fix i40e bug for flow filter
cover fix i40e bug for flow filter
fix i40e bug for flow filter and add comment in doc.
Wei Zhao (3):
net/i40e: remove ARP type check for FDIR filter
doc: input set requirement of each pctype for FDIR net/i40e: remove teardown when flush FDIR filter
doc/guides/nics/i40e.rst | 7 +++++++ drivers/net/i40e/i40e_flow.c | 5 ----- 2 files changed, 7 insertions(+), 5 deletions(-)
Signature · 1 line
2.19.1
PATCH 1/3 net/i40e: remove ARP type check for FDIR filter
I have try to enable ARP ethertype for FDIR filter, it work well for ARP for FDIR filter, so delete it.
Bugzilla ID: 402
--- drivers/net/i40e/i40e_flow.c | 2 -- 1 file changed, 2 deletions(-)
drivers/net/i40e/i40e_flow.c +0 -2
@@ -2666,7 +2666,6 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,if (next_type ==RTE_FLOW_ITEM_TYPE_VLAN || ether_type ==RTE_ETHER_TYPE_IPV4 || ether_type ==RTE_ETHER_TYPE_IPV6 || - ether_type ==RTE_ETHER_TYPE_ARP || ether_type == outer_tpid) { rte_flow_error_set(error, EINVAL, RTE_FLOW_ERROR_TYPE_ITEM,
@@ -2711,7 +2710,6 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,if (ether_type ==RTE_ETHER_TYPE_IPV4 || ether_type ==RTE_ETHER_TYPE_IPV6 || - ether_type ==RTE_ETHER_TYPE_ARP || ether_type == outer_tpid) { rte_flow_error_set(error, EINVAL, RTE_FLOW_ERROR_TYPE_ITEM,
Signature · 1 line
2.19.1
Re: [dpdk-dev] [PATCH 1/3] net/i40e: remove ARP type check for FDIR filter
Quoted 10 lines
-----Original Message-----
From: Zhao1, Wei <[email protected]>
Sent: Monday, April 27, 2020 3:15 PM
To: [email protected]
Cc: Xing, Beilei <[email protected]>; [email protected]; [email protected]; Zhao1, Wei <[email protected]>
Subject: [PATCH 1/3] net/i40e: remove ARP type check for FDIR filter
Quoted 2 lines
I have try to enable ARP ethertype for FDIR filter, it work well for ARP for FDIR filter, so delete it.
Could you rework the commit log by describing the problem and the root cause? And the title maybe 'fix FDIR issue for ARP packets' I guess the root cause is: if the ether_type is RTE_ETHER_TYPE_ARP, the PCTYPE parsed during pipeline is I40E_FILTER_PCTYPE_L2_PAYLOAD. Please double check.
Quoted 37 lines
Bugzilla ID: 402
Fixes 42044b69c67d ("net/i40e: support input set selection for FDIR")Signed-off-by Wei Zhao <[email protected]>--- drivers/net/i40e/i40e_flow.c | 2 -- 1 file changed, 2 deletions(-)drivers/net/i40e/i40e_flow.c index renamed from drivers/net/i40e/i40e_flow.c +0
-0
7e64ae53a..1533d5abb100644drivers/net/i40e/i40e_flow.c +0
-0@@ -2666,7 +2666,6 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev, if (next_type ==
RTE_FLOW_ITEM_TYPE_VLAN||drivers/net/i40e/i40e_flow.c +0
-1ether_type == RTE_ETHER_TYPE_IPV4 || ether_type == RTE_ETHER_TYPE_IPV6 || - ether_type == RTE_ETHER_TYPE_ARP || ether_type == outer_tpid) { rte_flow_error_set(error, EINVAL,
RTE_FLOW_ERROR_TYPE_ITEM,drivers/net/i40e/i40e_flow.c +0
-1@@ -2711,7 +2710,6 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev, if (ether_type == RTE_ETHER_TYPE_IPV4 || ether_type ==RTE_ETHER_TYPE_IPV6 || - ether_type ==RTE_ETHER_TYPE_ARP || ether_type == outer_tpid) { rte_flow_error_set(error, EINVAL,
RTE_FLOW_ERROR_TYPE_ITEM,
--
2.19.1
Re: [dpdk-dev] [PATCH 1/3] net/i40e: remove ARP type check for FDIR filter
Quoted 19 lines
-----Original Message-----
From: Xing, Beilei <[email protected]>
Sent: Tuesday, April 28, 2020 9:21 AM
To: Zhao1, Wei <[email protected]>; [email protected]
Cc: [email protected]; [email protected]
Subject: RE: [PATCH 1/3] net/i40e: remove ARP type check for FDIR filterQuoted 8 lines
-----Original Message-----
From: Zhao1, Wei <[email protected]>
Sent: Monday, April 27, 2020 3:15 PM
To: [email protected]
Cc: Xing, Beilei <[email protected]>; [email protected]; [email protected]; Zhao1, Wei <[email protected]>
Subject: [PATCH 1/3] net/i40e: remove ARP type check for FDIR filter
Quoted 8 lines
Quoted 2 lines
I have try to enable ARP ethertype for FDIR filter, it work well for ARP for FDIR filter, so delete it.
Could you rework the commit log by describing the problem and the root cause? And the title maybe 'fix FDIR issue for ARP packets'
I guess the root cause is: if theether_typeisRTE_ETHER_TYPE_ARP, the PCTYPE parsed during pipeline isI40E_FILTER_PCTYPE_L2_PAYLOAD. Please double check.
After check, this patch can enable FDIR with pctype I40E_FILTER_PCTYPE_L2_PAYLOAD for ARP packet. Comment will be update.
Quoted 38 lines
Quoted 38 lines
Bugzilla ID: 402
Fixes 42044b69c67d ("net/i40e: support input set selection for FDIR")Signed-off-by Wei Zhao <[email protected]>--- drivers/net/i40e/i40e_flow.c | 2 -- 1 file changed, 2 deletions(-)(quoted diff) +0
-0
b/drivers/net/i40e/i40e_flow.cindex7e64ae53a..1533d5abb100644drivers/net/i40e/i40e_flow.c +0
-0@@ -2666,7 +2666,6 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev, if (next_type ==
RTE_FLOW_ITEM_TYPE_VLAN||drivers/net/i40e/i40e_flow.c +0
-1ether_type == RTE_ETHER_TYPE_IPV4 || ether_type == RTE_ETHER_TYPE_IPV6 || - ether_type == RTE_ETHER_TYPE_ARP || ether_type == outer_tpid) { rte_flow_error_set(error, EINVAL,
RTE_FLOW_ERROR_TYPE_ITEM,drivers/net/i40e/i40e_flow.c +0
-1@@ -2711,7 +2710,6 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev, if (ether_type == RTE_ETHER_TYPE_IPV4 || ether_type ==RTE_ETHER_TYPE_IPV6 || - ether_type ==RTE_ETHER_TYPE_ARP || ether_type == outer_tpid) { rte_flow_error_set(error, EINVAL,
RTE_FLOW_ERROR_TYPE_ITEM,
--
2.19.1
PATCH 2/3 doc: input set requirement of each pctype for FDIR
In reply to Wei ZhaoAdd input set requirement info to i40e doc.
