| Message ID | 20110922213522.26654.59301.stgit@gitlad.jf.intel.com |
|---|---|
| State | RFC, archived |
| Delegated to: | David Miller |
| Headers | show |
On Thu, 2011-09-22 at 14:35 -0700, Greg Rose wrote: > The only currently supported method of configuring the number of VFs > is through the max_vfs module parameter. This method is inadequate to > support scenarios in which the user might wish to have varying numbers > of VFs per PF. There is additional support for drivers that want to > partition some driver resources to additional net devices and for > configuring the number of Tx/Rx queue pairs per VF. > > Signed-off-by: Greg Rose <gregory.v.rose@intel.com> > --- > > include/linux/ethtool.h | 30 +++++++++++++++++++++++++++++- > net/core/ethtool.c | 38 ++++++++++++++++++++++++++++++++++++++ > 2 files changed, 67 insertions(+), 1 deletions(-) > > diff --git a/include/linux/ethtool.h b/include/linux/ethtool.h > index 45f00b6..448730f 100644 > --- a/include/linux/ethtool.h > +++ b/include/linux/ethtool.h > @@ -720,6 +720,27 @@ enum ethtool_sfeatures_retval_bits { > #define ETHTOOL_F_WISH (1 << ETHTOOL_F_WISH__BIT) > #define ETHTOOL_F_COMPAT (1 << ETHTOOL_F_COMPAT__BIT) > > +enum ethtool_iov_modes { > + ETHTOOL_IOV_MODE_NONE, > + ETHTOOL_IOV_MODE_SRIOV, > + ETHTOOL_IOV_MODE_NETDEVS, > +}; > + > +#define ETHTOOL_IOV_CMD_CONFIGURE_SRIOV 1 > +#define ETHTOOL_IOV_CMD_CONFIGURE_NETDEVS 2 > +#define ETHTOOL_IOV_CMD_CONFIGURE_VF_QUEUES 3 > + > +struct ethtool_iov_set_cmd { > + __u32 cmd; > + __u32 set_cmd; > + __u32 cmd_param; > +}; Why introduce sub-commands? > +struct ethtool_iov_get_cmd { > + __u32 cmd; > + __u32 mode; > +}; Why is it only possible to get the mode? It really seems to make more sense to group these three parameters together and get/set them all at once. But if you can justify making them separate then struct ethtool_value is the type to use. In any case, any new types or macros need comments. [...] > +static int ethtool_iov_get_command(struct net_device *dev, > + void __user *useraddr) > +{ > + int ret; > + struct ethtool_iov_get_cmd iov_cmd; > + > + if (!dev->ethtool_ops->iov_get_cmd) > + return -EOPNOTSUPP; > + if (copy_from_user(&iov_cmd, useraddr, sizeof(iov_cmd))) > + return -EFAULT; > + > + ret = dev->ethtool_ops->iov_get_cmd(dev, &iov_cmd); > + if (!ret) > + ret = copy_to_user(useraddr, &iov_cmd, sizeof(iov_cmd)); if (!ret && copy_to_user(...)) ret = -EFAULT; > + return ret; > +} [...]
