler +exhaust fans in box)
Craig
you mean 31 - 38 C readed from /proc/acpi/temp[........]????
I'm having readings of 53 in idle and even 64 while compiling!! I have no case fan, but I don't think it's so important for this bug difference.
The problem is, you cannot trust those infos esp not across board manufacturers. In case of Abit nearly every bios shows different values...
I have an Athon XP running at 2.1Gz with 1.65vcore. Idle: 50°C (with CPU Disc usually about 44-40°C) and under load about 54°C. I am usign a more or less self-made watercooling.
Prakash
-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
Quoted 33 lines · Prakash K. Cheemplavam
Quoted 10 lines · Luis Miguel García
Quoted 8 lines · Craig Bradney
One day hopefully this will be sorted in the BIOSes and in mainline. I keep having to patch for every release (although as thats the only patch I have to do I'm sure there are many worse off than me). I use the 3com n/w on my A7N8X Deluxe v2 BIOS 1007 so no need for nforcedeth.
Quoted 6 lines
Quoted 5 lines
Best patches are at:
http://lkml.org/lkml/2003/12/21/7Ive applied them to 2.6.0 and 2.6.1 and give no crashes and no heat issues.
Unfortunately that patch doesn't work for me. Still locks if I try APIC +CPU DIsc.
Quoted 10 lines
Quoted 7 lines
(XP2600+ runs at 31/32C normal use and 38C compiling with Zalman cooler +exhaust fans in box)
Craig
you mean 31 - 38 C readed from /proc/acpi/temp[........]????
I'm having readings of 53 in idle and even 64 while compiling!! I have no case fan, but I don't think it's so important for this bug difference.
Quoted 7 lines
The problem is, you cannot trust those infos esp not across board manufacturers. In case of Abit nearly every bios shows different values...
I have an Athon XP running at 2.1Gz with 1.65vcore. Idle: 50°C (with CPU Disc usually about 44-40°C) and under load about 54°C. I am usign a more or less self-made watercooling.
Prakash
There is a way to "activate" cpu Disconnect? or it gets enabled by simply applying it?
Yes, I have a Abit motherboards, perhaps it's the problem with the bios.
Quoted 1 line
There is a way to "activate" cpu Disconnect? or it gets enabled by simply applying it?
In newer Abit BIOSes there is an option, or you use athcool.
Quoted 1 line
Yes, I have a Abit motherboards, perhaps it's the problem with the bios.
I have an Abit NF7-S Rev2 with latest Bios.
Prakash
Quoted 9 lines · Prakash K. Cheemplavam
Quoted 2 lines
There is a way to "activate" cpu Disconnect? or it gets enabled by simply applying it?
In newer Abit BIOSes there is an option, or you use athcool.
Quoted 1 line
Yes, I have a Abit motherboards, perhaps it's the problem with the bios.
I have an Abit NF7-S Rev2 with latest Bios.
As noted in my last post.. you dont NEED athcool OR Disconnect to get stability..
I've only ever run athcool to check the status.. and my BIOS doesnt have disconnect.
A7N8X Deluxe V2 BIOS 1007.. 11 days uptime here.. haven had a crash since Ross released those patches ages ago.
Craig
Quoted 11 lines · Craig Bradney
Quoted 8 lines · Prakash K. Cheemplavam
Quoted 3 lines
There is a way to "activate" cpu Disconnect? or it gets enabled by simply applying it?
In newer Abit BIOSes there is an option, or you use athcool.
Quoted 6 lines
Quoted 6 lines
Quoted 3 lines
Yes, I have a Abit motherboards, perhaps it's the problem with the bios.
I have an Abit NF7-S Rev2 with latest Bios.
Prakash, I have the same motherboard but not the latest bios (I think I cannot overclock in the same way when I flashed the latest, so I reverted one version). Perhaps I must upgrade and try.
About the "option" you're talking about in the bios, are you talking about CPU throttle?
Quoted 11 lines
As noted in my last post.. you dont NEED athcool OR Disconnect to get stability..
I've only ever run athcool to check the status.. and my BIOS doesnt have disconnect.
