| Message ID | 20171110185553.14657-3-jacob.e.keller@intel.com |
|---|---|
| State | Accepted |
| Delegated to: | Jeff Kirsher |
| Headers | show |
| Series | [1/3] ethtool: correct display of VF when showing vf/queue filters | expand |
> -----Original Message----- > From: Intel-wired-lan [mailto:intel-wired-lan-bounces@osuosl.org] On > Behalf Of Jacob Keller > Sent: Friday, November 10, 2017 10:56 AM > To: Intel Wired LAN <intel-wired-lan@lists.osuosl.org> > Subject: [Intel-wired-lan] [PATCH 3/3] ethtool: correct VF index values for > the ring_cookie parameter > > Instead of relying on the input being offset by 1, instead adjust the "vf" input > value by 1 so that it correctly targets the VF the user expects. Additionally, > update the displayed output to show the correct VF number. > > Signed-off-by: Jacob Keller <jacob.e.keller@intel.com> > --- > > This is the way the VF parameter should have worked from the beginning, > but unfortunately I did not code it as such. It may be considered a functional > change (since we already supported the vf parameter with the offset not > handled properly). I would consider this a bug fix, but it might be considered > a behavior change.... > > ethtool.8.in | 3 +-- > rxclass.c | 14 +++++++++++++- > 2 files changed, 14 insertions(+), 3 deletions(-) Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
diff --git a/ethtool.8.in b/ethtool.8.in index 6ad30650eb5b..76f86e6b6827 100644 --- a/ethtool.8.in +++ b/ethtool.8.in @@ -839,8 +839,7 @@ lB l. .TE .TP .BI vf \ N -Specifies the Virtual Function the filter applies to. A value of 0 indicates -the PF, and thus the VF index is offset by 1. Not compatible with action. +Specifies the Virtual Function the filter applies to. Not compatible with action. .TP .BI queue \ N Specifies the Rx queue to send packets to. Not compatible with action. diff --git a/rxclass.c b/rxclass.c index 144c8e5e76be..e05031b6a23f 100644 --- a/rxclass.c +++ b/rxclass.c @@ -251,9 +251,14 @@ static void rxclass_print_nfc_rule(struct ethtool_rx_flow_spec *fsp) u64 vf = ethtool_get_flow_spec_ring_vf(fsp->ring_cookie); u64 queue = ethtool_get_flow_spec_ring(fsp->ring_cookie); + /* A value of zero indicates that this rule targeted the main + * function. A positive value indicates which virtual function + * was targeted, so we'll subtract 1 in order to show the + * correct VF index + */ if (vf) fprintf(stdout, "\tAction: Direct to VF %llu queue %llu\n", - vf, queue); + vf - 1, queue); else fprintf(stdout, "\tAction: Direct to queue %llu\n", queue); @@ -1047,6 +1052,13 @@ static int rxclass_get_val(char *str, unsigned char *p, u32 *flags, err = rxclass_get_ulong(str, &val, 8); if (err) return -1; + + /* The ring_cookie uses 0 to indicate the rule targets the + * main function, so add 1 to the value in order to target the + * correct virtual function. + */ + val++; + *(u64 *)&p[opt->offset] &= ~ETHTOOL_RX_FLOW_SPEC_RING_VF; *(u64 *)&p[opt->offset] = (u64)val << ETHTOOL_RX_FLOW_SPEC_RING_VF_OFF; break;
Instead of relying on the input being offset by 1, instead adjust the "vf" input value by 1 so that it correctly targets the VF the user expects. Additionally, update the displayed output to show the correct VF number. Signed-off-by: Jacob Keller <jacob.e.keller@intel.com> --- This is the way the VF parameter should have worked from the beginning, but unfortunately I did not code it as such. It may be considered a functional change (since we already supported the vf parameter with the offset not handled properly). I would consider this a bug fix, but it might be considered a behavior change.... ethtool.8.in | 3 +-- rxclass.c | 14 +++++++++++++- 2 files changed, 14 insertions(+), 3 deletions(-)