> -----Original Message----- > From: Ben Hutchings [mailto:bhutchings@solarflare.com] > Sent: Friday, October 07, 2011 5:38 PM > To: Rose, Gregory V > Cc: netdev@vger.kernel.org > Subject: Re: [RFC PATCH V2 1/2] net/core/ethtool: New Commands to > Configure IOV features > > On Thu, 2011-09-22 at 14:35 -0700, Greg Rose wrote: > > The only currently supported method of configuring the number of VFs > > is through the max_vfs module parameter. This method is inadequate to > > support scenarios in which the user might wish to have varying numbers > > of VFs per PF. There is additional support for drivers that want to > > partition some driver resources to additional net devices and for > > configuring the number of Tx/Rx queue pairs per VF. > > > > Signed-off-by: Greg Rose <gregory.v.rose@intel.com> > > --- > > > > include/linux/ethtool.h | 30 +++++++++++++++++++++++++++++- > > net/core/ethtool.c | 38 ++++++++++++++++++++++++++++++++++++++ > > 2 files changed, 67 insertions(+), 1 deletions(-) > > > > diff --git a/include/linux/ethtool.h b/include/linux/ethtool.h > > index 45f00b6..448730f 100644 > > --- a/include/linux/ethtool.h > > +++ b/include/linux/ethtool.h > > @@ -720,6 +720,27 @@ enum ethtool_sfeatures_retval_bits { > > #define ETHTOOL_F_WISH (1 << ETHTOOL_F_WISH__BIT) > > #define ETHTOOL_F_COMPAT (1 << ETHTOOL_F_COMPAT__BIT) > > > > +enum ethtool_iov_modes { > > + ETHTOOL_IOV_MODE_NONE, > > + ETHTOOL_IOV_MODE_SRIOV, > > + ETHTOOL_IOV_MODE_NETDEVS, > > +}; > > + > > +#define ETHTOOL_IOV_CMD_CONFIGURE_SRIOV 1 > > +#define ETHTOOL_IOV_CMD_CONFIGURE_NETDEVS 2 > > +#define ETHTOOL_IOV_CMD_CONFIGURE_VF_QUEUES 3 > > + > > +struct ethtool_iov_set_cmd { > > + __u32 cmd; > > + __u32 set_cmd; > > + __u32 cmd_param; > > +}; > > Why introduce sub-commands? For maximum flexibility. It's difficult (to me anyway) to foresee all the things we might want to configure for I/O virtualization features. > > > +struct ethtool_iov_get_cmd { > > + __u32 cmd; > > + __u32 mode; > > +}; > > Why is it only possible to get the mode? To my way of thinking it was the only interesting thing to query. I suppose I could also return the number of VFs or net devices along with the mode. > > It really seems to make more sense to group these three parameters > together and get/set them all at once. But if you can justify making > them separate then struct ethtool_value is the type to use. I can do that. > > In any case, any new types or macros need comments. OK. > > [...] > > +static int ethtool_iov_get_command(struct net_device *dev, > > + void __user *useraddr) > > +{ > > + int ret; > > + struct ethtool_iov_get_cmd iov_cmd; > > + > > + if (!dev->ethtool_ops->iov_get_cmd) > > + return -EOPNOTSUPP; > > + if (copy_from_user(&iov_cmd, useraddr, sizeof(iov_cmd))) > > + return -EFAULT; > > + > > + ret = dev->ethtool_ops->iov_get_cmd(dev, &iov_cmd); > > + if (!ret) > > + ret = copy_to_user(useraddr, &iov_cmd, sizeof(iov_cmd)); > > if (!ret && copy_to_user(...)) > ret = -EFAULT; > > > + return ret; > > +} > [...] > > -- > Ben Hutchings, Staff Engineer, Solarflare > Not speaking for my employer; that's the marketing department's job. > They asked us to note that Solarflare product names are trademarked.