A7N8X Deluxe V2 BIOS 1007.. 11 days uptime here.. haven had a crash since Ross released those patches ages ago.
Craig, I'm not talking about cpu disconnect because of the stability. I have 100% stability here with the two patches mentioned before in this thread. I was talking about my cpu showing temperatures between 53 and 64º, what I think is very high.
Quoted 5 lines · Luis Miguel García
Quoted 3 lines · Craig Bradney
On Fri, 2004-02-06 at 01:58, Prakash K. Cheemplavam wrote:
Quoted 8 lines
Quoted 4 lines
Quoted 4 lines
Quoted 2 lines
There is a way to "activate" cpu Disconnect? or it gets enabled by simply applying it?
I have an Abit NF7-S Rev2 with latest Bios.
Prakash, I have the same motherboard but not the latest bios (I think I cannot overclock in the same way when I flashed the latest, so I reverted one version). Perhaps I must upgrade and try.
About the "option" you're talking about in the bios, are you talking about CPU throttle?
Nope, it is called cpu disconnect...
Quoted 1 line
Craig, I'm not talking about cpu disconnect because of the stability. I have 100% stability here with the two patches mentioned before in this thread. I was talking about my cpu showing temperatures between 53 and 64º, what I think is very high.
Do you only have stability with those two patches? I mean WITHOUT Disconnect I don't need those patches to have stability (with certain kernels...). At least the delay patch is not needed. The timer irq mapping patch might be usefull, but i am not 100% sure.
Prakash
Quoted 13 lines · Craig Bradney
Quoted 9 lines · Prakash K. Cheemplavam
Quoted 1 line
There is a way to "activate" cpu Disconnect? or it gets enabled by simply applying it?
In newer Abit BIOSes there is an option, or you use athcool.
Quoted 1 line
Yes, I have a Abit motherboards, perhaps it's the problem with the bios.
I have an Abit NF7-S Rev2 with latest Bios.
Quoted 8 lines
As noted in my last post.. you dont NEED athcool OR Disconnect to get stability..
I've only ever run athcool to check the status.. and my BIOS doesnt have disconnect.
A7N8X Deluxe V2 BIOS 1007.. 11 days uptime here.. haven had a crash since Ross released those patches ages ago.
WITHOUT Disconnect my System is stable, but hotter when idle, so that is not the point. Ross wanted the patched to work with APIC and Disconnect.
DO you guys use APIC (not ACPI)? I use both APIC (and local APIC) and ACPI. ACPI is not the problem (unless they break something...) but APIC and CPU Disconnect makes trouble on nforce2.
Prakash
Quoted 24 lines · Prakash K. Cheemplavam
Quoted 14 lines · Craig Bradney
Quoted 10 lines · Prakash K. Cheemplavam
Quoted 2 lines
There is a way to "activate" cpu Disconnect? or it gets enabled by simply applying it?
In newer Abit BIOSes there is an option, or you use athcool.
Quoted 1 line
Yes, I have a Abit motherboards, perhaps it's the problem with the bios.
I have an Abit NF7-S Rev2 with latest Bios.
Quoted 8 lines
As noted in my last post.. you dont NEED athcool OR Disconnect to get stability..
I've only ever run athcool to check the status.. and my BIOS doesnt have disconnect.
A7N8X Deluxe V2 BIOS 1007.. 11 days uptime here.. haven had a crash since Ross released those patches ages ago.
Quoted 6 lines
WITHOUT Disconnect my System is stable, but hotter when idle, so that is not the point. Ross wanted the patched to work with APIC and Disconnect.
DO you guys use APIC (not ACPI)? I use both APIC (and local APIC) and ACPI. ACPI is not the problem (unless they break something...) but APIC and CPU Disconnect makes trouble on nforce2.
athcool reports:
nVIDIA nForce2 (10de 01e0) found
'Halt Disconnect and Stop Grant Disconnect' bit is enabled.
I have never used athcool to turn it off, and I don't have a BIOS option. APIC, local APIC and ACPI are all on.
Craig
Quoted 11 lines
Quoted 9 lines
Quoted 5 lines
I've only ever run athcool to check the status.. and my BIOS doesnt have disconnect.
