FYI - Nvidia legacy340 and -current

Lenard Spencer · 08-23-2015, 12:01 PM

I'm sure many of you have noticed that the Nvidia 340 kernel will no longer build in -current with the new 4.x kernel. I just happened to find a patch over on the nvidia-devel forums that fixes that. I just did it, and it works like a charm. (I'm using it now.)

First the patch (I called it nvidia-upstream.patch)

Code:

--- a/nv-pat.c.orig     2015-02-20 02:49:40.000000000 +0100
+++ b/nv-pat.c  2015-02-25 07:56:40.000000000 +0100
@@ -35,8 +35,13 @@
     unsigned long cr0 = read_cr0();
     write_cr0(((cr0 & (0xdfffffff)) | 0x40000000));
     wbinvd();
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 20, 0)
     *cr4 = read_cr4();
     if (*cr4 & 0x80) write_cr4(*cr4 & ~0x80);
+#else
+    *cr4 = __read_cr4();
+    if (*cr4 & 0x80) __write_cr4(*cr4 & ~0x80);
+#endif
     __flush_tlb();
 }
 
@@ -46,7 +51,11 @@
     wbinvd();
     __flush_tlb();
     write_cr0((cr0 & 0x9fffffff));
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 20, 0)
     if (cr4 & 0x80) write_cr4(cr4);
+#else
+    if (cr4 & 0x80) __write_cr4(cr4);
+#endif   
 }
  static int nv_determine_pat_mode(void)

Then I added this line to the slackbuild script:

Code:

(cd kernel || exit 1 

# Apply patch from upstream to build with 4.x kernels
patch -p1 < $CWD/nvidia-upstream.patch

  make SYSSRC=$KERNELPATH module || exit 1
  cd uvm
  make SYSSRC=$KERNELPATH module || exit 1
)

Note the patch statement is placed BETWEEN 'cd kernel' and the first 'make' statement.

Full credit to the guys on nvidia-devel for the patch, NOT me.

I hope this helps.

PS I almost forgot, only the kernel SB script needs the patch, NOT the driver.

kingbeowulf · 08-23-2015, 07:10 PM

Kernel 4.x may well kill off most of the legacy Nvidia drivers.

@Lenard Spencer, Do you have a link to the original Nvidia-dev post?

Nvidia-304.125 has issues as well:
http://slackblogs.blogspot.com/2015/...kernel-41.html
I'll add these to the my SBo to-do list for when Slackware-current goes stable.

EDIT: Also, allend posted a patch recently: http://www.linuxquestions.org/questi...ml#post5409843
Why doesn't anyone provide source references to patches?

mats_b_tegner · 08-24-2015, 01:29 AM

Here it is Ed:

https://devtalk.nvidia.com/default/t...functions-fix/

Mats

Panagiotis Nik · 08-24-2015, 01:41 AM

Just upgraded to 4.x kernel and my 340 kernel refuse to work...
Fixed with this patch!

Thanks @Lenard Spencer

ardvark71 · 08-24-2015, 02:30 AM

Hi Lenard...

Thank you for posting this.

Regards...

Lenard Spencer · 08-24-2015, 09:23 AM

I found another thread for the 304 series. I have a crude patch working, and I will post it as soon as I get it cleaned up.

ReaperX7 · 08-24-2015, 09:39 AM

Many thanks Lenard.

Lenard Spencer · 08-24-2015, 10:27 AM

Ok, here are the final patches.

First, for the 340 driver (I changed the names for clarity)

FILENAME: nvidia-legacy340-kernel4.patch

Code:

--- a/nv-pat.c.orig     2015-02-20 02:49:40.000000000 +0100
+++ b/nv-pat.c  2015-02-25 07:56:40.000000000 +0100
@@ -35,8 +35,13 @@
     unsigned long cr0 = read_cr0();
     write_cr0(((cr0 & (0xdfffffff)) | 0x40000000));
     wbinvd();
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 20, 0)
     *cr4 = read_cr4();
     if (*cr4 & 0x80) write_cr4(*cr4 & ~0x80);
+#else
+    *cr4 = __read_cr4();
+    if (*cr4 & 0x80) __write_cr4(*cr4 & ~0x80);
+#endif
     __flush_tlb();
 }
 
@@ -46,7 +51,11 @@
     wbinvd();
     __flush_tlb();
     write_cr0((cr0 & 0x9fffffff));
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 20, 0)
     if (cr4 & 0x80) write_cr4(cr4);
+#else
+    if (cr4 & 0x80) __write_cr4(cr4);
+#endif   
 }
  static int nv_determine_pat_mode(void)