Bugzilla ID: 403
--- doc/guides/nics/i40e.rst | 7 +++++++ 1 file changed, 7 insertions(+)
doc/guides/nics/i40e.rst +7 -0
@@ -747,6 +747,13 @@ Use 16 Bytes RX Descriptor Size As i40e PMD supports both 16 and 32 bytes RX descriptor sizes, and 16 bytes size can provide helps to high performance of small packets. Configuration of ``CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC`` in config files can be changed to use 16 bytes size RX descriptors. +input set requirement of each pctype for FDIR +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Each pctype packet can only have one specific input set for FDIR, for example if you download 2 rte_flow rules of different input set for one specific pctype, +it is will fail and return the info "Conflict with the first rule's input set", that means the first type of input set conflict with the current one, +this limitaition is required by hardware. + Example of getting best performance with l3fwd example ------------------------------------------------------
Signature · 1 line
2.19.1
PATCH 3/3 net/i40e: remove teardown when flush FDIR filter
In reply to Wei ZhaoWhen we flush FDIR filter, we can not call i40e_fdir_teardown() function as it will free vsi used for FDIR, then the vsi->base_queue will be freed from pf->qp_pool, but vsi->base_queue can only get once when do dev init in i40e_pf_setup(). If we free it, it will never be alloc again.
Bugzilla ID: 404
--- drivers/net/i40e/i40e_flow.c | 3 --- 1 file changed, 3 deletions(-)
drivers/net/i40e/i40e_flow.c +0 -3
@@ -5145,7 +5145,6 @@ i40e_flow_destroy(struct rte_eth_dev *dev,/* If the last flow is destroyed, disable fdir. */ if (!ret &&TAILQ_EMPTY (&pf->fdir.fdir_list)) { -i40e_fdir_teardown (pf);i40e_fdir_rx_proc_enable (dev,0 ); }break ;
@@ -5343,8 +5342,6 @@ i40e_flow_flush_fdir_filter(struct i40e_pf *pf) pf->fdir.inset_flag[pctype] =0 ; } -i40e_fdir_teardown (pf); -return ret; }
Signature · 1 line
2.19.1
Re: [dpdk-dev] [PATCH 3/3] net/i40e: remove teardown when flush FDIR filter
Quoted 10 lines
-----Original Message-----
From: Zhao1, Wei <[email protected]>
Sent: Monday, April 27, 2020 3:15 PM
To: [email protected]
Cc: Xing, Beilei <[email protected]>; [email protected]; [email protected]; Zhao1, Wei <[email protected]>
Subject: [PATCH 3/3] net/i40e: remove teardown when flush FDIR filter
Quoted 4 lines
When we flush FDIR filter, we can not call
i40e_fdir_teardown()function as it will free vsi used for FDIR, then thevsi->base_queuewill be freed from pf-Quoted 1 line
qp_pool, butvsi->base_queuecan only get once when do dev init in
i40e_pf_setup(). If we free it, it will never be alloc again.
Then we should teardown fdir in dev_uninit, right?
Quoted 33 lines
Bugzilla ID: 404
Fixes 2e67a7fbf3ff ("net/i40e: config flow director automatically")Signed-off-by Wei Zhao <[email protected]>--- drivers/net/i40e/i40e_flow.c | 3 --- 1 file changed, 3 deletions(-)drivers/net/i40e/i40e_flow.c index renamed from drivers/net/i40e/i40e_flow.c +0
-0
1533d5abb..65f877866100644drivers/net/i40e/i40e_flow.c +0
-4@@ -5145,7 +5145,6 @@ i40e_flow_destroy(struct rte_eth_dev *dev,/* If the last flow is destroyed, disable fdir. */ if (!ret &&TAILQ_EMPTY (&pf->fdir.fdir_list)) { -i40e_fdir_teardown (pf);i40e_fdir_rx_proc_enable (dev,0 ); }break ;@@ -5343,8 +5342,6 @@ i40e_flow_flush_fdir_filter(struct i40e_pf *pf) pf->fdir.inset_flag[pctype] =0 ; } -i40e_fdir_teardown (pf); -return ret; } --2.19.1
Re: [dpdk-dev] [PATCH 3/3] net/i40e: remove teardown when flush FDIR filter
Quoted 19 lines
-----Original Message-----
From: Xing, Beilei <[email protected]>
Sent: Tuesday, April 28, 2020 9:13 AM
To: Zhao1, Wei <[email protected]>; [email protected]
Cc: [email protected]; [email protected]
Subject: RE: [PATCH 3/3] net/i40e: remove teardown when flush FDIR filterQuoted 8 lines
-----Original Message-----
From: Zhao1, Wei <[email protected]>
Sent: Monday, April 27, 2020 3:15 PM
To: [email protected]
Cc: Xing, Beilei <[email protected]>; [email protected]; [email protected]; Zhao1, Wei <[email protected]>
Subject: [PATCH 3/3] net/i40e: remove teardown when flush FDIR filter
Quoted 7 lines
Quoted 5 lines
When we flush FDIR filter, we can not call
i40e_fdir_teardown()function as it will free vsi used for FDIR, then thevsi->base_queuewill be freed from pf-Quoted 1 line
qp_pool, butvsi->base_queuecan only get once when do dev init in
i40e_pf_setup(). If we free it, it will never be alloc again.Then we should teardown fdir in
dev_uninit, right?
Yes, dev_close has call teardown function.
Quoted 35 lines
Quoted 35 lines
Bugzilla ID: 404
Fixes 2e67a7fbf3ff ("net/i40e: config flow director automatically")Signed-off-by Wei Zhao <[email protected]>--- drivers/net/i40e/i40e_flow.c | 3 --- 1 file changed, 3 deletions(-)(quoted diff) +0
-0
b/drivers/net/i40e/i40e_flow.cindex1533d5abb..65f877866100644drivers/net/i40e/i40e_flow.c +0
-4@@ -5145,7 +5145,6 @@ i40e_flow_destroy(struct rte_eth_dev *dev,/* If the last flow is destroyed, disable fdir. */ if (!ret &&TAILQ_EMPTY (&pf->fdir.fdir_list)) { -i40e_fdir_teardown (pf);i40e_fdir_rx_proc_enable (dev,0 ); }break ;@@ -5343,8 +5342,6 @@ i40e_flow_flush_fdir_filter(struct i40e_pf *pf) pf->fdir.inset_flag[pctype] =0 ; } -i40e_fdir_teardown (pf); -return ret; } --2.19.1
v2 cover fix i40e bug for flow filter
In reply to Wei Zhaofix i40e bug for flow filter and add comment in doc.
v2:
add more comment in git log.
Wei Zhao (3):
net/i40e: fix FDIR issue for ARP packets
doc: input set requirement of each pctype for FDIR net/i40e: remove teardown when flush FDIR filter
doc/guides/nics/i40e.rst | 7 +++++++ drivers/net/i40e/i40e_flow.c | 5 ----- 2 files changed, 7 insertions(+), 5 deletions(-)
Signature · 1 line
2.19.1
PATCH v2 1/3 net/i40e: fix FDIR issue for ARP packets
This patch can enable FDIR awith ethertype as input set for ARP packet, it will associate this rule, "flow create 0 ingress pattern eth type is 0x0806 / end actions mark id 0x86 / rss / end", with pctype I40E_FILTER_PCTYPE_L2_PAYLOAD for ARP packet rule. I have tried to enable ARP ethertype for FDIR filter,
it work well for ARP for FDIR filter, so delete the check.
