[dpdk-dev] [PATCH] eal: force IOVA PA mode if KNI module inserted

[santosh]

Hi Ferruh,


On Thursday 02 November 2017 05:36 AM, Ferruh Yigit wrote:
> Fix kernel crash with KNI because KNI requires physical addresses.
>
> When IOVA VA mode used, memzones and mbufs physical address fields
> contain virtual addresses. But KNI relies on these fields to enable
> kernel access for buffers. Those fields having virtual address cause
> crash in kernel.
>
> This is a workaround until KNI fixed properly to work with virtual
> addresses.
>
> Fixes: 72d013644bd6 ("mem: honor IOVA mode in malloc virt2phy")
>
> Signed-off-by: Ferruh Yigit <ferruh.yigit at intel.com>
> ---
> Cc: Jianfeng Tan <jianfeng.tan at intel.com>
> Cc: Jerin Jacob <jerin.jacob at caviumnetworks.com>
> Cc: Santosh Shukla <santosh.shukla at caviumnetworks.com>
> Cc: Thomas Monjalon <thomas at monjalon.net>
>
> This patch superseded following one:
> http://dpdk.org/dev/patchwork/patch/31071/
> ---
>  lib/librte_eal/linuxapp/eal/eal.c | 9 +++++++++
>  1 file changed, 9 insertions(+)
>
> diff --git a/lib/librte_eal/linuxapp/eal/eal.c b/lib/librte_eal/linuxapp/eal/eal.c
> index 017c402ed..29912a4e5 100644
> --- a/lib/librte_eal/linuxapp/eal/eal.c
> +++ b/lib/librte_eal/linuxapp/eal/eal.c
> @@ -808,6 +808,15 @@ rte_eal_init(int argc, char **argv)
>  	/* autodetect the iova mapping mode (default is iova_pa) */
>  	rte_eal_get_configuration()->iova_mode = rte_bus_get_iommu_class();
>  
> +	/* Workaround for KNI which requires physical address to work */
> +	if (rte_eal_get_configuration()->iova_mode == RTE_IOVA_VA &&
> +			rte_eal_check_module("rte_kni") == 1) {
> +		rte_eal_get_configuration()->iova_mode = RTE_IOVA_PA;
> +		RTE_LOG(WARNING, EAL,
> +			"Some devices want iova as va but pa will be used because.. "
> +			"KNI module inserted\n");
> +	}
> +
>  	if (internal_config.no_hugetlbfs == 0 &&
>  			internal_config.process_type != RTE_PROC_SECONDARY &&
>  			eal_hugepage_info_init() < 0) {

Rather checking for KNI module at linuxapp/eal, I was suggesting to move
KNI detection code in bus layer like below:

diff --git a/drivers/bus/pci/linux/pci.c b/drivers/bus/pci/linux/pci.c
index cdf8106..971586c 100644
--- a/drivers/bus/pci/linux/pci.c
+++ b/drivers/bus/pci/linux/pci.c
@@ -586,11 +586,13 @@ rte_pci_get_iommu_class(void)
        bool is_vfio_noiommu_enabled = true;
        bool has_iova_va;
        bool is_bound_uio;
+       bool has_kni;
 
        is_bound = pci_one_device_is_bound();
        if (!is_bound)
                return RTE_IOVA_DC;
 
+       has_kni = rte_eal_check_module("rte_kni");
        has_iova_va = pci_one_device_has_iova_va();
        is_bound_uio = pci_one_device_bound_uio();
 #ifdef VFIO_PRESENT
@@ -598,7 +600,7 @@ rte_pci_get_iommu_class(void)
                                        true : false;
 #endif
 
-       if (has_iova_va && !is_bound_uio && !is_vfio_noiommu_enabled)
+       if (has_iova_va && !is_bound_uio && !has_kni && !is_vfio_noiommu_enabled)
                return RTE_IOVA_VA;
 
        if (has_iova_va) {
@@ -607,6 +609,9 @@ rte_pci_get_iommu_class(void)
                        RTE_LOG(WARNING, EAL, "vfio-noiommu mode configured\n");
                if (is_bound_uio)
                        RTE_LOG(WARNING, EAL, "few device bound to UIO\n");
+               if (has_kni)
+                       RTE_LOG(WARNING, EAL, "Some devices want iova as va but pa will be used because.."
+                                       "KNI module inserted\n");
        }
 
        return RTE_IOVA_PA;
It builds but I haven;t tested, can you please give it a try.

Thanks.