Nvidia has open source drivers at nvidia.com

[gilboa]

Originally Posted by JoshPrismon

Thanks for your support. I actually have noticed some of teh changes coming in. The open source community is very thankfull for that.

I second the above.
Hopefully, nVidia will find a way to incorporate some 3D support into the open source nv drivers without running into IP/NDA/etc problem.

- Gilboa

[Thunderbird]

Most likely nvidia will never provide 3d support in the nv driver. You should place your bets on the nouveau project which attempts to extend the nv driver with dri and 3d support. The basic drm part is starting to work but a lot of stuff still needs to be changed before they can add 3d code.

Perhaps some day when the nouveau driver is decent it can be merged with nv. Though this will be tough. The nv driver is maintained by an nvidia employee and nvidia ofcourse wants to reveal as little info as possible and you can imagine thayt they don't want to reveal 'leaked' / 'reverse engineered' specs. But perhaps there will be a way but most likely something like this is atleast a year perhaps more away.

[Lithorus]

Why does nvidia insist on ONLY supporting SUSE and Redhat? I opened the zip file only to find directories which adresses to SUSE and Redhat. How about a generic folder for everyone else? It's quite common practice.

[irondog]

Originally Posted by netllama

irondog, what you've stated is untrue. The '1.11' zip archive that is available from NVIDIA's website contains versions of foreceth and sata_nv which are newer than what is currently available from an official kernel.org kernel. NVIDIA's developers contributed bug fixes as well as enhancements to the code that was posted. This isn't simply grabbing what was already in a kernel.org kernel and reposting it.
[...]
Lonni

Okay. Maybe I should have investigated better before posting. I was just seeing filenames which have been in the Linux kernel for a long time. I didn't see any makefiles or SRPM's which made me think NVIDIA was just starring in the show with nothing (some .c files and RPM's in a .zip).

I'm still wandering how somebody could reproduce these RPM's. There are neither spec-files nor SRPM's, nor Makefiles. I could also pick the kernel's .c file, alter some lines, provide some binary files and say: "here you go, new, NVIDIA reviewed drivers". Creating driver disk images for old Redhat versions isn't anything to be proud of either. These libata sata drivers are easy to backport.

Anyway, it's nice if NVIDIA is contributing to existing kernel code.

[netllama]

I'll try, as best as I can, to respond to all the recent questions regarding the 'new' open source driver (1.11) package. Please see my sticky post for additional details:
http://www.nvnews.net/vbulletin/showthread.php?p=945687

fhj52,
All that you need for any NVIDIA NIC is the forcedeth driver. There is nothing special about the driver package on NVIDIA's website which restricts it exclusively to MCP55. The driver package in the above URL includes pre-compiled RPMs and driver disks (which can be used for adding hardware support when installing the associated Linux distribution). Additionally, it provides the source for the forcedeth, sata_nv and amd74xx.c drivers. One thing to note is that the source is only included (under the corresponding distribution subdirectory) for distributions which do not already ship with the latest version of the corresponding driver.

In the case of SuSE-10.1, there isn't currently an explicit driver package, however you can build the new forcedeth for your SuSE-10.1 kernel fairly easily. Create a new Makefile the line:

obj-m := forcedeth.o


Then, type the command "make -C /usr/src/<kernel> SUBDIRS=$PWD modules" to build the forcedeth module for the kernel you specified. SuSE-10.1 already ships with the same version of sata_nv & amd74xx which is included in the 1.11 driver package.

gilboa,
As Thunderbird commented, there are no current plans to add any accelerated support to the 'nv' X driver, as NVIDIA already ships a driver with 3D acceleration (the 'nvidia' X driver).

Lithorus,
SuSE and Redhat are explicitly 'supported' as NVIDIA's enterprise customers have requested support for those distributions. The source for the drivers is also included in the 1.11 package for anyone running a different distribution, should they wish to build the drivers. For a recent (2.6.16.x or later) kernel, all you should need to do is replace the 'default' source with that included in the 1.11 package (or my sticky post) and (re)build the kernel.

irondog,
If you read the README that is included with the 1.11 driver package, you'd se that it provides instructions how to build the drivers from source. I'll talk with development about the possibility of posting SRPMs, or at least a SPEC file to generate the RPMs. In the interim, you should be able to build the forcedeth driver with the Makefile I noted earlier in this post.