A7N8X Deluxe V2 BIOS 1007.. 11 days uptime here.. haven had a crash since Ross released those patches ages ago.
WITHOUT Disconnect my System is stable, but hotter when idle, so that is not the point. Ross wanted the patched to work with APIC and Disconnect.
DO you guys use APIC (not ACPI)? I use both APIC (and local APIC) and ACPI. ACPI is not the problem (unless they break something...) but APIC and CPU Disconnect makes trouble on nforce2.
Quoted 6 lines
athcool reports:
nVIDIA nForce2 (10de 01e0) found
'Halt Disconnect and Stop Grant Disconnect' bit is enabled.I have never used athcool to turn it off, and I don't have a BIOS option. APIC, local APIC and ACPI are all on.
Ok, then it makes sense, so you are using APIc with CPU Disconnect and Ross' patch. This explains your low idle temps. As I said this config doesn't work for me.
Prakash
Quoted 1 line · Prakash K. Cheemplavam
Ok, then it makes sense, so you are using APIc with CPU Disconnect and Ross' patch. This explains your low idle temps. As I said this config doesn't work for me.
Have you experimented with the new apic_tack boot options in Ross's latest patches? apic_tack=2 seems to work best for me.
Daniel
Quoted 7 lines · Daniel Drake
Quoted 1 line · Prakash K. Cheemplavam
Ok, then it makes sense, so you are using APIc with CPU Disconnect and Ross' patch. This explains your low idle temps. As I said this config doesn't work for me.
Have you experimented with the new
apic_tackboot options in Ross's latest patches? apic_tack=2 seems to work best for me.
Stupid me. I haven't thoruoughly read the text. I have not activated the patch, so I'll try this. thx for pointing out..
Prakash
Quoted 8 lines · Prakash K. Cheemplavam
Quoted 6 lines · Daniel Drake
Quoted 1 line · Prakash K. Cheemplavam
Ok, then it makes sense, so you are using APIc with CPU Disconnect and Ross' patch. This explains your low idle temps. As I said this config doesn't work for me.
Quoted 5 lines
Quoted 2 lines
Have you experimented with the new
apic_tackboot options in Ross's latest patches? apic_tack=2 seems to work best for me.Stupid me. I haven't thoruoughly read the text. I have not activated the patch, so I'll try this. thx for pointing out..
OK, I appended apic_tack=2 and yes, it survives several hdparms! Great, so gonna try if it is really stable.Then I can try =1. CPU cooling down. Already at 46°C. :-)
Not bad,not bad, though I saw a small performace degration: hdparm gives me 60-61mb/s instead of >62mb/s, but I won't complain. :-)
Prakash
Quoted 19 lines · Prakash K. Cheemplavam
Quoted 10 lines · Prakash K. Cheemplavam
Quoted 8 lines · Daniel Drake
Quoted 3 lines · Prakash K. Cheemplavam
Ok, then it makes sense, so you are using APIc with CPU Disconnect and Ross' patch. This explains your low idle temps. As I said this config doesn't work for me.
Quoted 7 lines
Quoted 3 lines
Have you experimented with the new
apic_tackboot options in Ross's latest patches?
apic_tack=2 seems to work best for me.Stupid me. I haven't thoruoughly read the text. I have not activated the patch, so I'll try this. thx for pointing out..
Quoted 7 lines
OK, I appended apic_tack=2 and yes, it survives several hdparms! Great, so gonna try if it is really stable.Then I can try =1. CPU cooling down. Already at 46°C. :-)
Not bad,not bad, though I saw a small performace degration: hdparm gives me 60-61mb/s instead of >62mb/s, but I won't complain. :-)
Ahh yes.. missing the kernel line argument will make a difference. I'm running apic_tack=2 as well. From what I remember =2 was the "better" patch option if it made your system stable.
Craig
Here is a preliminary webpage documenting some stuff, it will mature later on (probably this weekend), but it would be nice to include success stories of people that have a setup that works for them. If you have one that you wish to share, please email me.
Typically, tell me about brand, make and bios revision of your board, the kernel tree you used, patches applied and URLs to those patches so I can mirror them locally. If you have the patch locally you could just send it, but my mbox is only 8MB so I would really prefer an URL.