Bugzilla ID: 402
--- drivers/net/i40e/i40e_flow.c | 2 -- 1 file changed, 2 deletions(-)
drivers/net/i40e/i40e_flow.c +0 -2
@@ -2666,7 +2666,6 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,if (next_type ==RTE_FLOW_ITEM_TYPE_VLAN || ether_type ==RTE_ETHER_TYPE_IPV4 || ether_type ==RTE_ETHER_TYPE_IPV6 || - ether_type ==RTE_ETHER_TYPE_ARP || ether_type == outer_tpid) { rte_flow_error_set(error, EINVAL, RTE_FLOW_ERROR_TYPE_ITEM,
@@ -2711,7 +2710,6 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,if (ether_type ==RTE_ETHER_TYPE_IPV4 || ether_type ==RTE_ETHER_TYPE_IPV6 || - ether_type ==RTE_ETHER_TYPE_ARP || ether_type == outer_tpid) { rte_flow_error_set(error, EINVAL, RTE_FLOW_ERROR_TYPE_ITEM,
Signature · 1 line
2.19.1
Re: [dpdk-dev] [PATCH v2 1/3] net/i40e: fix FDIR issue for ARP packets
Quoted 10 lines
-----Original Message-----
From: Zhao1, Wei <[email protected]>
Sent: Tuesday, April 28, 2020 1:48 PM
To: [email protected]
Cc: [email protected]; Xing, Beilei <[email protected]>; [email protected]; Zhao1, Wei <[email protected]>
Subject: [PATCH v2 1/3] net/i40e: fix FDIR issue for ARP packets
Quoted 6 lines
This patch can enable FDIR awith ethertype as input set for ARP packet, it will associate this rule, "flow create 0 ingress pattern eth type is
0x0806/ end actions mark id0x86/ rss / end", with pctypeI40E_FILTER_PCTYPE_L2_PAYLOADfor ARP packet rule.
I have tried to enable ARP ethertype for FDIR filter, it work well for ARP for FDIR filter, so delete the check.
How about
'Currently, flow "pattern eth type is 0x0806 / end actions mark id 0x86 / rss / end" can't be created successfully. FDIR parser shouldn't deny RTE_ETHER_TYPE_ARP since ARP packets will be parsed as PCTYPE_L2_PAYLOAD. This patch fixes the issue.' ?
Quoted 37 lines
Bugzilla ID: 402
Fixes 42044b69c67d ("net/i40e: support input set selection for FDIR")Signed-off-by Wei Zhao <[email protected]>--- drivers/net/i40e/i40e_flow.c | 2 -- 1 file changed, 2 deletions(-)drivers/net/i40e/i40e_flow.c index renamed from drivers/net/i40e/i40e_flow.c +0
-0
7e64ae53a..1533d5abb100644drivers/net/i40e/i40e_flow.c +0
-0@@ -2666,7 +2666,6 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev, if (next_type ==
RTE_FLOW_ITEM_TYPE_VLAN||drivers/net/i40e/i40e_flow.c +0
-1ether_type == RTE_ETHER_TYPE_IPV4 || ether_type == RTE_ETHER_TYPE_IPV6 || - ether_type == RTE_ETHER_TYPE_ARP || ether_type == outer_tpid) { rte_flow_error_set(error, EINVAL,
RTE_FLOW_ERROR_TYPE_ITEM,drivers/net/i40e/i40e_flow.c +0
-1@@ -2711,7 +2710,6 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev, if (ether_type == RTE_ETHER_TYPE_IPV4 || ether_type ==RTE_ETHER_TYPE_IPV6 || - ether_type ==RTE_ETHER_TYPE_ARP || ether_type == outer_tpid) { rte_flow_error_set(error, EINVAL,
RTE_FLOW_ERROR_TYPE_ITEM,
--
2.19.1
Re: [dpdk-dev] [PATCH v2 1/3] net/i40e: fix FDIR issue for ARP packets
Quoted 26 lines
-----Original Message-----
From: Xing, Beilei <[email protected]>
Sent: Tuesday, April 28, 2020 3:38 PM
To: Zhao1, Wei <[email protected]>; [email protected]
Cc: [email protected]; [email protected]
Subject: RE: [PATCH v2 1/3] net/i40e: fix FDIR issue for ARP packetsQuoted 8 lines
-----Original Message-----
From: Zhao1, Wei <[email protected]>
Sent: Tuesday, April 28, 2020 1:48 PM
To: [email protected]
Cc: [email protected]; Xing, Beilei <[email protected]>; [email protected]; Zhao1, Wei <[email protected]>
Subject: [PATCH v2 1/3] net/i40e: fix FDIR issue for ARP packetsQuoted 6 lines
This patch can enable FDIR awith ethertype as input set for ARP packet, it will associate this rule, "flow create 0 ingress pattern eth type is
0x0806/ end actions mark id0x86/ rss / end", with pctypeI40E_FILTER_PCTYPE_L2_PAYLOADfor ARP packet rule.
I have tried to enable ARP ethertype for FDIR filter, it work well for ARP for FDIR filter, so delete the check.
Quoted 6 lines
How about
'Currently, flow "pattern eth type is0x0806/ end actions mark id0x86/ rss / end" can't be created successfully. FDIR parser shouldn't denyRTE_ETHER_TYPE_ARPsince ARP packets will be parsed asPCTYPE_L2_PAYLOAD.
This patch fixes the issue.' ?
Ok
Quoted 38 lines
Quoted 38 lines
Bugzilla ID: 402
Fixes 42044b69c67d ("net/i40e: support input set selection for FDIR")Signed-off-by Wei Zhao <[email protected]>--- drivers/net/i40e/i40e_flow.c | 2 -- 1 file changed, 2 deletions(-)(quoted diff) +0
-0
b/drivers/net/i40e/i40e_flow.cindex7e64ae53a..1533d5abb100644drivers/net/i40e/i40e_flow.c +0
-0@@ -2666,7 +2666,6 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev, if (next_type ==
RTE_FLOW_ITEM_TYPE_VLAN||drivers/net/i40e/i40e_flow.c +0
-1ether_type == RTE_ETHER_TYPE_IPV4 || ether_type == RTE_ETHER_TYPE_IPV6 || - ether_type == RTE_ETHER_TYPE_ARP || ether_type == outer_tpid) { rte_flow_error_set(error, EINVAL,
RTE_FLOW_ERROR_TYPE_ITEM,drivers/net/i40e/i40e_flow.c +0
-1@@ -2711,7 +2710,6 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev, if (ether_type == RTE_ETHER_TYPE_IPV4 || ether_type ==RTE_ETHER_TYPE_IPV6 || - ether_type ==RTE_ETHER_TYPE_ARP || ether_type == outer_tpid) { rte_flow_error_set(error, EINVAL,
RTE_FLOW_ERROR_TYPE_ITEM,
--
2.19.1
PATCH v2 2/3 doc: input set requirement of each pctype for FDIR
In reply to Wei ZhaoAdd input set requirement info to i40e doc.