Add these lines at line 82 in nvidia-legacy340-kernel.SlackBuild:

Code:

# Apply patch from nvidia-devel to build with 4.x kernels
patch -p1 < $CWD/nvidia-legacy340-kernel4.patch

For the 304 driver:

FILENAME: nvidia-legacy304-kernel4.patch

Code:

--- a/nv.c 2014-12-01 19:58:34.000000000 -0800
+++ b/nv.c 2015-07-09 16:45:09.997840781 -0700
@@ -2026,7 +2026,11 @@
     unsigned long i_arg
 )
 {
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 20, 0)
     return nv_kern_ioctl(file->f_dentry->d_inode, file, cmd, i_arg);
+#else
+    return nv_kern_ioctl(file_inode(file), file, cmd, i_arg);
+#endif
 }
 
 long nv_kern_compat_ioctl(
@@ -2035,7 +2039,11 @@
 unsigned long i_arg
 )
 {
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 20, 0)
     return nv_kern_ioctl(file->f_dentry->d_inode, file, cmd, i_arg);
+#else
+    return nv_kern_ioctl(file_inode(file), file, cmd, i_arg);
+#endif
 }
 
 /*
--- a/nv-pat.c 2014-12-01 19:58:34.000000000 -0800
+++ b/nv-pat.c 2015-07-09 16:48:00.069823572 -0700
@@ -35,8 +35,13 @@
     unsigned long cr0 = read_cr0();
     write_cr0(((cr0 & (0xdfffffff)) | 0x40000000));
     wbinvd();
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 20, 0)
     *cr4 = read_cr4();
     if (*cr4 & 0x80) write_cr4(*cr4 & ~0x80);
+#else
+    *cr4 = __read_cr4();
+    if (*cr4 & 0x80) __write_cr4(*cr4 & ~0x80);
+#endif
     __flush_tlb();
 }
 
@@ -46,7 +51,11 @@
     wbinvd();
     __flush_tlb();
     write_cr0((cr0 & 0x9fffffff));
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 20, 0)
     if (cr4 & 0x80) write_cr4(cr4);
+#else
+    if (cr4 & 0x80) __write_cr4(cr4);
+#endif   
 }
  static int nv_determine_pat_mode(void)

In nvidia-legacy304-kernel.SlackBuild, add these lines, also at line 82:

Code:

# Apply patch from nvidia-devel to build with 4.x kernels
patch -p1 < $CWD/nvidia-legacy304-kernel4.patch

I hope these help.

Maybe someday when I hear the men in the white coats sneaking up behind me, I may look at the 174 driver as well. (Yes, I still have a couple of 5200FX cards lying around!)

GazL · 08-24-2015, 10:32 AM

update:

My patch is no longer necessary as nvidia have released 304.128 which supports up to and including 4.1.y, so I've removed it.

Lenard Spencer · 08-24-2015, 11:40 AM

Kingbeowulf, you can probably put these in the Slack 14.1 SB repo as well, for the benefit of those running multiboot machines with -current (as I am). I just went back and checked them out in 14.1 with no issues.

Thanks for your SB scripts!

willysr · 08-24-2015, 05:20 PM

SBo is designed to be used against -stable releases, so it's not needed for now, not until next SLackware is released

ReaperX7 · 08-24-2015, 05:30 PM

It can always be added with the patch command commented out. Getting -Current ready for 14.2 is what has been ongoing for some time in the other topic, plus it could always be relegated to the main trunk or testing branch trunk (?) of the SBo tree while remaining out of 14.1 branch. Good question, but does it affect 14.1 and prevent building the kernel module?

willysr · 08-24-2015, 05:54 PM

perhaps adding it to Ponce's current branch is more appropriate. It will gets merged to master when we think next Slackware is close to be released

ReaperX7 · 08-24-2015, 09:20 PM

+1 on that.

kingbeowulf · 08-27-2015, 01:24 AM

Thanks to lenard, gazl and willysr for the nvidia patches and links. I hadn't noticed since I was fiddling with current on 2 laptops with Intel GPUs. I'll add them to my local SBo branch and track ponce's to get ready for when current goes stable "real soon now" <crosses fingers>.

Lenard, trying to wrestle support for older nvidia legacy might be an exercise in futility. Nothing prior to 304.xx has been updated for newer X.Org servers and Mesa since 2013. Your 5200FX cards might be a lost cause. I've stopped supporting 173-legacy and 96-legacy in SBo and dumped my old cards. "Help me, nouveau, you're my only hope."