Maybe it is a good idea to include the kernel configs as well, so that we can get "monkey proof" solutions.
Let me know what you think.
Regards,
Arjen
Quoted 34 lines · Craig Bradney
Quoted 19 lines · Prakash K. Cheemplavam
Quoted 10 lines · Prakash K. Cheemplavam
Quoted 8 lines · Daniel Drake
Quoted 3 lines · Prakash K. Cheemplavam
Ok, then it makes sense, so you are using APIc with CPU Disconnect and Ross' patch. This explains your low idle temps. As I said this config doesn't work for me.
Quoted 7 lines
Quoted 3 lines
Have you experimented with the new
apic_tackboot options in Ross's latest patches?
apic_tack=2 seems to work best for me.Stupid me. I haven't thoruoughly read the text. I have not activated the patch, so I'll try this. thx for pointing out..
Quoted 7 lines
OK, I appended apic_tack=2 and yes, it survives several hdparms! Great, so gonna try if it is really stable.Then I can try =1. CPU cooling down. Already at 46�C. :-)
Not bad,not bad, though I saw a small performace degration: hdparm gives me 60-61mb/s instead of >62mb/s, but I won't complain. :-)
Ahh yes.. missing the kernel line argument will make a difference. I'm running apic_tack=2 as well. From what I remember =2 was the "better" patch option if it made your system stable.
Craig
Quoted 9 lines
Quoted 4 lines
OK, I appended apic_tack=2 and yes, it survives several hdparms! Great, so gonna try if it is really stable.Then I can try =1. CPU cooling down. Already at 46°C. :-)
Not bad,not bad, though I saw a small performace degration: hdparm gives me 60-61mb/s instead of >62mb/s, but I won't complain. :-)
Ahh yes.. missing the kernel line argument will make a difference. I'm running apic_tack=2 as well. From what I remember =2 was the "better" patch option if it made your system stable.
Yes, you're right. So far 2 seems to be pretty stable.Nice. :-)
Prakash
It would probably be more useful if I included the URL as well duh... (thx Craig)
http://atlas.et.tudelft.nl/verwei90/nforce2/
http://atlas.et.tudelft.nl/verwei90/nforce2/index.html
both should work.
Quoted 34 lines · Craig Bradney
Quoted 19 lines · Prakash K. Cheemplavam
Quoted 10 lines · Prakash K. Cheemplavam
Quoted 8 lines · Daniel Drake
Quoted 3 lines · Prakash K. Cheemplavam
Ok, then it makes sense, so you are using APIc with CPU Disconnect and Ross' patch. This explains your low idle temps. As I said this config doesn't work for me.
Quoted 7 lines
Quoted 3 lines
Have you experimented with the new
apic_tackboot options in Ross's latest patches?
apic_tack=2 seems to work best for me.Stupid me. I haven't thoruoughly read the text. I have not activated the patch, so I'll try this. thx for pointing out..
Quoted 7 lines
OK, I appended apic_tack=2 and yes, it survives several hdparms! Great, so gonna try if it is really stable.Then I can try =1. CPU cooling down. Already at 46�C. :-)
Not bad,not bad, though I saw a small performace degration: hdparm gives me 60-61mb/s instead of >62mb/s, but I won't complain. :-)
Ahh yes.. missing the kernel line argument will make a difference. I'm running apic_tack=2 as well. From what I remember =2 was the "better" patch option if it made your system stable.
Craig
Quoted 5 lines · Arjen Verweij
It would probably be more useful if I included the URL as well duh... (thx Craig)
http://atlas.et.tudelft.nl/verwei90/nforce2/
http://atlas.et.tudelft.nl/verwei90/nforce2/index.html
* 3.2 Turning off the CPU bus disconnect in the BIOS? (verify)
(drawback: causes cpu to run significantly hotter)
Inprecice: CPU runs as hot as without disconnect, only without load, cpu disconnect lets the cpu cool down.
With athcool you can alsao turn it off, or there is a quirk-patch available from Bart.