Thanks,
Lonni

[fhj52]

Originally Posted by netllama

I'll try, as best as I can, to respond to all the recent questions regarding the 'new' open source driver (1.11) package. Please see my sticky post for additional details:
http://www.nvnews.net/vbulletin/showthread.php?p=945687

fhj52,
All that you need for any NVIDIA NIC is the forcedeth driver. There is nothing special about the driver package on NVIDIA's website which restricts it exclusively to MCP55. The driver package in the above URL includes pre-compiled RPMs and driver disks (which can be used for adding hardware support when installing the associated Linux distribution). Additionally, it provides the source for the forcedeth, sata_nv and amd74xx.c drivers. One thing to note is that the source is only included (under the corresponding distribution subdirectory) for distributions which do not already ship with the latest version of the corresponding driver.

In the case of SuSE-10.1, there isn't currently an explicit driver package, however you can build the new forcedeth for your SuSE-10.1 kernel fairly easily. Create a new Makefile the line:

obj-m := forcedeth.o


Then, type the command "make -C /usr/src/<kernel> SUBDIRS=$PWD modules" to build the forcedeth module for the kernel you specified. SuSE-10.1 already ships with the same version of sata_nv & amd74xx which is included in the 1.11 driver package.

[snip]
Thanks,
Lonni

Thank you.
I'll confess to not applying the sticky info to my knowledge base. The date on it is July, but 1.11 was relased this week so I did not expect it to be applicable.

Thanks for the make & compile info, as well as letting me know that " same version of sata_nv & amd74xx " are already being used.

I had a bit of confusion trying to figure out which forcedeth.c to use
This kernel, 2.6.16.21, has forcedeth.c version 0.49

#> grep \#define.FORCEDETH.VERSION \" /usr/src/linux-2.6.16.21-0.13/drivers/net/forcedeth.c

/usr/src/linux-2.6.16.21-0.13/drivers/net/forcedeth.c:#define FORCEDETH_VERSION         "0.49"

After looking at some of the code, it really is, apparently, version 0.49.
However, ALL of the 1.11 included source files are shown as version 0.56:

#> grepr \#define.FORCEDETH.VERSION \" *

Fedora4/source/forcedeth.c:123:#define FORCEDETH_VERSION                "0.56"
Fedora5/source/forcedeth.c:123:#define FORCEDETH_VERSION                "0.56"
RHEL3_U4/source/forcedeth.c:122:#define FORCEDETH_VERSION               "0.56"
RHEL3_U5/source/forcedeth.c:122:#define FORCEDETH_VERSION               "0.56"
RHEL3_U6/source/forcedeth.c:122:#define FORCEDETH_VERSION               "0.56"
RHEL3_U7/source/forcedeth.c:122:#define FORCEDETH_VERSION               "0.56"
RHEL4_U1/source/forcedeth.c:123:#define FORCEDETH_VERSION               "0.56"
RHEL4_U2/source/forcedeth.c:123:#define FORCEDETH_VERSION               "0.56"
RHEL4_U3/source/forcedeth.c:123:#define FORCEDETH_VERSION               "0.56"
SLES9_SP1/source/forcedeth.c:123:#define FORCEDETH_VERSION              "0.56"
SLES9_SP2/source/forcedeth.c:123:#define FORCEDETH_VERSION              "0.56"
SLES9_SP3/source/forcedeth.c:123:#define FORCEDETH_VERSION              "0.56"
SuSE10/source/forcedeth.c:123:#define FORCEDETH_VERSION         "0.56"

Considering that some of the files are different sizes, that made me even more confused.
Using

diff Fedora4/source/forcedeth.c RHEL3_U7/source/forcedeth.c

to compare a couple that have different sizes brings up the next issue.
The files are different(no surprise) but they are the same version number(huge surprise). You just cannot do that.
It looks like 0.56 got hacked to work with some of the older distros but whoever hacked it did not either revert to 0.51 ('cause apparently 0.52 got whacked out completely) or annotate the version.
Two files with same name but different code absolutely cannot have the same version number. Of course maybe I missed something but if the above is true, please get somebody to fix it. ...

