[dpdk-dev] [PATCH v2 00/41] Memory Hotplug for DPDK

[Pavan Nikhilesh]

On Thu, Mar 08, 2018 at 08:33:21PM +0000, Burakov, Anatoly wrote:
> On 08-Mar-18 8:11 PM, Burakov, Anatoly wrote:
> > On 08-Mar-18 2:36 PM, Burakov, Anatoly wrote:
> > > On 08-Mar-18 1:36 PM, Pavan Nikhilesh wrote:
> > > > Hi Anatoly,
> > > >
> > > > We are currently facing issues with running testpmd on thunderx
> > > > platform.
> > > > The issue seems to be with vfio
> > > >
> > > > EAL: Detected 24 lcore(s)
> > > > EAL: Detected 1 NUMA nodes
> > > > EAL: No free hugepages reported in hugepages-2048kB
> > > > EAL: Multi-process socket /var/run/.rte_unix
> > > > EAL: Probing VFIO support...
> > > > EAL: VFIO support initialized
> > > > EAL:   VFIO support not initialized
> > > >
> > > > <snip>
> > > >
> > > > EAL:   probe driver: 177d:a053 octeontx_fpavf
> > > > EAL: PCI device 0001:01:00.1 on NUMA socket 0
> > > > EAL:   probe driver: 177d:a034 net_thunderx
> > > > EAL:   using IOMMU type 1 (Type 1)
> > > > EAL:   cannot set up DMA remapping, error 22 (Invalid argument)
> > > > EAL:   0001:01:00.1 DMA remapping failed, error 22 (Invalid argument)
> > > > EAL: Requested device 0001:01:00.1 cannot be used
> > > > EAL: PCI device 0001:01:00.2 on NUMA socket 0
> > > > <snip>
> > > > testpmd: No probed ethernet devices
> > > > testpmd: create a new mbuf pool <mbuf_pool_socket_0>: n=251456,
> > > > size=2176, socket=0
> > > > testpmd: preferred mempool ops selected: ring_mp_mc
> > > > EAL:   VFIO support not initialized
> > > > EAL:   VFIO support not initialized
> > > > EAL:   VFIO support not initialized
> > > > Done
> > > >
> > > >
> > > > This is because rte_service_init() calls rte_calloc() before
> > > > rte_bus_probe() and vfio_dma_mem_map fails because iommu type is
> > > > not set.
> > > >
> > > > Call stack:
> > > > gdb) bt
> > > > #0  vfio_dma_mem_map (vaddr=281439006359552, iova=11274289152,
> > > > len=536870912, do_map=1) at
> > > > /root/clean/dpdk/lib/librte_eal/linuxapp/eal/eal_vfio.c:967
> > > > #1  0x00000000004fd974 in rte_vfio_dma_map
> > > > (vaddr=281439006359552, iova=11274289152, len=536870912) at
> > > > /root/clean/dpdk/lib/librte_eal/linuxapp/eal/eal_vfio.c:988
> > > > #2  0x00000000004fbe78 in vfio_mem_event_callback
> > > > (type=RTE_MEM_EVENT_ALLOC, addr=0xfff7a0000000, len=536870912)
> > > > at /root/clean/dpdk/lib/librte_eal/linuxapp/eal/eal_vfio.c:240
> > > > #3  0x00000000005070ac in eal_memalloc_notify
> > > > (event=RTE_MEM_EVENT_ALLOC, start=0xfff7a0000000, len=536870912)
> > > > at
> > > > /root/clean/dpdk/lib/librte_eal/common/eal_common_memalloc.c:177
> > > > #4  0x0000000000515c98 in try_expand_heap_primary
> > > > (heap=0xffffb7fb167c, pg_sz=536870912, elt_size=8192, socket=0,
> > > > flags=0, align=128, bound=0, contig=false) at
> > > > /root/clean/dpdk/lib/librte_eal/common/malloc_heap.c:247
> > > > #5  0x0000000000515e94 in try_expand_heap (heap=0xffffb7fb167c,
> > > > pg_sz=536870912, elt_size=8192, socket=0, flags=0, align=128,
> > > > bound=0, contig=false) at
> > > > /root/clean/dpdk/lib/librte_eal/common/malloc_heap.c:327
> > > > #6  0x00000000005163a0 in alloc_more_mem_on_socket
> > > > (heap=0xffffb7fb167c, size=8192, socket=0, flags=0, align=128,
> > > > bound=0, contig=false) at
> > > > /root/clean/dpdk/lib/librte_eal/common/malloc_heap.c:455
> > > > #7  0x0000000000516514 in heap_alloc_on_socket (type=0x85bf90
> > > > "rte_services", size=8192, socket=0, flags=0, align=128,
> > > > bound=0, contig=false) at
> > > > /root/clean/dpdk/lib/librte_eal/common/malloc_heap.c:491
> > > > #8  0x0000000000516664 in malloc_heap_alloc (type=0x85bf90
> > > > "rte_services", size=8192, socket_arg=-1, flags=0, align=128,
> > > > bound=0, contig=false) at
> > > > /root/clean/dpdk/lib/librte_eal/common/malloc_heap.c:527
> > > > #9  0x0000000000513b54 in rte_malloc_socket (type=0x85bf90
> > > > "rte_services", size=8192, align=128, socket_arg=-1) at
> > > > /root/clean/dpdk/lib/librte_eal/common/rte_malloc.c:54
> > > > #10 0x0000000000513bc8 in rte_zmalloc_socket (type=0x85bf90
> > > > "rte_services", size=8192, align=128, socket=-1) at
> > > > /root/clean/dpdk/lib/librte_eal/common/rte_malloc.c:72
> > > > #11 0x0000000000513c00 in rte_zmalloc (type=0x85bf90
> > > > "rte_services", size=8192, align=128) at
> > > > /root/clean/dpdk/lib/librte_eal/common/rte_malloc.c:81
> > > > #12 0x0000000000513c90 in rte_calloc (type=0x85bf90