* 3.3 Apply patches cooked up by Ross Dickson, possibly in
conjunction with 3.2? (verify)
o 3.3.1 Seems to work for Craig Bradney (hardware unknown,
some Asus board?) and Prakash K. Cheemplavam (Abit NF7-S)
I have Rev2 (2.0) board and d22 (or 2.2 ? Abit has a strange way of numbering.) Bios. I only applied first patch. I already use an equivalent to the second one.
Prakash
Oh, I am using a furthermore patched version of this kernel:
http://forums.gentoo.org/viewtopic.php?p=840090#840090
It is based on 2.6-mm1 and has a lot of goodies.
Prakash
Quoted 32 lines · Craig Bradney
Quoted 24 lines · Prakash K. Cheemplavam
Quoted 14 lines · Craig Bradney
Quoted 10 lines · Prakash K. Cheemplavam
Quoted 2 lines
There is a way to "activate" cpu Disconnect? or it gets enabled by simply applying it?
In newer Abit BIOSes there is an option, or you use athcool.
Quoted 1 line
Yes, I have a Abit motherboards, perhaps it's the problem with the bios.
I have an Abit NF7-S Rev2 with latest Bios.
Quoted 8 lines
As noted in my last post.. you dont NEED athcool OR Disconnect to get stability..
I've only ever run athcool to check the status.. and my BIOS doesnt have disconnect.
A7N8X Deluxe V2 BIOS 1007.. 11 days uptime here.. haven had a crash since Ross released those patches ages ago.
Quoted 6 lines
WITHOUT Disconnect my System is stable, but hotter when idle, so that is not the point. Ross wanted the patched to work with APIC and Disconnect.
DO you guys use APIC (not ACPI)? I use both APIC (and local APIC) and ACPI. ACPI is not the problem (unless they break something...) but APIC and CPU Disconnect makes trouble on nforce2.
Quoted 7 lines
athcool reports:
nVIDIA nForce2 (10de 01e0) found
'Halt Disconnect and Stop Grant Disconnect' bit is enabled.I have never used athcool to turn it off, and I don't have a BIOS option. APIC, local APIC and ACPI are all on.
I have to add there were some issues found with the patches I'm using relating to timing.. (because that is in effect most of the problem). The issue was with the PC clock losing time but I never noticed it because I run ntpd anyway.
Craig
Quoted 48 lines · Luis Miguel García
Quoted 46 lines · Prakash K. Cheemplavam
Quoted 11 lines · Luis Miguel García
Quoted 9 lines · Craig Bradney
One day hopefully this will be sorted in the BIOSes and in mainline. I keep having to patch for every release (although as thats the only patch
I have to do I'm sure there are many worse off than me). I use the 3com n/w on my A7N8X Deluxe v2 BIOS 1007 so no need for nforcedeth.Quoted 6 lines
Quoted 5 lines
Best patches are at:
http://lkml.org/lkml/2003/12/21/7Ive applied them to 2.6.0 and 2.6.1 and give no crashes and no heat issues.
Unfortunately that patch doesn't work for me. Still locks if I try APIC +CPU DIsc.
Quoted 4 lines
Quoted 4 lines
(XP2600+ runs at 31/32C normal use and 38C compiling with Zalman cooler +exhaust fans in box)
Quoted 8 lines
Quoted 3 lines
Craig
you mean 31 - 38 C readed from /proc/acpi/temp[........]????
I'm having readings of 53 in idle and even 64 while compiling!! I have no case fan, but I don't think it's so important for this bug difference.
The problem is, you cannot trust those infos esp not across board manufacturers. In case of Abit nearly every bios shows different values...
I have an Athon XP running at 2.1Gz with 1.65vcore. Idle: 50°C (with CPU Disc usually about 44-40°C) and under load about 54°C. I am usign a more or less self-made watercooling.
Quoted 7 lines
Quoted 2 lines
Prakash
There is a way to "activate" cpu Disconnect? or it gets enabled by simply applying it?
Yes, I have a Abit motherboards, perhaps it's the problem with the bios.
I'm not activating Disconnect.. I'm using Ross's latest 2 patches.. one of them avoids using Disconnect.