diff --git a/include/linux/ethtool.h b/include/linux/ethtool.h index 45f00b6..448730f 100644 --- a/include/linux/ethtool.h +++ b/include/linux/ethtool.h @@ -720,6 +720,27 @@ enum ethtool_sfeatures_retval_bits { #define ETHTOOL_F_WISH (1 << ETHTOOL_F_WISH__BIT) #define ETHTOOL_F_COMPAT (1 << ETHTOOL_F_COMPAT__BIT) +enum ethtool_iov_modes { + ETHTOOL_IOV_MODE_NONE, + ETHTOOL_IOV_MODE_SRIOV, + ETHTOOL_IOV_MODE_NETDEVS, +}; + +#define ETHTOOL_IOV_CMD_CONFIGURE_SRIOV 1 +#define ETHTOOL_IOV_CMD_CONFIGURE_NETDEVS 2 +#define ETHTOOL_IOV_CMD_CONFIGURE_VF_QUEUES 3 + +struct ethtool_iov_set_cmd { + __u32 cmd; + __u32 set_cmd; + __u32 cmd_param; +}; + +struct ethtool_iov_get_cmd { + __u32 cmd; + __u32 mode; +}; + #ifdef __KERNEL__ #include <linux/rculist.h> @@ -879,6 +900,8 @@ bool ethtool_invalid_flags(struct net_device *dev, u32 data, u32 supported); * and flag of the device. * @get_dump_data: Get dump data. * @set_dump: Set dump specific flags to the device. + * @iov_set_cmd: Set IOV parameters - number of VFs or VM Queues + * @iov_get_cmd: Get current IOV parameters and configuration * * All operations are optional (i.e. the function pointer may be set * to %NULL) and callers must take this into account. Callers must @@ -956,7 +979,10 @@ struct ethtool_ops { int (*get_dump_data)(struct net_device *, struct ethtool_dump *, void *); int (*set_dump)(struct net_device *, struct ethtool_dump *); - + int (*iov_set_cmd)(struct net_device *, + struct ethtool_iov_set_cmd *); + int (*iov_get_cmd)(struct net_device *, + struct ethtool_iov_get_cmd *); }; #endif /* __KERNEL__ */ @@ -1030,6 +1056,8 @@ struct ethtool_ops { #define ETHTOOL_SET_DUMP 0x0000003e /* Set dump settings */ #define ETHTOOL_GET_DUMP_FLAG 0x0000003f /* Get dump settings */ #define ETHTOOL_GET_DUMP_DATA 0x00000040 /* Get dump data */ +#define ETHTOOL_IOV_SET_CMD 0x00000041 /* Set IOV parameters */ +#define ETHTOOL_IOV_GET_CMD 0x00000042 /* Get IOV parameters */ /* compatibility with older code */ #define SPARC_ETH_GSET ETHTOOL_GSET diff --git a/net/core/ethtool.c b/net/core/ethtool.c index f444817..a469e40 100644 --- a/net/core/ethtool.c +++ b/net/core/ethtool.c @@ -1633,6 +1633,38 @@ out: return ret; } +static int ethtool_iov_set_command(struct net_device *dev, + void __user *useraddr) +{ + struct ethtool_iov_set_cmd iov_cmd; + + if (!dev->ethtool_ops->iov_set_cmd) + return -EOPNOTSUPP; + if (copy_from_user(&iov_cmd, useraddr, sizeof(iov_cmd))) + return -EFAULT; + + return dev->ethtool_ops->iov_set_cmd(dev, &iov_cmd); +} + +static int ethtool_iov_get_command(struct net_device *dev, + void __user *useraddr) +{ + int ret; + struct ethtool_iov_get_cmd iov_cmd; + + if (!dev->ethtool_ops->iov_get_cmd) + return -EOPNOTSUPP; + if (copy_from_user(&iov_cmd, useraddr, sizeof(iov_cmd))) + return -EFAULT; + + ret = dev->ethtool_ops->iov_get_cmd(dev, &iov_cmd); + if (!ret) + ret = copy_to_user(useraddr, &iov_cmd, sizeof(iov_cmd)); + + return ret; +} + + /* The main entry point in this file. Called from net/core/dev.c */ int dev_ethtool(struct net *net, struct ifreq *ifr) @@ -1855,6 +1887,12 @@ int dev_ethtool(struct net *net, struct ifreq *ifr) case ETHTOOL_GET_DUMP_DATA: rc = ethtool_get_dump_data(dev, useraddr); break; + case ETHTOOL_IOV_SET_CMD: + rc = ethtool_iov_set_command(dev, useraddr); + break; + case ETHTOOL_IOV_GET_CMD: + rc = ethtool_iov_get_command(dev, useraddr); + break; default: rc = -EOPNOTSUPP; }
The only currently supported method of configuring the number of VFs is through the max_vfs module parameter. This method is inadequate to support scenarios in which the user might wish to have varying numbers of VFs per PF. There is additional support for drivers that want to partition some driver resources to additional net devices and for configuring the number of Tx/Rx queue pairs per VF. Signed-off-by: Greg Rose <gregory.v.rose@intel.com> --- include/linux/ethtool.h | 30 +++++++++++++++++++++++++++++- net/core/ethtool.c | 38 ++++++++++++++++++++++++++++++++++++++ 2 files changed, 67 insertions(+), 1 deletions(-) -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html