> > > > "rte_services", num=64, size=128, align=128) at
> > > > /root/clean/dpdk/lib/librte_eal/common/rte_malloc.c:99
> > > > #13 0x0000000000518cec in rte_service_init () at
> > > > /root/clean/dpdk/lib/librte_eal/common/rte_service.c:81
> > > > #14 0x00000000004f55f4 in rte_eal_init (argc=3,
> > > > argv=0xfffffffff488) at
> > > > /root/clean/dpdk/lib/librte_eal/linuxapp/eal/eal.c:959
> > > > #15 0x000000000045af5c in main (argc=3, argv=0xfffffffff488) at
> > > > /root/clean/dpdk/app/test-pmd/testpmd.c:2483
> > > >
> > > >
> > > > Also, I have tried running with --legacy-mem but I'm stuck in
> > > > `pci_find_max_end_va` loop  because `rte_fbarray_find_next_used`
> > > > always return
> > > > 0. >
> > > > HugePages_Total:      15
> > > > HugePages_Free:       11
> > > > HugePages_Rsvd:        0
> > > > HugePages_Surp:        0
> > > > Hugepagesize:     524288 kB
> > > >
> > > > Call Stack:
> > > > (gdb) bt
> > > > #0  find_next (arr=0xffffb7fb009c, start=0, used=true) at
> > > > /root/clean/dpdk/lib/librte_eal/common/eal_common_fbarray.c:248
> > > > #1  0x00000000005132a8 in rte_fbarray_find_next_used
> > > > (arr=0xffffb7fb009c, start=0) at
> > > > /root/clean/dpdk/lib/librte_eal/common/eal_common_fbarray.c:700
> > > > #2  0x000000000052d030 in pci_find_max_end_va () at
> > > > /root/clean/dpdk/drivers/bus/pci/linux/pci.c:138
> > > > #3  0x0000000000530ab8 in pci_vfio_map_resource_primary
> > > > (dev=0xeae700) at
> > > > /root/clean/dpdk/drivers/bus/pci/linux/pci_vfio.c:499
> > > > #4  0x0000000000530ffc in pci_vfio_map_resource (dev=0xeae700)
> > > > at /root/clean/dpdk/drivers/bus/pci/linux/pci_vfio.c:601
> > > > #5  0x000000000052ce90 in rte_pci_map_device (dev=0xeae700) at
> > > > /root/clean/dpdk/drivers/bus/pci/linux/pci.c:75
> > > > #6  0x0000000000531a20 in rte_pci_probe_one_driver (dr=0x997e20
> > > > <rte_nicvf_pmd>, dev=0xeae700) at
> > > > /root/clean/dpdk/drivers/bus/pci/pci_common.c:164
> > > > #7  0x0000000000531c68 in pci_probe_all_drivers (dev=0xeae700)
> > > > at /root/clean/dpdk/drivers/bus/pci/pci_common.c:249
> > > > #8  0x0000000000531f68 in rte_pci_probe () at
> > > > /root/clean/dpdk/drivers/bus/pci/pci_common.c:359
> > > > #9  0x000000000050a140 in rte_bus_probe () at
> > > > /root/clean/dpdk/lib/librte_eal/common/eal_common_bus.c:98
> > > > #10 0x00000000004f55f4 in rte_eal_init (argc=1,
> > > > argv=0xfffffffff498) at
> > > > /root/clean/dpdk/lib/librte_eal/linuxapp/eal/eal.c:967
> > > > #11 0x000000000045af5c in main (argc=1, argv=0xfffffffff498) at
> > > > /root/clean/dpdk/app/test-pmd/testpmd.c:2483
> > > >
> > > > Am I missing something here?
> > >
> > > I'll look into those, thanks!
> > >
> > > Btw, i've now set up a github repo with the patchset applied:
> > >
> > > https://github.com/anatolyburakov/dpdk
> > >
> > > I will be pushing quick fixes there before spinning new revisions,
> > > so we can discover and fix bugs more rapidly. I'll fix compile
> > > issues reported earlier, then i'll take a look at your issues. The
> > > latter one seems like a typo, the former is probably a matter of
> > > moving things around a bit.
> > >
> > > (also, pull requests welcome if you find it easier to fix things
> > > yourself and submit patches against my tree!)
> > >
> > > Thanks for testing.
> > >
> >
> > I've looked into the failures.
> >
> > The VFIO one is not actually a failure. It only prints out errors
> > because rte_malloc is called before VFIO is initialized. However, once
> > VFIO *is* initialized, all of that memory would be added to VFIO, so
> > these error messages are harmless. Regardless, i've added a check to see
> > if init is finished before printing out those errors, so they won't be
> > printed out any more.
> >
> > Second one is a typo on my part that got lost in one of the rebases.
> >
> > I've pushed fixes for both into the github repo.
> >
>
> Although i do wonder where do the DMA remapping errors come from. The error
> message says "invalid argument", so that doesn't come from rte_service or
> anything to do with rte_malloc - this is us not providing valid arguments to
> VFIO. I'm not seeing these errors on my system. I'll check on others to be
> sure.

I have taken a look at the github tree the issues with VFIO are gone, Although
compilation issues with dpaa/dpaa2 are still present due to their dependency on
`rte_eal_get_physmem_layout`.

>
> --
> Thanks,
> Anatoly

Thanks,
Pavan