Craig
Quoted 58 lines · Luis Miguel García
Quoted 47 lines · Craig Bradney
Quoted 26 lines · Andrew Morton
Quoted 9 lines · Luis Miguel García
Quoted 5 lines · David Ford
I have the same problem. I "solved" it a while ago by mucking with the AGP stuff. IIRC, it was turning off AGP fast writes or 8x or something similar in cmos. Went from incredibly broken to stable instantly. I'll check my cmos settings in a bit and refresh my memory.
Quoted 9 lines
Quoted 3 lines
What patches are you using?
I'm using nforce2-apic.patch and nforce2-disconnect-quirk.patch that Andrew Morton have sent to me. I think they have been included in previous mm kernels but now are droped because they caused some temperature problems for some people with no nforce motherboards.
Yes, the patch which disables "Halt Disconnect and Stop Grant Disconnect" apparently causes the CPU to run hot.
Quoted 8 lines
Quoted 5 lines
By the way, is anyone involved in solving the IO-APIC thing in nforce motherboards? Anyone trying a different approach? Anyone contacting nvidia about this problem?
As far as I know, we're dead in the water on these problems.
One day hopefully this will be sorted in the BIOSes and in mainline. I keep having to patch for every release (although as thats the only patch I have to do I'm sure there are many worse off than me). I use the 3com n/w on my A7N8X Deluxe v2 BIOS 1007 so no need for nforcedeth.
Best patches are at:
http://lkml.org/lkml/2003/12/21/7Quoted 9 lines
Ive applied them to 2.6.0 and 2.6.1 and give no crashes and no heat issues.
(XP2600+ runs at 31/32C normal use and 38C compiling with Zalman cooler +exhaust fans in box)
Craig
Quoted 7 lines
you mean 31 - 38 C readed from /proc/acpi/temp[........]????
I'm having readings of 53 in idle and even 64 while compiling!! I have no case fan, but I don't think it's so important for this bug difference.
by the way, has anyone tried to contact nvidia with detailed information of this bug? Perhaps they can tell us something, not to?
no.. /sys/bus/i2c
I was highly sceptical that these values were wrong.. but if I was to shut down and immediately look at the BIOS values.. they are close enough to make the values I'm quoting to be on the mark. Of course, the BIOS values could be wrong.
I had the normal Athlon cooler and one rear case fan and it was maxing out at around 50C. Putting on the Zalman cooler dropped it by 10C. Adding in the rear fans means it never goes above 39C while compiling. If the windows open and theres a draft through the room and cold outside.. :) ..it will idle at 29ish.
Craig
Quoted 5 lines · Andrew Morton
Quoted 3 lines
By the way, is anyone involved in solving the IO-APIC thing in nforce motherboards? Anyone trying a different approach? Anyone contacting nvidia about this problem?
As far as I know, we're dead in the water on these problems.
Not necessarily. :-)
Quoted 17 lines
Here's one:
[x86] do not wrongly override
mp_ExtINTIRQFrom: Mathieu <[email protected]>.
With this patch timer IRQ0 is correctly set to IO-APIC-edge (not XT-PIC) on nForce2 boards when using APIC and ACPI.
arch/i386/kernel/mpparse.c | 3 ++- 1 files changed, 2 insertions(+), 1 deletion(-)diff -puN
arch/i386/kernel/mpparse.c~nforce2-apicarch/i386/kernel/mpparse.c
arch/i386/kernel/mpparse.c void __init mp_override_legacy_irq ( · line 969
Quoted 9 lines
arch/i386/kernel/mpparse.c renamed from arch/i386/kernel/mpparse.c~nforce2-apic +2
-1@@ -962,7 +962,8 @@ void __init mp_override_legacy_irq (*/ for (i =0 ; i < mp_irq_entries; i++) {if ((mp_irqs[i].mpc_dstapic == intsrc.mpc_dstapic) - && (mp_irqs[i].mpc_srcbusirq == intsrc.mpc_srcbusirq)) { + && (mp_irqs[i].mpc_srcbusirq == intsrc.mpc_srcbusirq) + && (mp_irqs[i].mpc_irqtype == intsrc.mpc_irqtype)) { mp_irqs[i] = intsrc; found =1 ;break ;
That's not the right fix. There's a bug in Linux's ACPI IRQ setup as
I've discovered by comparing the code to the spec. Here's a patch I sent in December both to the LKML and the ACPI maintainer. The feedback from the list was positive, but the maintainer didn't bother to comment.
