Using virtio ethdev ports as non-root
cover Using virtio ethdev ports as non-root
I managed to run testpmd as non-root with virtio ports after quite some tries.
Sending a first RFC with the (not so big) changes that are required on the DPDK side.
I will come with a little howto for the non-RFC series for people interested in this, since I had to align quite some stars to make it work.
Signature · 13 lines
David Marchand David Marchand (2): bus/pci: check IO permissions for UIO only net/virtio: do not require IO permissions drivers/bus/pci/bsd/pci.c | 5 +++++ drivers/bus/pci/linux/pci.c | 10 ++++++++++ drivers/net/virtio/virtio_ethdev.c | 8 ++------ 3 files changed, 17 insertions(+), 6 deletions(-) -- 1.8.3.1
RFC PATCH 1/2 bus/pci: check IO permissions for UIO only
On x86, calling inb/outb special instructions (used in uio ioport read/write parts) is only possible if the right IO permissions has been granted.
The only user of this API (the net/virtio pmd) checks this unconditionnaly but this should be hidden by the rte_pci_ioport API itself and only checked when the device is bound to a UIO driver.
--- drivers/bus/pci/bsd/pci.c | 5 +++++ drivers/bus/pci/linux/pci.c | 10 ++++++++++ 2 files changed, 15 insertions(+)
drivers/bus/pci/bsd/pci.c +5 -0
@@ -540,6 +540,11 @@ rte_pci_ioport_map(struct rte_pci_device *dev, int bar,switch (dev->kdrv) {#if defined(RTE_ARCH_X86 )case RTE_KDRV_NIC_UIO : +if (rte_eal_iopl_init () !=0 ) { +RTE_LOG (DEBUG ,EAL ,"%s(): cannot gain io permissions\n" , + __func__); +return -1 ; + }if ((uintptr_t ) dev->mem_resource[bar].addr <=UINT16_MAX ) { p->base = (uintptr_t )dev->mem_resource[bar].addr; ret =0 ;
drivers/bus/pci/linux/pci.c +10 -0
@@ -661,6 +661,12 @@ pci_ioport_map(struct rte_pci_device *dev, int bar __rte_unused, dev->addr.domain, dev->addr.bus, dev->addr.devid, dev->addr.function); +if (rte_eal_iopl_init () !=0 ) { +RTE_LOG (DEBUG ,EAL ,"%s(): cannot gain io permissions\n" , + __func__); +return -1 ; + } + fp =fopen ("/proc/ioports" ,"r" );if (fp ==NULL ) {RTE_LOG (ERR ,EAL ,"%s(): can't open ioports\n" , __func__);
@@ -718,7 +724,11 @@ rte_pci_ioport_map(struct rte_pci_device *dev, int bar,break ;#endif case RTE_KDRV_IGB_UIO : +#if defined(RTE_ARCH_X86 ) + ret =pci_ioport_map (dev, bar, p); +#else ret =pci_uio_ioport_map (dev, bar, p); +#endif break ;case RTE_KDRV_UIO_GENERIC :#if defined(RTE_ARCH_X86 )
Signature · 1 line
1.8.3.1
RFC PATCH 2/2 net/virtio: do not require IO permissions
In reply to David MarchandOn x86, iopl permissions are only needed when the virtio devices are bound to a uio kernel module.
When running a dpdk application as non root, the virtio driver was refusing to register because of this check while it could work when binding the device to vfio (requires to have a vIOMMU configured).
We still need to call rte_eal_iopl_init() in the constructor so that the interrupt thread would inherit this permission in the case it could be used with UIO later.
Log a warning message for the user to understand what is wrong.
--- drivers/net/virtio/virtio_ethdev.c | 8 ++------ 1 file changed, 2 insertions(+), 6 deletions(-)
drivers/net/virtio/virtio_ethdev.c +2 -6
@@ -1977,11 +1977,6 @@ exit: static int eth_virtio_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,struct rte_pci_device *pci_dev) { -if (rte_eal_iopl_init () !=0 ) { -PMD_INIT_LOG (ERR ,"IOPL call failed - cannot use virtio PMD" ); -return 1 ; - } -/* virtio pmd skips probe if device needs to work in vdpa mode */ if (vdpa_mode_selected (pci_dev->device.devargs))return 1 ;
@@ -2013,7 +2008,8 @@ static struct rte_pci_driver rte_virtio_pmd = {RTE_INIT (rte_virtio_pmd_init) { -rte_eal_iopl_init (); +if (rte_eal_iopl_init () !=0 ) +PMD_INIT_LOG (WARNING ,"IOPL call failed - virtio devices won't be functional if bound to UIO drivers" );rte_pci_register (&rte_virtio_pmd); }
Signature · 1 line
1.8.3.1