[dpdk-dev] DMAR errors when running testpmd on kernel >= 3.15

[dpdk-dev] DMAR errors when running testpmd on kernel >= 3.15

[didier.pallard]

Hello,

I am testing testpmd on recent kernels (>= 3.15) and i have an IOMMU 
problem (there are several commits around IOMMU in v3.15, they may be 
related with this problem):

Hardware:
Platform: Intel S2600IP
CPU: Intel(R) Xeon(R) CPU E5-2697 v2 @ 2.70GHz
BIOS version: SE5C600.86B.02.03.0003.041920141333
NICs:
     05:00.0 Ethernet controller: Intel Corporation 82599ES 10-Gigabit 
SFI/SFP+ Network Connection (rev 01)
     05:00.1 Ethernet controller: Intel Corporation 82599ES 10-Gigabit 
SFI/SFP+ Network Connection (rev 01)
     07:00.0 Ethernet controller: Intel Corporation 82599ES 10-Gigabit 
SFI/SFP+ Network Connection (rev 01)
     07:00.1 Ethernet controller: Intel Corporation 82599ES 10-Gigabit 
SFI/SFP+ Network Connection (rev 01)

Software:
Linux distribution: Fedora 20
Linux kernel: >= 3.15 (tested with 3.15.6-200 fc20 and 3.16.0-rc6 vanilla)
Linux boot line: BOOT_IMAGE=/boot/vmlinuz-3.15.6-200.fc20.x86_64 
root=/dev/sda5 console=tty1 console=ttyS0,115200n8 intel_iommu=on iommu=pt
DPDK version: 1.7.0 (also tested on master, with same result)

Test: I run the following commands to start the test:
echo 1024 >  /proc/sys/vm/nr_hugepages
mount -t hugetlbfs nodev /mnt/huge
modprobe uio
insmod dpdk/build/kmod/igb_uio.ko
dpdk/tools/dpdk_nic_bind.py --bind=igb_uio 05:00.0 05:00.1 07:00.0 07:00.1
dpdk/build/app/testpmd  --huge-dir /mnt/huge -n 4 -c 0x0001fe0001ff 
--socket-mem=512,0 -- --socket-num=0 -i --rxq=4 --txq=16 --portmask=0xf 
--rxd=128 --rxfreet=32 --rxpt=8 --rxht=8 --rxwt=0 --txd=512 --txfreet=32 
--txpt=32 --txht=0  --txwt=0 --txrst=32 --txqflags=0xF00 --burst=32 
--mbcache=250 --total-num-mbufs=64000 --coremask=0x0001fe0001fe

When i start the traffic, i have the following messages displayed on screen:
[ 1819.770528] dmar: DMAR:[DMA Read] Request device [05:00.0] fault addr 
3c4160000
[ 1819.770528] DMAR:[fault reason 02] Present bit in context entry is clear
[ 1819.786243] dmar: DMAR:[DMA Read] Request device [07:00.1] fault addr 
3ce399000
[ 1819.786243] DMAR:[fault reason 02] Present bit in context entry is clear
[ 1819.801955] dmar: DMAR:[DMA Read] Request device [05:00.1] fault addr 
3ce119000
[ 1819.801955] DMAR:[fault reason 02] Present bit in context entry is clear
[ 1819.817668] dmar: DMAR:[DMA Read] Request device [07:00.0] fault addr 
3ce269000
[ 1819.817668] DMAR:[fault reason 02] Present bit in context entry is clear

and nothing flows through the testpmd ports.

The problem does not appear:
   - with kernel older than 3.15
   - if interfaces are bound using vfio-pci instead of igb_uio
   - if iommu is disabled in bootline (remove intel_iommu=on iommu=pt 
from kernel boot line)

Sometimes, only two ports instead of four have DMAR problems, the two 
others forward packets normally.

Does someone know what may happen?

thanks
didier