Besides, version 0.56 should build and run on the older distros with proper config options ... but if a mod' is mandatory and nVidia *nicely* fixed 0.56 so it will, then the "fixed" 0.56 needs to be #define'd as 0.56-reversion or, perhaps, just rename it to forcedeth_nv.c so we all know it is forcedeth but is nVidia's *special* version of it.

AND, while on that subject, someone should look at the sata_nv.c, & etc too since some of those are apparently different also:

#> lloc sata_nv.c`
-rw-rw-rw- 1 19K 2006-03-23 14:16 /NFORCE-Linux-x86-1.11/NV_MCP55_Linux_DRV_PKG_v1.11/Fedora4/source/sata_nv.c
-rw-rw-rw- 1 19K 2006-05-01 14:38 /NFORCE-Linux-x86-1.11/NV_MCP55_Linux_DRV_PKG_v1.11/RHEL3_U4/source/sata_nv.c
-rw-rw-rw- 1 19K 2006-04-05 13:56 /NFORCE-Linux-x86-1.11/NV_MCP55_Linux_DRV_PKG_v1.11/RHEL3_U5/source/sata_nv.c
-rw-rw-rw- 1 19K 2006-04-05 13:56 /NFORCE-Linux-x86-1.11/NV_MCP55_Linux_DRV_PKG_v1.11/RHEL3_U6/source/sata_nv.c
-rw-rw-rw- 1 20K 2006-05-19 21:37 /NFORCE-Linux-x86-1.11/NV_MCP55_Linux_DRV_PKG_v1.11/RHEL3_U7/source/sata_nv.c
-rw-rw-rw- 1 19K 2006-03-23 14:16 /NFORCE-Linux-x86-1.11/NV_MCP55_Linux_DRV_PKG_v1.11/RHEL4_U1/source/sata_nv.c
-rw-rw-rw- 1 19K 2006-03-23 14:16 /NFORCE-Linux-x86-1.11/NV_MCP55_Linux_DRV_PKG_v1.11/RHEL4_U2/source/sata_nv.c
-rw-rw-rw- 1 20K 2006-05-19 21:38 /NFORCE-Linux-x86-1.11/NV_MCP55_Linux_DRV_PKG_v1.11/RHEL4_U3/source/sata_nv.c
-rw-rw-rw- 1 21K 2006-05-23 06:43 /NFORCE-Linux-x86-1.11/NV_MCP55_Linux_DRV_PKG_v1.11/SLES9_SP1/source/sata_nv.c
-rw-rw-rw- 1 19K 2006-05-19 21:42 /NFORCE-Linux-x86-1.11/NV_MCP55_Linux_DRV_PKG_v1.11/SLES9_SP2/source/sata_nv.c
-rw-rw-rw- 1 20K 2006-05-19 16:23 /NFORCE-Linux-x86-1.11/NV_MCP55_Linux_DRV_PKG_v1.11/SLES9_SP3/source/sata_nv.c
-rw-r--r-- 1 18K 2006-03-19 23:53 /usr/src/linux-2.6.16.21-0.13/drivers/scsi/sata_nv.c
It's 13:23:22 CDT (UTC-0500) on Thu Aug 24, week 34 in 2006.

Different sizes and different code too ... which does not matter as long as version jives but it does not:

#> grepr \#define.DRV_VERSION \" *

Fedora4/source/sata_nv.c:51:#define DRV_VERSION                           "0.6"
RHEL3_U4/source/sata_nv.c:47:#define DRV_VERSION                        "0.03"
RHEL3_U5/source/sata_nv.c:51:#define DRV_VERSION                        "0.6"
RHEL3_U6/source/sata_nv.c:51:#define DRV_VERSION                        "0.6"
RHEL3_U7/source/sata_nv.c:51:#define DRV_VERSION                        "0.10"
RHEL4_U1/source/sata_nv.c:51:#define DRV_VERSION                        "0.6"
RHEL4_U2/source/sata_nv.c:51:#define DRV_VERSION                        "0.6"
RHEL4_U3/source/sata_nv.c:51:#define DRV_VERSION                        "0.10"
SLES9_SP1/source/sata_nv.c:51:#define DRV_VERSION                       "0.10"
SLES9_SP2/source/sata_nv.c:51:#define DRV_VERSION                       "0.10"
SLES9_SP3/source/sata_nv.c:51:#define DRV_VERSION                       "0.10"

TO compound the(my) problem, the openSUSE kernel source sata_nv.c code is claiming 0.8 revision but it contains 0.10 changelog info ...
I don't know what it really is yet.

I did not look at others...
I wish I had never looked at any of this stuff. Ignorance is bliss...

Thanks again for taking the time to be here & respond and special thanks to nVidia for the years of linux support. I still have not bought any of that other major brand because of it(for either Linux or MS OS systems).

Ric

[netllama]

We plan to address the seemingly different code yet same version issue in a future update. As you theorized, the functionality is the same, however some (older) distributions had special build requirements which required driver code changes to work around.

Thanks,
Lonni

[fhj52]

Originally Posted by netllama

We plan to address the seemingly different code yet same version issue in a future update. As you theorized, the functionality is the same, however some (older) distributions had special build requirements which required driver code changes to work around.

Thanks,
Lonni

" seemingly different code " : say What?!?
There is no "seemingly" to it. The code is different. .(period)

Nvidia knows that is wrong way to do things. Old boxen that need special code to work are (almost) never supported by special releases of any software. The code, the latest code/version, is supposed to address any such issue. Otherwise it is a hack and nothing but a hack. Hacks are okay (by me) but they don't belong in the mainstream and certainly should NEVER have the same version or name as the mainstream code. Nobody releases hacks with same name and version number as the main code!

Pull the files, change the hacked versions names to forcedeth_nv.c and sata_nv2.c (and probably the same for the others...). Rev the number from 1.11 to 1.12 and put 'em back up.

It's an hour, at most, for the whole thing.... no big deal.
Edit: well, no big deal at this time ... .


Ric

[PrakashP]

I still haven't read anything about the nforce2 APU problem with NVidia's new "out-sourcing tactics".

Are we going to get NCQ for sata_nv? Jeff Garzik seems to have gotten docs under NDA, but he already expressed that it is low priority for him...

[Lithorus]

Originally Posted by netllama

Lithorus,
SuSE and Redhat are explicitly 'supported' as NVIDIA's enterprise customers have requested support for those distributions. The source for the drivers is also included in the 1.11 package for anyone running a different distribution, should they wish to build the drivers. For a recent (2.6.16.x or later) kernel, all you should need to do is replace the 'default' source with that included in the 1.11 package (or my sticky post) and (re)build the kernel.

Don't get me wrong I'm deeply gratefull for the support that there is already. It's just when you have another distro which is not supported directly and you open the zip file you are only met with directories related to distros other than your own. When met with only these options it can be quite confusing as to which you should use.
Supporting every distro out there is ofcourse not an option, which is why IMHO a "generic" folder would be nice to have. I've seen many other open source drivers out there which are able to compile a new module without the need for recompiling the whole kernel. Much nicer to maintain from a users perspective.
Ofcourse if (which I suspect) the main goal is to provide source for the kernel team for them to incorporate into the main tree, users shouldn't need to deal with these kinds of things in the first place..

[R3MF]

as a 430 MCP owner I do hope the nVidia Linux drivers get sorted soon.

i was totally unable to get SUSE 10.1 to behave with an Asus 6150/430 motherboard.

[PrakashP]

Originally Posted by R3MF

i was totally unable to get SUSE 10.1 to behave with an Asus 6150/430 motherboard.

No problem here after minor kernel patching. Take a look into the big 430/6150 thread, last pages.