Bugzilla ID: 403
--- doc/guides/nics/i40e.rst | 7 +++++++ 1 file changed, 7 insertions(+)
doc/guides/nics/i40e.rst +7 -0
@@ -747,6 +747,13 @@ Use 16 Bytes RX Descriptor Size As i40e PMD supports both 16 and 32 bytes RX descriptor sizes, and 16 bytes size can provide helps to high performance of small packets. Configuration of ``CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC`` in config files can be changed to use 16 bytes size RX descriptors. +input set requirement of each pctype for FDIR +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Each pctype packet can only have one specific input set for FDIR, for example if you download 2 rte_flow rules of different input set for one specific pctype, +it is will fail and return the info "Conflict with the first rule's input set", that means the first type of input set conflict with the current one, +this limitaition is required by hardware. + Example of getting best performance with l3fwd example ------------------------------------------------------
Signature · 1 line
2.19.1
PATCH v2 3/3 net/i40e: remove teardown when flush FDIR filter
In reply to Wei ZhaoWhen we flush FDIR filter, we can not call i40e_fdir_teardown() function as it will free vsi used for FDIR, then the vsi->base_queue will be freed from pf->qp_pool, but vsi->base_queue can only get once when do dev init in i40e_pf_setup(). If we free it, it will never be alloc again.
Bugzilla ID: 404
--- drivers/net/i40e/i40e_flow.c | 3 --- 1 file changed, 3 deletions(-)
drivers/net/i40e/i40e_flow.c +0 -3
@@ -5145,7 +5145,6 @@ i40e_flow_destroy(struct rte_eth_dev *dev,/* If the last flow is destroyed, disable fdir. */ if (!ret &&TAILQ_EMPTY (&pf->fdir.fdir_list)) { -i40e_fdir_teardown (pf);i40e_fdir_rx_proc_enable (dev,0 ); }break ;
@@ -5343,8 +5342,6 @@ i40e_flow_flush_fdir_filter(struct i40e_pf *pf) pf->fdir.inset_flag[pctype] =0 ; } -i40e_fdir_teardown (pf); -return ret; }
Signature · 1 line
2.19.1
Re: [dpdk-dev] [PATCH v2 3/3] net/i40e: remove teardown when flush FDIR filter
Didn't address my comment in v1 patch.
Quoted 45 lines
-----Original Message-----
From: Zhao1, Wei <[email protected]>
Sent: Tuesday, April 28, 2020 1:48 PM
To: [email protected]
Cc: [email protected]; Xing, Beilei <[email protected]>; [email protected]; Zhao1, Wei <[email protected]>
Subject: [PATCH v2 3/3] net/i40e: remove teardown when flush FDIR filterWhen we flush FDIR filter, we can not call
i40e_fdir_teardown()function as it will free vsi used for FDIR, then thevsi->base_queuewill be freed from pf-Quoted 1 line
qp_pool, butvsi->base_queuecan only get once when do dev init in
i40e_pf_setup(). If we free it, it will never be alloc again.Bugzilla ID: 404
Fixes 2e67a7fbf3ff ("net/i40e: config flow director automatically")Signed-off-by Wei Zhao <[email protected]>--- drivers/net/i40e/i40e_flow.c | 3 --- 1 file changed, 3 deletions(-)drivers/net/i40e/i40e_flow.c index renamed from drivers/net/i40e/i40e_flow.c +0
-0
1533d5abb..65f877866100644drivers/net/i40e/i40e_flow.c +0
-4@@ -5145,7 +5145,6 @@ i40e_flow_destroy(struct rte_eth_dev *dev,/* If the last flow is destroyed, disable fdir. */ if (!ret &&TAILQ_EMPTY (&pf->fdir.fdir_list)) { -i40e_fdir_teardown (pf);i40e_fdir_rx_proc_enable (dev,0 ); }break ;@@ -5343,8 +5342,6 @@ i40e_flow_flush_fdir_filter(struct i40e_pf *pf) pf->fdir.inset_flag[pctype] =0 ; } -i40e_fdir_teardown (pf); -return ret; } --2.19.1
Re: [dpdk-dev] [PATCH v2 3/3] net/i40e: remove teardown when flush FDIR filter
Quoted 10 lines
-----Original Message-----
From: Xing, Beilei <[email protected]>
Sent: Tuesday, April 28, 2020 3:16 PM
To: Zhao1, Wei <[email protected]>; [email protected]
Cc: [email protected]; [email protected]
Subject: RE: [PATCH v2 3/3] net/i40e: remove teardown when flush FDIR filterDidn't address my comment in v1 patch.
Hi,
it has exist in unit.
Quoted 49 lines
Quoted 49 lines
-----Original Message-----
From: Zhao1, Wei <[email protected]>
Sent: Tuesday, April 28, 2020 1:48 PM
To: [email protected]
Cc: [email protected]; Xing, Beilei <[email protected]>; [email protected]; Zhao1, Wei <[email protected]>
Subject: [PATCH v2 3/3] net/i40e: remove teardown when flush FDIR filterWhen we flush FDIR filter, we can not call
i40e_fdir_teardown()function as it will free vsi used for FDIR, then thevsi->base_queuewill be freed from pf-Quoted 1 line
qp_pool, butvsi->base_queuecan only get once when do dev init in
i40e_pf_setup(). If we free it, it will never be alloc again.Bugzilla ID: 404
Fixes 2e67a7fbf3ff ("net/i40e: config flow director automatically")Signed-off-by Wei Zhao <[email protected]>--- drivers/net/i40e/i40e_flow.c | 3 --- 1 file changed, 3 deletions(-)(quoted diff) +0
-0
b/drivers/net/i40e/i40e_flow.cindex1533d5abb..65f877866100644drivers/net/i40e/i40e_flow.c +0
-4@@ -5145,7 +5145,6 @@ i40e_flow_destroy(struct rte_eth_dev *dev,/* If the last flow is destroyed, disable fdir. */ if (!ret &&TAILQ_EMPTY (&pf->fdir.fdir_list)) { -i40e_fdir_teardown (pf);i40e_fdir_rx_proc_enable (dev,0 ); }break ;@@ -5343,8 +5342,6 @@ i40e_flow_flush_fdir_filter(struct i40e_pf *pf) pf->fdir.inset_flag[pctype] =0 ; } -i40e_fdir_teardown (pf); -return ret; } --2.19.1
v3 cover fix i40e bug for flow filter
In reply to Wei Zhaofix i40e bug for flow filter and add comment in doc.
v2:
add more comment in git log.
v3:
update git log.
Wei Zhao (3):
net/i40e: fix FDIR issue for ARP packets
doc: input set requirement of each pctype for FDIR net/i40e: remove teardown when flush FDIR filter
doc/guides/nics/i40e.rst | 7 +++++++ drivers/net/i40e/i40e_flow.c | 5 ----- 2 files changed, 7 insertions(+), 5 deletions(-)
Signature · 1 line
2.19.1
PATCH v3 1/3 net/i40e: fix FDIR issue for ARP packets
Currently, flow "pattern eth type is 0x0806 / end actions mark id 0x86 / rss / end" can't be created successfully. FDIR parser shouldn't deny RTE_ETHER_TYPE_ARP since ARP packets will be parsed as PCTYPE_L2_PAYLOAD. This patch fixes the issue.
Bugzilla ID: 402
--- drivers/net/i40e/i40e_flow.c | 2 -- 1 file changed, 2 deletions(-)
drivers/net/i40e/i40e_flow.c +0 -2
@@ -2666,7 +2666,6 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,if (next_type ==RTE_FLOW_ITEM_TYPE_VLAN || ether_type ==RTE_ETHER_TYPE_IPV4 || ether_type ==RTE_ETHER_TYPE_IPV6 || - ether_type ==RTE_ETHER_TYPE_ARP || ether_type == outer_tpid) { rte_flow_error_set(error, EINVAL, RTE_FLOW_ERROR_TYPE_ITEM,
@@ -2711,7 +2710,6 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,if (ether_type ==RTE_ETHER_TYPE_IPV4 || ether_type ==RTE_ETHER_TYPE_IPV6 || - ether_type ==RTE_ETHER_TYPE_ARP || ether_type == outer_tpid) { rte_flow_error_set(error, EINVAL, RTE_FLOW_ERROR_TYPE_ITEM,
Signature · 1 line
2.19.1
PATCH v3 2/3 doc: input set requirement of each pctype for FDIR
In reply to Wei ZhaoAdd input set requirement info to i40e doc.
Bugzilla ID: 403
--- doc/guides/nics/i40e.rst | 7 +++++++ 1 file changed, 7 insertions(+)
doc/guides/nics/i40e.rst +7 -0
@@ -747,6 +747,13 @@ Use 16 Bytes RX Descriptor Size As i40e PMD supports both 16 and 32 bytes RX descriptor sizes, and 16 bytes size can provide helps to high performance of small packets. Configuration of ``CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC`` in config files can be changed to use 16 bytes size RX descriptors. +input set requirement of each pctype for FDIR +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Each pctype packet can only have one specific input set for FDIR, for example if you download 2 rte_flow rules of different input set for one specific pctype, +it is will fail and return the info "Conflict with the first rule's input set", that means the first type of input set conflict with the current one, +this limitaition is required by hardware. + Example of getting best performance with l3fwd example ------------------------------------------------------
Signature · 1 line
2.19.1
PATCH v3 3/3 net/i40e: remove teardown when flush FDIR filter
In reply to Wei ZhaoWhen we flush FDIR filter, we can not call i40e_fdir_teardown() function as it will free vsi used for FDIR, then the vsi->base_queue will be freed from pf->qp_pool, but vsi->base_queue can only get once when do dev init in i40e_pf_setup(). If we free it, it will never be alloc again.
Bugzilla ID: 404
--- drivers/net/i40e/i40e_flow.c | 3 --- 1 file changed, 3 deletions(-)
drivers/net/i40e/i40e_flow.c +0 -3
@@ -5145,7 +5145,6 @@ i40e_flow_destroy(struct rte_eth_dev *dev,/* If the last flow is destroyed, disable fdir. */ if (!ret &&TAILQ_EMPTY (&pf->fdir.fdir_list)) { -i40e_fdir_teardown (pf);i40e_fdir_rx_proc_enable (dev,0 ); }break ;
@@ -5343,8 +5342,6 @@ i40e_flow_flush_fdir_filter(struct i40e_pf *pf) pf->fdir.inset_flag[pctype] =0 ; } -i40e_fdir_teardown (pf); -return ret; }
Signature · 1 line
2.19.1
v4 cover fix i40e bug for flow filter
In reply to Wei Zhaofix i40e bug for flow filter and add comment in doc.
v2:
add more comment in git log.
v3:
update git log.
v4:
update doc info.
Wei Zhao (3):
net/i40e: fix FDIR issue for ARP packets
doc: input set requirement of each pctype for FDIR net/i40e: remove teardown when flush FDIR filter
doc/guides/nics/i40e.rst | 7 +++++++ drivers/net/i40e/i40e_flow.c | 5 ----- 2 files changed, 7 insertions(+), 5 deletions(-)
Signature · 1 line
2.19.1
PATCH v4 1/3 net/i40e: fix FDIR issue for ARP packets
Currently, flow "pattern eth type is 0x0806 / end actions mark id 0x86 / rss / end" can't be created successfully. FDIR parser shouldn't deny RTE_ETHER_TYPE_ARP since ARP packets will be parsed as PCTYPE_L2_PAYLOAD. This patch fixes the issue.
Bugzilla ID: 402
--- drivers/net/i40e/i40e_flow.c | 2 -- 1 file changed, 2 deletions(-)
drivers/net/i40e/i40e_flow.c +0 -2
@@ -2666,7 +2666,6 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,if (next_type ==RTE_FLOW_ITEM_TYPE_VLAN || ether_type ==RTE_ETHER_TYPE_IPV4 || ether_type ==RTE_ETHER_TYPE_IPV6 || - ether_type ==RTE_ETHER_TYPE_ARP || ether_type == outer_tpid) { rte_flow_error_set(error, EINVAL, RTE_FLOW_ERROR_TYPE_ITEM,
@@ -2711,7 +2710,6 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,if (ether_type ==RTE_ETHER_TYPE_IPV4 || ether_type ==RTE_ETHER_TYPE_IPV6 || - ether_type ==RTE_ETHER_TYPE_ARP || ether_type == outer_tpid) { rte_flow_error_set(error, EINVAL, RTE_FLOW_ERROR_TYPE_ITEM,
Signature · 1 line
2.19.1
PATCH v4 2/3 doc: input set requirement of each pctype for FDIR
In reply to Wei ZhaoAdd input set requirement info to i40e doc.
Bugzilla ID: 403
--- doc/guides/nics/i40e.rst | 7 +++++++ 1 file changed, 7 insertions(+)
doc/guides/nics/i40e.rst +7 -0
@@ -747,6 +747,13 @@ Use 16 Bytes RX Descriptor Size As i40e PMD supports both 16 and 32 bytes RX descriptor sizes, and 16 bytes size can provide helps to high performance of small packets. Configuration of ``CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC`` in config files can be changed to use 16 bytes size RX descriptors. +input set requirement of each pctype for FDIR +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Each PCTYPE can only have one specific FDIR input set at one time. For example, if creating 2 rte_flow rules with different input set +for one PCTYPE, it is will fail and return the info "Conflict with the first rule's input set", which means the current rule's input +set is conflict with the first rule's. Remove the first rule if want to change the input set of the PCTYPE. + Example of getting best performance with l3fwd example ------------------------------------------------------
Signature · 1 line
2.19.1
Re: [dpdk-dev] [dpdk-stable] [PATCH v4 2/3] doc: input set requirement of each pctype for FDIR
Quoted 3 lines · Wei Zhao
Add input set requirement info to i40e doc.
Bugzilla ID: 403
Should this document update backported to the LTS?
If so which LTS versions are in the scope, it is better if you can add a fixes tag to help us to figure out that.
Quoted 24 lines
Signed-off-by Wei Zhao <[email protected]>--- doc/guides/nics/i40e.rst | 7 +++++++ 1 file changed, 7 insertions(+)doc/guides/nics/i40e.rst +7
-0@@ -747,6 +747,13 @@ Use 16 Bytes RX Descriptor Size As i40e PMD supports both 16 and 32 bytes RX descriptor sizes, and 16 bytes size can provide helps to high performance of small packets. Configuration of ``CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC`` in config files can be changed to use 16 bytes size RX descriptors. +input set requirement of each pctype for FDIR +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Each PCTYPE can only have one specific FDIR input set at one time. For example, if creating 2 rte_flow rules with different input set +for one PCTYPE, it is will fail and return the info "Conflict with the first rule's input set", which means the current rule's input +set is conflict with the first rule's. Remove the first rule if want to change the input set of the PCTYPE. + Example of getting best performance with l3fwd example ------------------------------------------------------
Re: [dpdk-dev] [dpdk-stable] [PATCH v4 2/3] doc: input set requirement of each pctype for FDIR
Hi,
Quoted 9 lines
-----Original Message-----
From: Yigit, Ferruh <[email protected]>
Sent: Wednesday, April 29, 2020 6:35 PM
To: Zhao1, Wei <[email protected]>; [email protected]
Cc: [email protected]; Xing, Beilei <[email protected]>; [email protected]
Subject: Re: [dpdk-stable] [PATCH v4 2/3] doc: input set requirement of each pctype for FDIR
Quoted 8 lines
Quoted 3 lines · Wei Zhao
Add input set requirement info to i40e doc.
Bugzilla ID: 403
Should this document update backported to the LTS?
If so which LTS versions are in the scope, it is better if you can add a fixes tag to help us to figure out that.
Maybe we can use this, it is the latest update for i40 doc. Btw, this is not a bug although it is report in bug list.
Quoted 29 lines
Quoted 11 lines
Signed-off-by Wei Zhao <[email protected]>--- doc/guides/nics/i40e.rst | 7 +++++++ 1 file changed, 7 insertions(+)doc/guides/nics/i40e.rst index renamed from doc/guides/nics/i40e.rst +0
-0
416b3904e..f085a357c100644doc/guides/nics/i40e.rst +0
-0@@ -747,6 +747,13 @@ Use 16 Bytes RX Descriptor Size As i40e PMDQuoted 1 line
supports both 16 and 32 bytes RX descriptor sizes, and 16 bytes size can
provide helps to high performance of small packets.
Quoted 1 line
Configuration of ``CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC`` in config
files can be changed to use 16 bytes size RX descriptors.
doc/guides/nics/i40e.rst Use 16 Bytes RX Descriptor Size
Quoted 7 lines
doc/guides/nics/i40e.rst +6
-0+input set requirement of each pctype for FDIR +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Each PCTYPE can only have one specific FDIR input set at one time. +For example, if creating 2 rte_flow rules with different input set +for one PCTYPE, it is will fail and return the info "Conflict with the first rule'sinput set", which means the current rule's input set is conflict with the first rule's. Remove the first rule if want to change the input set of the PCTYPE.
Quoted 5 lines
+
Example of getting best performance with l3fwd example
------------------------------------------------------
Re: [dpdk-dev] [dpdk-stable] [PATCH v4 2/3] doc: input set requirement of each pctype for FDIR
Quoted 17 lines · Wei
Hi,
Quoted 14 lines
-----Original Message-----
From: Yigit, Ferruh <[email protected]>
Sent: Wednesday, April 29, 2020 6:35 PM
To: Zhao1, Wei <[email protected]>; [email protected]
Cc: [email protected]; Xing, Beilei <[email protected]>; [email protected]
Subject: Re: [dpdk-stable] [PATCH v4 2/3] doc: input set requirement of each pctype for FDIRQuoted 3 lines · Wei Zhao
Add input set requirement info to i40e doc.
Bugzilla ID: 403
Quoted 8 lines
Quoted 3 lines
Should this document update backported to the LTS?
If so which LTS versions are in the scope, it is better if you can add a fixes tag to help us to figure out that.Maybe we can use this, it is the latest update for i40 doc. Btw, this is not a bug although it is report in bug list.
Fixes 3d91e097a352 ("doc: add feature support matrix link in i40e guide ")
Hi Wei,
I agree this is not bug, but this documents a limitation for users. And if this limitation is valid for the stable trees, it make sense to update documentation for those releases too, so that they can have correct/complete documentation. I will add the "[email protected]" tag to request the backport.
But next question is for which stable trees is this limitation valid, it would be wrong to document a limitation that doesn't have the feature, how a stable tree maintainer can know if this issue is valid for a specific stable tree or not. For that we need the commit that introduces the mentioned feature (and the limitation), I don't think the commit above really is that commit, that is unrelated doc commit, but we need the commit for code.
If you can provide correct fixes line, I will update it in the next-net.
Thanks,
ferruh
Quoted 31 lines
Quoted 30 lines
Quoted 12 lines
Signed-off-by Wei Zhao <[email protected]>--- doc/guides/nics/i40e.rst | 7 +++++++ 1 file changed, 7 insertions(+)doc/guides/nics/i40e.rst index renamed from doc/guides/nics/i40e.rst +0
-0
416b3904e..f085a357c100644doc/guides/nics/i40e.rst +0
-0@@ -747,6 +747,13 @@ Use 16 Bytes RX Descriptor Size As i40e PMDQuoted 1 line
supports both 16 and 32 bytes RX descriptor sizes, and 16 bytes size can
provide helps to high performance of small packets.
Quoted 1 line
Configuration of ``CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC`` in config
files can be changed to use 16 bytes size RX descriptors.
doc/guides/nics/i40e.rst Use 16 Bytes RX Descriptor Size
Quoted 7 lines
doc/guides/nics/i40e.rst +6
-0+input set requirement of each pctype for FDIR +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Each PCTYPE can only have one specific FDIR input set at one time. +For example, if creating 2 rte_flow rules with different input set +for one PCTYPE, it is will fail and return the info "Conflict with the first rule'sinput set", which means the current rule's input set is conflict with the first rule's. Remove the first rule if want to change the input set of the PCTYPE.
Quoted 5 lines
+
Example of getting best performance with l3fwd example
------------------------------------------------------
Re: [dpdk-dev] [dpdk-stable] [PATCH v4 2/3] doc: input set requirement of each pctype for FDIR
HI, Ferruh
This is the patch enable FDIR for flow filter, so we can use this.
Quoted 92 lines
-----Original Message-----
From: Yigit, Ferruh <[email protected]>
Sent: Thursday, April 30, 2020 6:19 PM
To: Zhao1, Wei <[email protected]>; [email protected]
Cc: [email protected]; Xing, Beilei <[email protected]>; [email protected]; Kevin Traynor <[email protected]>
Subject: Re: [dpdk-stable] [PATCH v4 2/3] doc: input set requirement of each pctype for FDIRQuoted 21 lines · Wei
Hi,
Quoted 9 lines
-----Original Message-----
From: Yigit, Ferruh <[email protected]>
Sent: Wednesday, April 29, 2020 6:35 PM
To: Zhao1, Wei <[email protected]>; [email protected]
Cc: [email protected]; Xing, Beilei <[email protected]>; [email protected]
Subject: Re: [dpdk-stable] [PATCH v4 2/3] doc: input set requirement of each pctype for FDIRQuoted 8 lines
Quoted 3 lines · Wei Zhao
Add input set requirement info to i40e doc.
Bugzilla ID: 403
Should this document update backported to the LTS?
If so which LTS versions are in the scope, it is better if you can add a fixes tag to help us to figure out that.Quoted 5 lines
Maybe we can use this, it is the latest update for i40 doc. Btw, this is not a bug although it is report in bug list.
Fixes 3d91e097a352 ("doc: add feature support matrix link in i40eguide ")
Hi Wei,
I agree this is not bug, but this documents a limitation for users. And if this limitation is valid for the stable trees, it make sense to update documentation for those releases too, so that they can have correct/complete documentation. I will add the "[email protected]" tag to request the backport.
But next question is for which stable trees is this limitation valid, it would be wrong to document a limitation that doesn't have the feature, how a stable tree maintainer can know if this issue is valid for a specific stable tree or not. For that we need the commit that introduces the mentioned feature (and the limitation), I don't think the commit above really is that commit, that is unrelated doc commit, but we need the commit for code.
If you can provide correct fixes line, I will update it in the next-net.
Thanks,
ferruhQuoted 19 lines
Quoted 19 lines
Quoted 12 lines
Signed-off-by Wei Zhao <[email protected]>--- doc/guides/nics/i40e.rst | 7 +++++++ 1 file changed, 7 insertions(+)doc/guides/nics/i40e.rst +0
-0@@ -747,6 +747,13 @@ Use 16 Bytes RX Descriptor Size As i40e PMDQuoted 2 lines
supports both 16 and 32 bytes RX descriptor sizes, and 16 bytes size can
provide helps to high performance of small packets.
Quoted 2 lines
Configuration of ``CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC`` in
configfiles can be changed to use 16 bytes size RX descriptors.
Quoted 1 line
Quoted 9 lines
Quoted 9 lines
doc/guides/nics/i40e.rst
Quoted 7 lines
doc/guides/nics/i40e.rst +7
-0+input set requirement of each pctype for FDIR +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Each PCTYPE can only have one specific FDIR input set at one time. +For example, if creating 2 rte_flow rules with different input set +for one PCTYPE, it is will fail and return the info "Conflict with +the first rule'sinput set", which means the current rule's input set is conflict with the first rule's. Remove the first rule if want to change the input set of the
PCTYPE.
Quoted 6 lines
Quoted 5 lines
Quoted 5 lines
+
Example of getting best performance with l3fwd example
------------------------------------------------------
Re: [dpdk-dev] [dpdk-stable] [PATCH v4 2/3] doc: input set requirement of each pctype for FDIR
Quoted 6 lines · Wei
HI, Ferruh
This is the patch enable FDIR for flow filter, so we can use this.
Fixes 14c66a451ef ("net/i40e: parse flow director filter ")
Updated the commit log, thanks.
Quoted 95 lines
Quoted 94 lines
-----Original Message-----
From: Yigit, Ferruh <[email protected]>
Sent: Thursday, April 30, 2020 6:19 PM
To: Zhao1, Wei <[email protected]>; [email protected]
Cc: [email protected]; Xing, Beilei <[email protected]>; [email protected]; Kevin Traynor <[email protected]>
Subject: Re: [dpdk-stable] [PATCH v4 2/3] doc: input set requirement of each pctype for FDIRQuoted 21 lines · Wei
Hi,
Quoted 9 lines
-----Original Message-----
From: Yigit, Ferruh <[email protected]>
Sent: Wednesday, April 29, 2020 6:35 PM
To: Zhao1, Wei <[email protected]>; [email protected]
Cc: [email protected]; Xing, Beilei <[email protected]>; [email protected]
Subject: Re: [dpdk-stable] [PATCH v4 2/3] doc: input set requirement of each pctype for FDIRQuoted 8 lines
Quoted 3 lines · Wei Zhao
Add input set requirement info to i40e doc.
Bugzilla ID: 403
Should this document update backported to the LTS?
If so which LTS versions are in the scope, it is better if you can add a fixes tag to help us to figure out that.Quoted 5 lines
Maybe we can use this, it is the latest update for i40 doc. Btw, this is not a bug although it is report in bug list.
Fixes 3d91e097a352 ("doc: add feature support matrix link in i40eguide ")
Hi Wei,
I agree this is not bug, but this documents a limitation for users. And if this limitation is valid for the stable trees, it make sense to update documentation for those releases too, so that they can have correct/complete documentation. I will add the "[email protected]" tag to request the backport.
But next question is for which stable trees is this limitation valid, it would be wrong to document a limitation that doesn't have the feature, how a stable tree maintainer can know if this issue is valid for a specific stable tree or not. For that we need the commit that introduces the mentioned feature (and the limitation), I don't think the commit above really is that commit, that is unrelated doc commit, but we need the commit for code.
If you can provide correct fixes line, I will update it in the next-net.
Thanks,
ferruhQuoted 19 lines
Quoted 19 lines
Quoted 12 lines
Signed-off-by Wei Zhao <[email protected]>--- doc/guides/nics/i40e.rst | 7 +++++++ 1 file changed, 7 insertions(+)doc/guides/nics/i40e.rst +0
-0@@ -747,6 +747,13 @@ Use 16 Bytes RX Descriptor Size As i40e PMDQuoted 2 lines
supports both 16 and 32 bytes RX descriptor sizes, and 16 bytes size can
provide helps to high performance of small packets.
Quoted 2 lines
Configuration of ``CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC`` in
configfiles can be changed to use 16 bytes size RX descriptors.
Quoted 1 line
Quoted 9 lines
Quoted 9 lines
doc/guides/nics/i40e.rst
Quoted 7 lines
doc/guides/nics/i40e.rst +7
-0+input set requirement of each pctype for FDIR +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Each PCTYPE can only have one specific FDIR input set at one time. +For example, if creating 2 rte_flow rules with different input set +for one PCTYPE, it is will fail and return the info "Conflict with +the first rule'sinput set", which means the current rule's input set is conflict with the first rule's. Remove the first rule if want to change the input set of the
PCTYPE.
Quoted 6 lines
Quoted 5 lines
Quoted 5 lines
+
Example of getting best performance with l3fwd example
------------------------------------------------------
Re: [dpdk-dev] [dpdk-stable] [PATCH v4 2/3] doc: input set requirement of each pctype for FDIR
Thank for your explanation.
Quoted 118 lines
-----Original Message-----
From: Yigit, Ferruh <[email protected]>
Sent: Wednesday, May 6, 2020 5:29 PM
To: Zhao1, Wei <[email protected]>; [email protected]
Cc: [email protected]; Xing, Beilei <[email protected]>; [email protected]; Kevin Traynor <[email protected]>
Subject: Re: [dpdk-stable] [PATCH v4 2/3] doc: input set requirement of each pctype for FDIRQuoted 6 lines · Wei
HI, Ferruh
This is the patch enable FDIR for flow filter, so we can use this.
Fixes 14c66a451ef ("net/i40e: parse flow director filter ")Updated the commit log, thanks.
Quoted 36 lines
Quoted 36 lines
-----Original Message-----
From: Yigit, Ferruh <[email protected]>
Sent: Thursday, April 30, 2020 6:19 PM
To: Zhao1, Wei <[email protected]>; [email protected]
Cc: [email protected]; Xing, Beilei <[email protected]>; [email protected]; Kevin Traynor <[email protected]>
Subject: Re: [dpdk-stable] [PATCH v4 2/3] doc: input set requirement of each pctype for FDIRQuoted 24 lines · Wei
Hi,
Quoted 9 lines
-----Original Message-----
From: Yigit, Ferruh <[email protected]>
Sent: Wednesday, April 29, 2020 6:35 PM
To: Zhao1, Wei <[email protected]>; [email protected]
Cc: [email protected]; Xing, Beilei <[email protected]>; [email protected]
Subject: Re: [dpdk-stable] [PATCH v4 2/3] doc: input set requirement of each pctype for FDIRQuoted 8 lines
Quoted 3 lines · Wei Zhao
Add input set requirement info to i40e doc.
Bugzilla ID: 403
Should this document update backported to the LTS?
If so which LTS versions are in the scope, it is better if you can add a fixes tag to help us to figure out that.Maybe we can use this, it is the latest update for i40 doc. Btw, this is not a bug although it is report in bug list.
Quoted 8 lines
Quoted 8 lines
Quoted 2 lines
Fixes 3d91e097a352 ("doc: add feature support matrix link in i40eguide ")
Hi Wei,
I agree this is not bug, but this documents a limitation for users. And if this limitation is valid for the stable trees, it make sense to update documentation for those releases too, so that they can have
correct/complete documentation.
Quoted 5 lines
Quoted 5 lines
I will add the "[email protected]" tag to request the backport.
But next question is for which stable trees is this limitation valid, it would be wrong to document a limitation that doesn't have the feature, how a stable tree maintainer can know if this issue is valid for a
specific stable tree or not.
Quoted 17 lines
Quoted 17 lines
For that we need the commit that introduces the mentioned feature (and the limitation), I don't think the commit above really is that commit, that is unrelated doc commit, but we need the commit for code.
If you can provide correct fixes line, I will update it in the next-net.
Thanks,
ferruhQuoted 7 lines
Quoted 7 lines
Quoted 7 lines
Signed-off-by Wei Zhao <[email protected]>--- doc/guides/nics/i40e.rst | 7 +++++++ 1 file changed, 7 insertions(+)Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
doc/guides/nics/i40e.rst Use 16 Bytes RX Descriptor Size As i40e
Quoted 5 lines
doc/guides/nics/i40e.rst +0
-0@@ -747,6 +747,13 @@ Use 16 Bytes RX Descriptor Size As i40ePMD
Quoted 25 lines
Quoted 24 lines
Quoted 7 lines
Quoted 7 lines
Quoted 2 lines
supports both 16 and 32 bytes RX descriptor sizes, and 16 bytes size can
provide helps to high performance of small packets.
Quoted 2 lines
Configuration of ``CONFIG_RTE_LIBRTE_I40E_16BYTE_RX_DESC`` in
configfiles can be changed to use 16 bytes size RX descriptors.
Quoted 1 line
Quoted 10 lines
Quoted 10 lines
doc/guides/nics/i40e.rst
Quoted 7 lines
doc/guides/nics/i40e.rst +7
-0+input set requirement of each pctype for FDIR +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Each PCTYPE can only have one specific FDIR input set at one time. +For example, if creating 2 rte_flow rules with different input +set for one PCTYPE, it is will fail and return the info "Conflict +with the first rule'sinput set", which means the current rule's input set is conflict with the first rule's. Remove the first rule if want to change the input set of the
PCTYPE.
Quoted 6 lines
Quoted 5 lines
Quoted 5 lines
+
Example of getting best performance with l3fwd example
------------------------------------------------------
PATCH v4 3/3 net/i40e: remove teardown when flush FDIR filter
In reply to Wei ZhaoWhen we flush FDIR filter, we can not call i40e_fdir_teardown() function as it will free vsi used for FDIR, then the vsi->base_queue will be freed from pf->qp_pool, but vsi->base_queue can only get once when do dev init in i40e_pf_setup(). If we free it, it will never be alloc again.
Bugzilla ID: 404
--- drivers/net/i40e/i40e_flow.c | 3 --- 1 file changed, 3 deletions(-)
drivers/net/i40e/i40e_flow.c +0 -3
@@ -5145,7 +5145,6 @@ i40e_flow_destroy(struct rte_eth_dev *dev,/* If the last flow is destroyed, disable fdir. */ if (!ret &&TAILQ_EMPTY (&pf->fdir.fdir_list)) { -i40e_fdir_teardown (pf);i40e_fdir_rx_proc_enable (dev,0 ); }break ;
@@ -5343,8 +5342,6 @@ i40e_flow_flush_fdir_filter(struct i40e_pf *pf) pf->fdir.inset_flag[pctype] =0 ; } -i40e_fdir_teardown (pf); -return ret; }
Signature · 1 line
2.19.1
Quoted 26 lines
-----Original Message-----
From: Zhao1, Wei <[email protected]>
Sent: Wednesday, April 29, 2020 10:04 AM
To: [email protected]
Cc: [email protected]; Xing, Beilei <[email protected]>; [email protected]
Subject: [PATCH v4 0/3] fix i40e bug for flow filterfix i40e bug for flow filter and add comment in doc.
v2:
add more comment in git log.v3:
update git log.v4:
update doc info.Wei Zhao (3):
net/i40e: fix FDIR issue for ARP packets
doc: input set requirement of each pctype for FDIR net/i40e: remove teardown when flush FDIR filter
Quoted 6 lines
doc/guides/nics/i40e.rst | 7 +++++++ drivers/net/i40e/i40e_flow.c | 5 ----- 2 files changed, 7 insertions(+), 5 deletions(-)--
2.19.1
Ye Xiaolong
AppliedQuoted 26 lines · Beilei
Quoted 26 lines
-----Original Message-----
From: Zhao1, Wei <[email protected]>
Sent: Wednesday, April 29, 2020 10:04 AM
To: [email protected]
Cc: [email protected]; Xing, Beilei <[email protected]>; [email protected]
Subject: [PATCH v4 0/3] fix i40e bug for flow filterfix i40e bug for flow filter and add comment in doc.
v2:
add more comment in git log.v3:
update git log.v4:
update doc info.Wei Zhao (3):
net/i40e: fix FDIR issue for ARP packets
doc: input set requirement of each pctype for FDIR net/i40e: remove teardown when flush FDIR filter
Quoted 8 lines
Quoted 6 lines
doc/guides/nics/i40e.rst | 7 +++++++ drivers/net/i40e/i40e_flow.c | 5 ----- 2 files changed, 7 insertions(+), 5 deletions(-)--
2.19.1Acked-by Beilei Xing <[email protected]>
Applied to dpdk-next-net-intel, Thanks.