331.13 BETA drivers feedback thread

Graphics / Linux Linux
21

Please don’t use the patch linked from the first post in this thread. At the very least, it should use “get_num_physpages()” rather than “get_num_physpages” (note the parentheses) because the latter uses the address of the get_num_physpages function as if it were the number it’s supposed to return, which is definitely wrong.

As I mentioned in a different post, just replacing num_physpages with get_num_physpages() isn’t quite right for one of the two places NV_NUM_PHYSPAGES is used, so please uses patches that just replace one with the other at your own risk. We’re working on a proper solution for Linux 3.11.

22

Any chance this driver fixes the “GPU has fallen off the bus” error and potential overheating (newer drivers cause my laptop to run warmer than it did, and with kernel 3.10, my laptop just shuts off after a minute)?

23

Tomasu, the “GPU has fallen off the bus” error is a symptom of many possible problems that run the gamut from hardware to software. That’s why it’s impossible to say definitively that “the bug” has been fixed or not. Many things change from one driver release to the next, especially when the release is the first from a new release branch, where there could be hundreds of changes that don’t make it into the changelog. Your best bet is to try it and see.

24

Ah. I figured it would be something fairly obvious as its something new, and happens with 3.10 and up. but yes, I presume a lot has changed in both the driver and the kernel between 3.9 and 3.11+.

25

For Linux 3.12 would be great also. Many devs here do testing on the latest kernels.

26

aplattner, check leigh123linux’s post on the first page of this thread. His patch is working fine for me with 3.12-git.

27

Agreed. It would be nice if the next beta would compile on 3.12 without any need for external patches. My new Haswell laptop with GT740m & Intel 7260 AC needs the 3.12 kernel to be able to stay consistently connected to my university’s WPA Enterprise PEAP wireless network as well as being able to work with bumblebee.

28

New patch for 3.12-rc4

diff -uNrp a/kernel/nv-drm.c b/kernel/nv-drm.c
--- a/kernel/nv-drm.c
+++ b/kernel/nv-drm.c
@@ -201,12 +201,16 @@ RM_STATUS NV_API_CALL nv_alloc_os_descri
     memset(&nv_obj->base, 0, sizeof(nv_obj->base));
     nv_obj->pages = pages;
 
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 12, 0)
+    drm_gem_private_object_init(nvl->drm, &nv_obj->base, size);
+#else
     ret = drm_gem_private_object_init(nvl->drm, &nv_obj->base, size);
     if (ret)
     {
-        status = RM_ERR_OPERATING_SYSTEM;
+       status = RM_ERR_OPERATING_SYSTEM;
         goto done;
     }
+#endif
 
     ret = drm_gem_handle_create(file_priv, &nv_obj->base, handle);
     if (ret)
diff -uNrp a/kernel/nv-linux.h b/kernel/nv-linux.h
--- a/kernel/nv-linux.h
+++ b/kernel/nv-linux.h
@@ -956,7 +956,11 @@ static inline int nv_execute_on_all_cpus
 #endif
 
 #if !defined(NV_VMWARE)
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 11, 0)
+#define NV_NUM_PHYSPAGES                get_num_physpages()
+#else
 #define NV_NUM_PHYSPAGES                num_physpages
+#endif
 #define NV_GET_CURRENT_PROCESS()        current->tgid
 #define NV_IN_ATOMIC()                  in_atomic()
 #define NV_LOCAL_BH_DISABLE()           local_bh_disable()
29

Thanks. The previous one still compiles the new rc4 also.
EDIT:
Actually this patch can compile older NVIDIA drivers while using Linux 3.12-rc4. Previous patch didn’t work.
I now have NVIDIA 319.60 running Arch 64-bit

30

I can see why you say not to use it.
I did some testing with the patch on and off using kernels 3.10, 3.11.3 and 3.12-rc4 and noticed this and the newer patch posted by leigh123linux caused huge performance degradation specially when using flash applications.

31

then what is the good patch for kernel 3.11?

greetings

32

nvidia not realsing the patch or the right driver but also tells people the patch is wrong. more shenanigans are on the way?

33

And for the first time, AMD have support for RC kernel before NVIDIA. Come on NVIDIA, get that ass moving. :)

34

On my GTX 650, I did not see tearing on VLC player, but I still see tearing when scaling up using mplayer.

Again, straight video seems fine - but maximizing the video adds a tearing phenomenon.

35

I recommend you stay on the latest kernel 3.10. Then you can try 331.13 without compilation issues. Also kernels 3.11 is not as reliable as 3.10.16…

But if you really wanna live with the latest bugs go ahead and use 3.11 and try the patch―it works.

36

My girlfriends computer has a GeForce 660ti and experienced tearing right from day one. My computer which uses 3 GTX 480 had tearing with the 26X drivers and has been good since the dark days of using that series of driver. Aaron suggested I try ForceCompositionPipeline = 1 on my GFs Kepler in the Kepler tearing thread and it fixed my tearing issue, although I do believe it decreases GPU performance. Using ForceFullCompositionPipeline increased lag very noticeably, although I could not prove it well using the Unigine Heaven benchmark.

37

Since 331.13 (no problems with 319.60 before) I get a lot of these X errors from an OpenGL based Qt program (Qt 4.7.4). All pixmaps/pngs on controls are missing. Any ideas on what might cause this? When I switch back to 319.60 everything is fine again.

X Error: BadMatch (invalid parameter attributes) 8
Extension: 149 (Uknown extension)
Minor opcode: 22 (Unknown request)
Resource id: 0x40002d
X Error: BadDrawable (invalid Pixmap or Window parameter) 9
Extension: 150 (Uknown extension)
Minor opcode: 4 (Unknown request)
Resource id: 0x40036a
X Error: GLXBadPixmap 165
Extension: 149 (Uknown extension)
Minor opcode: 16 (Unknown request)
Resource id: 0x40036a
X Error: BadMatch (invalid parameter attributes) 8
Extension: 149 (Uknown extension)
Minor opcode: 22 (Unknown request)
Resource id: 0x400033
…

38

What compositing window manager do you use (if at all)? E.g. on Ubuntu/Unity’s Compiz, the “Composite” plugin is set to unredirect fullscreen windows by default, though VLC and MPlayer are on an exception list by default. So maybe your setup unredirects fullscreen MPlayer but not VLC?

Edit: You might want to try that tearing fix

39

No compositing currently.