334.21 driver cause kernel null pointer deference

Hi,
When i try to see a video with both mplayer or with vlc, a kernel null pointer deference occurs. I use vdpau video output. When I stop the video reproduction , mplayer crash.
I use Fedora 20 with 3.13.5 vanilla kernel,

dmesg say:

260.428505] BUG: unable to handle kernel NULL pointer dereference at 0000000000000088
[  260.428510] IP: [<ffffffff814a13d7>] iommu_no_mapping+0x7/0x100
[  260.428514] PGD 0 
[  260.428515] Oops: 0000 [#1] PREEMPT SMP 
[  260.428517] Modules linked in: nvidia_uvm(PO) rfcomm ccm ip6t_rpfilter ip6t_REJECT xt_conntrack bnep vboxpci(O) vboxnetadp(O) vboxnetflt(O) ip6table_nat nf_conntrack_ipv6 nf_defrag_ipv6 nf_nat_ipv6 ip6table_mangle ip6table_security ip6table_raw ip6table_filter ip6_tables iptable_nat nf_conntrack_ipv4 nf_defrag_ipv4 nf_nat_ipv4 nf_nat nf_conntrack iptable_mangle iptable_security iptable_raw vboxdrv(O) nvidia(PO) arc4 iwldvm mac80211 btusb bluetooth x86_pkg_temp_thermal coretemp kvm_intel kvm dell_wmi sparse_keymap dell_laptop dcdbas sdhci_pci sdhci mmc_core uvcvideo videobuf2_vmalloc wmi videobuf2_memops videobuf2_core i2c_i801 videodev media iwlwifi cfg80211 joydev iTCO_wdt iTCO_vendor_support crct10dif_pclmul crc32_pclmul crc32c_intel rfkill ghash_clmulni_intel video firewire_ohci firewire_core
[  260.428539]  crc_itu_t
[  260.428541] CPU: 5 PID: 2124 Comm: mplayer Tainted: P           O 3.13.5 #1
[  260.428542] Hardware name: Dell Inc. Precision M4700/082H3V, BIOS A05 10/08/2012
[  260.428543] task: ffff88022177ee00 ti: ffff8801e449e000 task.ti: ffff8801e449e000
[  260.428544] RIP: 0010:[<ffffffff814a13d7>]  [<ffffffff814a13d7>] iommu_no_mapping+0x7/0x100
[  260.428546] RSP: 0018:ffff8801e449fd80  EFLAGS: 00010246
[  260.428547] RAX: 00000001dee38000 RBX: 0000000000000000 RCX: 0000000000000000
[  260.428547] RDX: 0000000000000001 RSI: ffff8801dec1e898 RDI: 0000000000000000
[  260.428548] RBP: ffffea0000000000 R08: 0000000000000000 R09: 0000000000000001
[  260.428549] R10: ffffffff814a14d0 R11: 0000000000000293 R12: ffff8801dec1e898
[  260.428550] R13: 0000000000000000 R14: ffff8801dec1e840 R15: ffff8801dec1e880
[  260.428551] FS:  00007fdc67176900(0000) GS:ffff88022dd40000(0000) knlGS:0000000000000000
[  260.428552] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  260.428552] CR2: 0000000000000088 CR3: 00000001deeaf000 CR4: 00000000001407e0
[  260.428553] Stack:
[  260.428554]  0000000000000000 ffffea0000000000 ffff8801dec1e898 ffffffff814a14e5
[  260.428555]  0000000000000000 ffffea0000000000 000077ff80000000 0000000000000000
[  260.428557]  ffff8801dec1e840 ffff8801dec1e880 ffffffffa0cbca91 0000000000000098
[  260.428558] Call Trace:
[  260.428560]  [<ffffffff814a14e5>] ? intel_unmap_sg+0x15/0x120
[  260.428593]  [<ffffffffa0cbca91>] ? nv_free_system_pages+0xc1/0x3a0 [nvidia]
[  260.428616]  [<ffffffffa0cb6a35>] ? nv_free_pages+0xc5/0xd0 [nvidia]
[  260.428638]  [<ffffffffa0cb6d5e>] ? nvidia_close+0x31e/0x430 [nvidia]
[  260.428659]  [<ffffffffa0cbf7cf>] ? nvidia_frontend_close+0x3f/0x90 [nvidia]
[  260.428662]  [<ffffffff811492eb>] ? __fput+0x8b/0x200
[  260.428664]  [<ffffffff8108599f>] ? task_work_run+0x8f/0xd0
[  260.428666]  [<ffffffff81002a34>] ? do_notify_resume+0x74/0x80
[  260.428669]  [<ffffffff81618332>] ? int_signal+0x12/0x17
[  260.428670] Code: 31 56 00 48 89 de e8 49 9a c9 ff 5b 89 e8 5d 41 5c 41 5d c3 b8 f4 ff ff ff e9 46 ff ff ff 0f 0b 0f 1f 40 00 41 54 55 53 48 89 fb <48> 81 bf 88 00 00 00 c0 fc 85 81 0f 85 c8 00 00 00 48 8b 87 f8
[  260.428684] RIP  [<ffffffff814a13d7>] iommu_no_mapping+0x7/0x100
[  260.428685]  RSP <ffff8801e449fd80>
[  260.428686] CR2: 0000000000000088
[  260.428687] ---[ end trace 5c261e9668056b89 ]---

Thanks

It seems to be the same bug reported since 334.16.
I think your problem can be workarounded by adding “intel_iommu=off” as kernel parameter.

See the following thread for details:

Hi,
thank you for the suggestion.
The workaround seem work quite fine, but I see some glitch in google-chrome for example.
I like opengles support and the performance with CUDA, but stability for me is important. I think I make downgrade the driver.