I haven't pushed the patch more firmly, because the MIPS port is my
priority and I don't even have any ACPI-aware equipment.
Maciej
Signature · 46 lines
+ Maciej W. Rozycki, Technical University of Gdansk, Poland + +--------------------------------------------------------------+ + e-mail: [email protected], PGP key available + patch-mips-2.6.0-test11-20031209-acpi-irq0-1 diff -up --recursive --new-file linux-mips-2.6.0-test11-20031209.macro/arch/i386/kernel/mpparse.c linux-mips-2.6.0-test11-20031209/arch/i386/kernel/mpparse.c --- linux-mips-2.6.0-test11-20031209.macro/arch/i386/kernel/mpparse.c 2003-11-25 04:57:01.000000000 +0000 +++ linux-mips-2.6.0-test11-20031209/arch/i386/kernel/mpparse.c 2003-12-11 09:43:26.000000000 +0000 @@ -940,7 +940,7 @@ void __init mp_override_legacy_irq ( * erroneously sets the trigger to level, resulting in a HUGE * increase of timer interrupts! */ - if ((bus_irq == 0) && (global_irq == 2) && (trigger == 3)) + if ((bus_irq == 0) && (trigger == 3)) trigger = 1; intsrc.mpc_type = MP_INTSRC; @@ -961,7 +961,7 @@ void __init mp_override_legacy_irq ( * Otherwise create a new entry (e.g. global_irq == 2). */ for (i = 0; i < mp_irq_entries; i++) { - if ((mp_irqs[i].mpc_dstapic == intsrc.mpc_dstapic) + if ((mp_irqs[i].mpc_srcbus == intsrc.mpc_srcbus) && (mp_irqs[i].mpc_srcbusirq == intsrc.mpc_srcbusirq)) { mp_irqs[i] = intsrc; found = 1; @@ -1008,9 +1008,10 @@ void __init mp_config_acpi_legacy_irqs ( */ for (i = 0; i < 16; i++) { - if (i == 2) continue; /* Don't connect IRQ2 */ + if (i == 2) + continue; /* Don't connect IRQ2 */ - intsrc.mpc_irqtype = i ? mp_INT : mp_ExtINT; /* 8259A to #0 */ + intsrc.mpc_irqtype = mp_INT; intsrc.mpc_srcbusirq = i; /* Identity mapped */ intsrc.mpc_dstirq = i; ------------------------------------------------------- The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA. http://www.eclipsecon.org/osdn
Quoted 4 lines · On Sat, 7 Feb 2004 00:33:04 +0100 (CET) "Maciej W. Rozycki"…
That's not the right fix. There's a bug in Linux's ACPI IRQ setup as
I've discovered by comparing the code to the spec. Here's a patch I sent in December both to the LKML and the ACPI maintainer. The feedback from the list was positive, but the maintainer didn't bother to comment.
Thanks. I added the patch to the x86-64 sources and it indeed seems to fix the Nforce3.
-Andi
Quoted 4 lines · Andi Kleen
On Sat, 7 Feb 2004 00:33:04 +0100 (CET)
"Maciej W. Rozycki" <[email protected]> wrote:
Quoted 8 lines
Quoted 4 lines
That's not the right fix. There's a bug in Linux's ACPI IRQ setup as
I've discovered by comparing the code to the spec. Here's a patch I sent in December both to the LKML and the ACPI maintainer. The feedback from the list was positive, but the maintainer didn't bother to comment.Thanks. I added the patch to the x86-64 sources and it indeed seems to fix the Nforce3.
I tried the patch against 2.6.3-rc1 and it doesn't seem to fix anything, timer is still on XT-PIC. This is on a A7N8X Deluxe rev2, nforce2.
