[dpdk-dev,29/53] net/sfc/base: request info about outer frame in Rx events
Checks
Commit Message
From: Mark Spender <mspender@solarflare.com>
For encapsulated packets, the firmware gives info about the inner frame
fields by default. When not using encapsulation offload, ask for info
about the outer frame instead.
Unless bug69842 is fixed, a driver reload will be needed, at least
on some platforms, when switching from full-featured to low-latency
firmware (though users are unlikely to hit that).
Signed-off-by: Mark Spender <mspender@solarflare.com>
Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com>
---
drivers/net/sfc/base/ef10_rx.c | 26 ++++++++++++++++++++++++--
1 file changed, 24 insertions(+), 2 deletions(-)
Comments
On 11/16/2017 12:04 AM, Andrew Rybchenko wrote:
> From: Mark Spender <mspender@solarflare.com>
>
> For encapsulated packets, the firmware gives info about the inner frame
> fields by default. When not using encapsulation offload, ask for info
> about the outer frame instead.
>
> Unless bug69842 is fixed, a driver reload will be needed, at least
I guess this is your internal bug tracking id, this is not very useful in commit
log since we don't have access to it.
Leaving it as it is, but it would be more useful to describe defect here instead
of referring to its id.
> on some platforms, when switching from full-featured to low-latency
> firmware (though users are unlikely to hit that).
>
> Signed-off-by: Mark Spender <mspender@solarflare.com>
> Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com>
<...>
On 11/27/2017 10:57 PM, Ferruh Yigit wrote:
> On 11/16/2017 12:04 AM, Andrew Rybchenko wrote:
>> From: Mark Spender <mspender@solarflare.com>
>>
>> For encapsulated packets, the firmware gives info about the inner frame
>> fields by default. When not using encapsulation offload, ask for info
>> about the outer frame instead.
>>
>> Unless bug69842 is fixed, a driver reload will be needed, at least
> I guess this is your internal bug tracking id, this is not very useful in commit
> log since we don't have access to it.
>
> Leaving it as it is, but it would be more useful to describe defect here instead
> of referring to its id.
>
Let's rephrase the paragraph in the following way:
On SFN8xxx with firmware version before v6.4.2.1007 driver reload is needed
after switching from full-feature to low-latency firmware variant since
the driver
still thinks that firmware supports encapsulation, but firmware does not
tolerate request to provide info about outer frame in Rx events.
>> on some platforms, when switching from full-featured to low-latency
>> firmware (though users are unlikely to hit that).
>>
>> Signed-off-by: Mark Spender <mspender@solarflare.com>
>> Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com>
> <...>
>
@@ -46,6 +46,7 @@ efx_mcdi_init_rxq(
__in boolean_t disable_scatter,
__in uint32_t ps_bufsize)
{
+ efx_nic_cfg_t *encp = &(enp->en_nic_cfg);
efx_mcdi_req_t req;
uint8_t payload[MAX(MC_CMD_INIT_RXQ_EXT_IN_LEN,
MC_CMD_INIT_RXQ_EXT_OUT_LEN)];
@@ -55,6 +56,7 @@ efx_mcdi_init_rxq(
uint64_t addr;
efx_rc_t rc;
uint32_t dma_mode;
+ boolean_t want_outer_classes;
/* If this changes, then the payload size might need to change. */
EFSYS_ASSERT3U(MC_CMD_INIT_RXQ_OUT_LEN, ==, 0);
@@ -65,6 +67,25 @@ efx_mcdi_init_rxq(
else
dma_mode = MC_CMD_INIT_RXQ_EXT_IN_SINGLE_PACKET;
+ if (encp->enc_tunnel_encapsulations_supported != 0) {
+ /*
+ * WANT_OUTER_CLASSES can only be specified on hardware which
+ * supports tunnel encapsulation offloads, even though it is
+ * effectively the behaviour the hardware gives.
+ *
+ * Also, on hardware which does support such offloads, older
+ * firmware rejects the flag if the offloads are not supported
+ * by the current firmware variant, which means this may fail if
+ * the capabilities are not updated when the firmware variant
+ * changes. This is not an issue on newer firmware, as it was
+ * changed in bug 69842 (v6.4.2.1007) to permit this flag to be
+ * specified on all firmware variants.
+ */
+ want_outer_classes = B_TRUE;
+ } else {
+ want_outer_classes = B_FALSE;
+ }
+
(void) memset(payload, 0, sizeof (payload));
req.emr_cmd = MC_CMD_INIT_RXQ;
req.emr_in_buf = payload;
@@ -76,7 +97,7 @@ efx_mcdi_init_rxq(
MCDI_IN_SET_DWORD(req, INIT_RXQ_EXT_IN_TARGET_EVQ, target_evq);
MCDI_IN_SET_DWORD(req, INIT_RXQ_EXT_IN_LABEL, label);
MCDI_IN_SET_DWORD(req, INIT_RXQ_EXT_IN_INSTANCE, instance);
- MCDI_IN_POPULATE_DWORD_8(req, INIT_RXQ_EXT_IN_FLAGS,
+ MCDI_IN_POPULATE_DWORD_9(req, INIT_RXQ_EXT_IN_FLAGS,
INIT_RXQ_EXT_IN_FLAG_BUFF_MODE, 0,
INIT_RXQ_EXT_IN_FLAG_HDR_SPLIT, 0,
INIT_RXQ_EXT_IN_FLAG_TIMESTAMP, 0,
@@ -85,7 +106,8 @@ efx_mcdi_init_rxq(
INIT_RXQ_EXT_IN_FLAG_DISABLE_SCATTER, disable_scatter,
INIT_RXQ_EXT_IN_DMA_MODE,
dma_mode,
- INIT_RXQ_EXT_IN_PACKED_STREAM_BUFF_SIZE, ps_bufsize);
+ INIT_RXQ_EXT_IN_PACKED_STREAM_BUFF_SIZE, ps_bufsize,
+ INIT_RXQ_EXT_IN_FLAG_WANT_OUTER_CLASSES, want_outer_classes);
MCDI_IN_SET_DWORD(req, INIT_RXQ_EXT_IN_OWNER_ID, 0);
MCDI_IN_SET_DWORD(req, INIT_RXQ_EXT_IN_PORT_ID, EVB_PORT_ID_ASSIGNED);