CPU0
0: 524366 XT-PIC timer 1: 1727 IO-APIC-edge i8042 2: 0 XT-PIC cascade 8: 1 IO-APIC-edge rtc 9: 0 IO-APIC-level acpi 14: 913 IO-APIC-edge ide0 15: 22792 IO-APIC-edge ide1 19: 23559 IO-APIC-level nvidia 20: 52452 IO-APIC-level ohci_hcd, eth0 21: 0 IO-APIC-level ehci_hcd, NVidia nForce2 22: 354 IO-APIC-level ohci_hcd NMI: 0 LOC: 524223 ERR: 0 MIS: 0
Signature · 6 lines
Josh McKinney | Webmaster: http://joshandangie.org
--------------------------------------------------------------------------
| They that can give up essential liberty
Linux, the choice -o) | to obtain a little temporary safety deserve
of the GNU generation /\ | neither liberty or safety.
_\_v | -Benjamin FranklinQuoted 3 lines · Josh McKinney
I tried the patch against 2.6.3-rc1 and it doesn't seem to fix anything, timer is still on XT-PIC. This is on a A7N8X Deluxe rev2, nforce2.
The patch fixes our bad interpretation of the ACPI spec -- it's not
sufficient to work around errors in the nforce2's ACPI tables, which should really get fixed by the manufacturer with a BIOS update.
Signature · 3 lines
+ Maciej W. Rozycki, Technical University of Gdansk, Poland + +--------------------------------------------------------------+ + e-mail: [email protected], PGP key available +
Quoted 49 lines · Maciej W. Rozycki
Quoted 6 lines · Andrew Morton
Quoted 3 lines
By the way, is anyone involved in solving the IO-APIC thing in nforce motherboards? Anyone trying a different approach? Anyone contacting nvidia about this problem?
As far as I know, we're dead in the water on these problems.
Not necessarily. :-)
Quoted 19 lines
Here's one:
[x86] do not wrongly override
mp_ExtINTIRQFrom: Mathieu <[email protected]>.
With this patch timer IRQ0 is correctly set to IO-APIC-edge (not XT-PIC) on nForce2 boards when using APIC and ACPI.
arch/i386/kernel/mpparse.c| 3 ++-1 files changed, 2 insertions(+), 1 deletion(-)diff -puN
arch/i386/kernel/mpparse.c~nforce2-apicarch/i386/kernel/mpparse.carch/i386/kernel/mpparse.c void __init mp_override_legacy_irq ( · line 969
Quoted 9 lines
arch/i386/kernel/mpparse.c renamed from arch/i386/kernel/mpparse.c~nforce2-apic +2
-1@@ -962,7 +962,8 @@ void __init mp_override_legacy_irq (*/ for (i =0 ; i < mp_irq_entries; i++) {if ((mp_irqs[i].mpc_dstapic == intsrc.mpc_dstapic) - && (mp_irqs[i].mpc_srcbusirq == intsrc.mpc_srcbusirq)) { + && (mp_irqs[i].mpc_srcbusirq == intsrc.mpc_srcbusirq) + && (mp_irqs[i].mpc_irqtype == intsrc.mpc_irqtype)) { mp_irqs[i] = intsrc; found =1 ;break ;That's not the right fix. There's a bug in Linux's ACPI IRQ setup as I've discovered by comparing the code to the spec. Here's a patch I sent in December both to the LKML and the ACPI maintainer. The feedback from the list was positive, but the maintainer didn't bother to comment.
Quoted 7 lines
I haven't pushed the patch more firmly, because the MIPS port is my priority and I don't even have any ACPI-aware equipment.
Maciej
Can you send us the specific patch or at least telling us if you're going to push it?
Thanks a lot for the fix ;)
LuisMi Garcia
-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
Quoted 1 line · Luis Miguel García
By the way, is anyone involved in solving the IO-APIC thing in nforce motherboards? Anyone trying a different approach? Anyone contacting nvidia about this problem?
Ross said in previous mails that he had tried to contact nvidia through various methods, and had not recieved a response.
He also mentioned that AMD were being much more responsive, and he had a support request open with them, which he was assured was being investigated.
But yeah, doesn't look like we are getting anywhere with the manufacturers...
Daniel
-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn