* [PATCH v6 1/8] net/rnp: add skeleton
2023-09-01 2:30 [PATCH v6 0/8] [v6]drivers/net Add Support mucse N10 Pmd Driver Wenbo Cao
@ 2023-09-01 2:30 ` Wenbo Cao
2023-09-05 15:35 ` Ferruh Yigit
2024-03-29 11:28 ` Ferruh Yigit
2023-09-01 2:30 ` [PATCH v6 2/8] net/rnp: add ethdev probe and remove Wenbo Cao
` (7 subsequent siblings)
8 siblings, 2 replies; 27+ messages in thread
From: Wenbo Cao @ 2023-09-01 2:30 UTC (permalink / raw)
To: Thomas Monjalon, Wenbo Cao; +Cc: dev, ferruh.yigit, andrew.rybchenko, yaojun
Add Basic PMD library and doc build infrastructure
Update maintainers file to claim responsibility.
Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
---
MAINTAINERS | 6 +++++
doc/guides/nics/features/rnp.ini | 8 +++++++
doc/guides/nics/index.rst | 1 +
doc/guides/nics/rnp.rst | 38 ++++++++++++++++++++++++++++++++
drivers/net/meson.build | 1 +
drivers/net/rnp/meson.build | 11 +++++++++
drivers/net/rnp/rnp_ethdev.c | 3 +++
7 files changed, 68 insertions(+)
create mode 100644 doc/guides/nics/features/rnp.ini
create mode 100644 doc/guides/nics/rnp.rst
create mode 100644 drivers/net/rnp/meson.build
create mode 100644 drivers/net/rnp/rnp_ethdev.c
diff --git a/MAINTAINERS b/MAINTAINERS
index a5219926ab..29c130b280 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -955,6 +955,12 @@ F: drivers/net/qede/
F: doc/guides/nics/qede.rst
F: doc/guides/nics/features/qede*.ini
+Mucse rnp
+M: Wenbo Cao <caowenbo@mucse.com>
+F: drivers/net/rnp
+F: doc/guides/nics/rnp.rst
+F: doc/guides/nics/features/rnp.ini
+
Solarflare sfc_efx
M: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
F: drivers/common/sfc_efx/
diff --git a/doc/guides/nics/features/rnp.ini b/doc/guides/nics/features/rnp.ini
new file mode 100644
index 0000000000..2ad04ee330
--- /dev/null
+++ b/doc/guides/nics/features/rnp.ini
@@ -0,0 +1,8 @@
+;
+; Supported features of the 'rnp' network poll mode driver.
+;
+; Refer to default.ini for the full list of available PMD features.
+;
+[Features]
+Linux = Y
+x86-64 = Y
diff --git a/doc/guides/nics/index.rst b/doc/guides/nics/index.rst
index 5c9d1edf5e..cc89d3154a 100644
--- a/doc/guides/nics/index.rst
+++ b/doc/guides/nics/index.rst
@@ -61,6 +61,7 @@ Network Interface Controller Drivers
pcap_ring
pfe
qede
+ rnp
sfc_efx
softnic
tap
diff --git a/doc/guides/nics/rnp.rst b/doc/guides/nics/rnp.rst
new file mode 100644
index 0000000000..0eb8f2d415
--- /dev/null
+++ b/doc/guides/nics/rnp.rst
@@ -0,0 +1,38 @@
+.. SPADIX-License-Identifier: BSD-3-Clause
+ Copyright(c) 2023 Mucse IC Design Ltd.
+
+RNP Poll Mode driver
+====================
+
+The RNP ETHDEV PMD (**librte_net_rnp**) provides poll mode ethdev
+driver support for the inbuilt network device found in the **Mucse RNP**
+
+Prerequisites
+-------------
+More information can be found at `Mucse, Official Website
+<https://mucse.com/productDetail>`_.
+
+Supported Chipsets and NICs
+---------------------------
+
+- MUCSE Ethernet Controller N10 Series for 10GbE or 40GbE (Dual-port)
+
+Limitations or Known issues
+---------------------------
+
+Build with ICC is not supported yet.
+BSD are not supported yet.
+
+CRC stripping
+~~~~~~~~~~~~~
+
+The RNP Soc family Nic strip the CRC for every packets coming into the
+host interface irrespective of the offload configuration.
+When you want to disable CRC_OFFLOAD the operate will influence the rxCksum offload.
+
+VLAN Strip/Filter
+~~~~~~~~~~~~~~~~~
+
+For VLAN strip/filter, RNP just support vlan is CVLAN(0x8100).If the outvlan type is SVLAN(0X88a8)
+VLAN filter or strip will not effort for this packet.It will bypass filter to the host default queue,
+whatever the other filter rule is.
diff --git a/drivers/net/meson.build b/drivers/net/meson.build
index b1df17ce8c..f9e013d38e 100644
--- a/drivers/net/meson.build
+++ b/drivers/net/meson.build
@@ -54,6 +54,7 @@ drivers = [
'pfe',
'qede',
'ring',
+ 'rnp',
'sfc',
'softnic',
'tap',
diff --git a/drivers/net/rnp/meson.build b/drivers/net/rnp/meson.build
new file mode 100644
index 0000000000..4f37c6b456
--- /dev/null
+++ b/drivers/net/rnp/meson.build
@@ -0,0 +1,11 @@
+# SPDX-License-Identifier: BSD-3-Clause
+# Copyright(C) 2023 Mucse IC Design Ltd.
+#
+if not is_linux
+ build = false
+ reason = 'only supported on Linux'
+endif
+
+sources = files(
+ 'rnp_ethdev.c',
+)
diff --git a/drivers/net/rnp/rnp_ethdev.c b/drivers/net/rnp/rnp_ethdev.c
new file mode 100644
index 0000000000..9ce3c0b497
--- /dev/null
+++ b/drivers/net/rnp/rnp_ethdev.c
@@ -0,0 +1,3 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(C) 2023 Mucse IC Design Ltd.
+ */
--
2.27.0
^ permalink raw reply related [flat|nested] 27+ messages in thread* Re: [PATCH v6 1/8] net/rnp: add skeleton
2023-09-01 2:30 ` [PATCH v6 1/8] net/rnp: add skeleton Wenbo Cao
@ 2023-09-05 15:35 ` Ferruh Yigit
2023-09-06 8:15 ` 11
2024-03-29 11:28 ` Ferruh Yigit
1 sibling, 1 reply; 27+ messages in thread
From: Ferruh Yigit @ 2023-09-05 15:35 UTC (permalink / raw)
To: Wenbo Cao, Thomas Monjalon; +Cc: dev, andrew.rybchenko, yaojun
On 9/1/2023 3:30 AM, Wenbo Cao wrote:
> Add Basic PMD library and doc build infrastructure
> Update maintainers file to claim responsibility.
>
> Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
<...>
> diff --git a/doc/guides/nics/rnp.rst b/doc/guides/nics/rnp.rst
> new file mode 100644
> index 0000000000..0eb8f2d415
> --- /dev/null
> +++ b/doc/guides/nics/rnp.rst
> @@ -0,0 +1,38 @@
> +.. SPADIX-License-Identifier: BSD-3-Clause
s/SPADIX/SPDX/
> + Copyright(c) 2023 Mucse IC Design Ltd.
> +
> +RNP Poll Mode driver
> +====================
> +
> +The RNP ETHDEV PMD (**librte_net_rnp**) provides poll mode ethdev
> +driver support for the inbuilt network device found in the **Mucse RNP**
> +
> +Prerequisites
> +-------------
> +More information can be found at `Mucse, Official Website
> +<https://mucse.com/productDetail>`_.
> +
> +Supported Chipsets and NICs
> +---------------------------
> +
> +- MUCSE Ethernet Controller N10 Series for 10GbE or 40GbE (Dual-port)
> +
Can you provide a link for the product, and if there is an English
version it can reach to more people?
> +Limitations or Known issues
> +---------------------------
> +
> +Build with ICC is not supported yet.
>
I know this is documented in other PMDs too, but at this stage ICC is
supported with best effort, and I am not sure if anybody testing with
it, so I think it is OK to drop this as limitation.
> +BSD are not supported yet.
>
FreeBSD is not supported.
Also Windows seems not supported, you may document that too.
> +
> +CRC stripping
> +~~~~~~~~~~~~~
> +
> +The RNP Soc family Nic strip the CRC for every packets coming into the
s/Soc/SoC/
s/Nic/NIC/
> +host interface irrespective of the offload configuration.
> +When you want to disable CRC_OFFLOAD the operate will influence the rxCksum offload.
>
I didn't understand the second sentences, from first one I understand
'RTE_ETH_RX_OFFLOAD_KEEP_CRC' is not supported by the PMD, but can you
please clarify more?
> +
> +VLAN Strip/Filter
> +~~~~~~~~~~~~~~~~~
> +
> +For VLAN strip/filter, RNP just support vlan is CVLAN(0x8100).If the outvlan type is SVLAN(0X88a8)
s/vlan/VLAN/
Please put space, " ", after '.' in documentation. Same comment for all
document.
> +VLAN filter or strip will not effort for this packet.It will bypass filter to the host default queue,
> +whatever the other filter rule is.
>
VLAN strip/filter doesn't work for double-tag (QinQ), it only works for
single VLAN tag, is this correct? If so can you please update above
sentences to clarify this?
^ permalink raw reply [flat|nested] 27+ messages in thread
* RE: [PATCH v6 1/8] net/rnp: add skeleton
2023-09-05 15:35 ` Ferruh Yigit
@ 2023-09-06 8:15 ` 11
0 siblings, 0 replies; 27+ messages in thread
From: 11 @ 2023-09-06 8:15 UTC (permalink / raw)
To: 'Ferruh Yigit', 'Thomas Monjalon'
Cc: dev, andrew.rybchenko, yaojun
Hi Ferruh,
Thanks for your kindly review, please see the below comment.
Regards Wenbo
> -----Original Message-----
> From: Ferruh Yigit <ferruh.yigit@amd.com>
> Sent: 2023年9月5日 23:35
> To: Wenbo Cao <caowenbo@mucse.com>; Thomas Monjalon
> <thomas@monjalon.net>
> Cc: dev@dpdk.org; andrew.rybchenko@oktetlabs.ru; yaojun@mucse.com
> Subject: Re: [PATCH v6 1/8] net/rnp: add skeleton
>
> On 9/1/2023 3:30 AM, Wenbo Cao wrote:
> > Add Basic PMD library and doc build infrastructure Update maintainers
> > file to claim responsibility.
> >
> > Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
>
> <...>
>
> > diff --git a/doc/guides/nics/rnp.rst b/doc/guides/nics/rnp.rst new
> > file mode 100644 index 0000000000..0eb8f2d415
> > --- /dev/null
> > +++ b/doc/guides/nics/rnp.rst
> > @@ -0,0 +1,38 @@
> > +.. SPADIX-License-Identifier: BSD-3-Clause
>
> s/SPADIX/SPDX/
I get it.
>
> > + Copyright(c) 2023 Mucse IC Design Ltd.
> > +
> > +RNP Poll Mode driver
> > +====================
> > +
> > +The RNP ETHDEV PMD (**librte_net_rnp**) provides poll mode ethdev
> > +driver support for the inbuilt network device found in the **Mucse
> > +RNP**
> > +
> > +Prerequisites
> > +-------------
> > +More information can be found at `Mucse, Official Website
> > +<https://mucse.com/productDetail>`_.
> > +
> > +Supported Chipsets and NICs
> > +---------------------------
> > +
> > +- MUCSE Ethernet Controller N10 Series for 10GbE or 40GbE (Dual-port)
> > +
>
> Can you provide a link for the product, and if there is an English version it can
> reach to more people?
>
For English version, it will be coming soon in a month about.
we will provide a html link of pdf for the time.
> > +Limitations or Known issues
> > +---------------------------
> > +
> > +Build with ICC is not supported yet.
> >
>
> I know this is documented in other PMDs too, but at this stage ICC is supported
> with best effort, and I am not sure if anybody testing with it, so I think it is OK to
> drop this as limitation.
>
>
We have never test the code use icc-compile-tool, just use gcc-compile-tool,
I'm going to try to test it with icc :)
> > +BSD are not supported yet.
> >
>
> FreeBSD is not supported.
> Also Windows seems not supported, you may document that too.
>
> > +
> > +CRC stripping
> > +~~~~~~~~~~~~~
> > +
> > +The RNP Soc family Nic strip the CRC for every packets coming into
> > +the
>
> s/Soc/SoC/
> s/Nic/NIC/
>
For this spell issue, is there exist tool I can use to check it ?
Thanks a lot.
> > +host interface irrespective of the offload configuration.
> > +When you want to disable CRC_OFFLOAD the operate will influence the
> rxCksum offload.
> >
>
> I didn't understand the second sentences, from first one I understand
> 'RTE_ETH_RX_OFFLOAD_KEEP_CRC' is not supported by the PMD, but can you
> please clarify more?
>
>
disable CRC_OFFLOAD will trigger SOC bug,
which causing bad-cksum bit in rx description always be 1 ,even package has no cksum error.
> > +
> > +VLAN Strip/Filter
> > +~~~~~~~~~~~~~~~~~
> > +
> > +For VLAN strip/filter, RNP just support vlan is CVLAN(0x8100).If the
> > +outvlan type is SVLAN(0X88a8)
>
> s/vlan/VLAN/
> Please put space, " ", after '.' in documentation. Same comment for all document.
>
I forgot the format before.
> > +VLAN filter or strip will not effort for this packet.It will bypass
> > +filter to the host default queue, whatever the other filter rule is.
> >
>
> VLAN strip/filter doesn't work for double-tag (QinQ), it only works for single VLAN
> tag, is this correct? If so can you please update above sentences to clarify this?
>
For VLAN strip/filter, if the VLAN type have (0x88a8), the feature of hardware isn't effect.
^ permalink raw reply [flat|nested] 27+ messages in thread
* Re: [PATCH v6 1/8] net/rnp: add skeleton
2023-09-01 2:30 ` [PATCH v6 1/8] net/rnp: add skeleton Wenbo Cao
2023-09-05 15:35 ` Ferruh Yigit
@ 2024-03-29 11:28 ` Ferruh Yigit
2024-03-29 14:45 ` 11
1 sibling, 1 reply; 27+ messages in thread
From: Ferruh Yigit @ 2024-03-29 11:28 UTC (permalink / raw)
To: Wenbo Cao, Thomas Monjalon; +Cc: dev, andrew.rybchenko, yaojun
On 9/1/2023 3:30 AM, Wenbo Cao wrote:
> Add Basic PMD library and doc build infrastructure
> Update maintainers file to claim responsibility.
>
> Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
>
Hi Wenbo,
What is the status of the 'rnp' driver, v7 was expected but not
received, will upstreaming continue for v24.03?
^ permalink raw reply [flat|nested] 27+ messages in thread
* RE: [PATCH v6 1/8] net/rnp: add skeleton
2024-03-29 11:28 ` Ferruh Yigit
@ 2024-03-29 14:45 ` 11
2024-04-02 10:15 ` Ferruh Yigit
0 siblings, 1 reply; 27+ messages in thread
From: 11 @ 2024-03-29 14:45 UTC (permalink / raw)
To: Ferruh.Yigit, 'Thomas Monjalon'; +Cc: dev, andrew.rybchenko, yaojun
Hi Ferruh,
Thanks for your reminder, I'm sorry for that I had been work on anothing before.
Recendly, I have been reworked on this work. It will miss on release of v24.03.
For another thing, I'm always confused for the secondary process call like hw->mac_ops this function pointer.
is this method can work normally on secondary process ?
For example,ixgbe on secondary process call ixgbe_get_module_eeprom I fond that it will cause process core-dump.
because of secondary process use primary process function-point address.
Is dpdk plaform allowed user call the eth_dev_ops on secondary process?
I don't find any limit and don't know wether or not to protect this function call work normally.
My driver eth_dev_ops achieve will used some function-pointer, so I'm confued about this.
Wish your kind guidance.
Regadrs Wenbo
> -----Original Message-----
> From: Ferruh.Yigit@amd.com <Ferruh.Yigit@amd.com>
> Sent: 2024年3月29日 19:28
> To: Wenbo Cao <caowenbo@mucse.com>; Thomas Monjalon
> <thomas@monjalon.net>
> Cc: dev@dpdk.org; andrew.rybchenko@oktetlabs.ru; yaojun@mucse.com
> Subject: Re: [PATCH v6 1/8] net/rnp: add skeleton
>
> On 9/1/2023 3:30 AM, Wenbo Cao wrote:
> > Add Basic PMD library and doc build infrastructure Update maintainers
> > file to claim responsibility.
> >
> > Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
> >
>
> Hi Wenbo,
>
> What is the status of the 'rnp' driver, v7 was expected but not received, will
> upstreaming continue for v24.03?
^ permalink raw reply [flat|nested] 27+ messages in thread
* Re: [PATCH v6 1/8] net/rnp: add skeleton
2024-03-29 14:45 ` 11
@ 2024-04-02 10:15 ` Ferruh Yigit
0 siblings, 0 replies; 27+ messages in thread
From: Ferruh Yigit @ 2024-04-02 10:15 UTC (permalink / raw)
To: 11, 'Thomas Monjalon'; +Cc: dev, andrew.rybchenko, yaojun
On 3/29/2024 2:45 PM, 11 wrote:
> Hi Ferruh,
>
> Thanks for your reminder, I'm sorry for that I had been work on anothing before.
> Recendly, I have been reworked on this work. It will miss on release of v24.03.
>
> For another thing, I'm always confused for the secondary process call like hw->mac_ops this function pointer.
> is this method can work normally on secondary process ?
> For example,ixgbe on secondary process call ixgbe_get_module_eeprom I fond that it will cause process core-dump.
> because of secondary process use primary process function-point address.
> Is dpdk plaform allowed user call the eth_dev_ops on secondary process?
> I don't find any limit and don't know wether or not to protect this function call work normally.
> My driver eth_dev_ops achieve will used some function-pointer, so I'm confued about this.
> Wish your kind guidance.
>
Hi Wenbo,
"struct rte_eth_dev" is per process, not shared, so both primary and and
secondary has a copy of it.
And during driver initialization, 'eth_dev->dev_ops' is set, so both
primary and secondary processes has correct function pointers.
("struct rte_eth_dev_data" is shared, so any pointer withing that struct
should be in a shared memory between primary and secondaries.)
And there is no restriction for secondary process call 'eth_dev_ops' but
this call needs to be synchronized by application, since there is no
locking around those functions.
Common usecase is primary process runs the control path, uses
'eth_dev_ops' calls, and secondaries only does the datapath on specific
set of queues.
If you are referring to 'mac_ops', that is another layer of abstraction
some drivers use, but those should be tied to ethdev control path
functions eventually, so same rule for using ethdev control path applies
to them.
>
>
>> -----Original Message-----
>> From: Ferruh.Yigit@amd.com <Ferruh.Yigit@amd.com>
>> Sent: 2024年3月29日 19:28
>> To: Wenbo Cao <caowenbo@mucse.com>; Thomas Monjalon
>> <thomas@monjalon.net>
>> Cc: dev@dpdk.org; andrew.rybchenko@oktetlabs.ru; yaojun@mucse.com
>> Subject: Re: [PATCH v6 1/8] net/rnp: add skeleton
>>
>> On 9/1/2023 3:30 AM, Wenbo Cao wrote:
>>> Add Basic PMD library and doc build infrastructure Update maintainers
>>> file to claim responsibility.
>>>
>>> Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
>>>
>>
>> Hi Wenbo,
>>
>> What is the status of the 'rnp' driver, v7 was expected but not received, will
>> upstreaming continue for v24.03?
>
>
^ permalink raw reply [flat|nested] 27+ messages in thread
* [PATCH v6 2/8] net/rnp: add ethdev probe and remove
2023-09-01 2:30 [PATCH v6 0/8] [v6]drivers/net Add Support mucse N10 Pmd Driver Wenbo Cao
2023-09-01 2:30 ` [PATCH v6 1/8] net/rnp: add skeleton Wenbo Cao
@ 2023-09-01 2:30 ` Wenbo Cao
2023-09-05 15:36 ` Ferruh Yigit
2023-09-01 2:30 ` [PATCH v6 3/8] net/rnp: add device init and uninit Wenbo Cao
` (6 subsequent siblings)
8 siblings, 1 reply; 27+ messages in thread
From: Wenbo Cao @ 2023-09-01 2:30 UTC (permalink / raw)
To: Wenbo Cao, Anatoly Burakov
Cc: dev, ferruh.yigit, thomas, andrew.rybchenko, yaojun
Add basic PCIe ethdev probe and remove.
Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
---
drivers/net/rnp/rnp.h | 13 ++++++
drivers/net/rnp/rnp_ethdev.c | 83 ++++++++++++++++++++++++++++++++++++
2 files changed, 96 insertions(+)
create mode 100644 drivers/net/rnp/rnp.h
diff --git a/drivers/net/rnp/rnp.h b/drivers/net/rnp/rnp.h
new file mode 100644
index 0000000000..76d281cc0a
--- /dev/null
+++ b/drivers/net/rnp/rnp.h
@@ -0,0 +1,13 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(C) 2023 Mucse IC Design Ltd.
+ */
+#ifndef __RNP_H__
+#define __RNP_H__
+
+#define PCI_VENDOR_ID_MUCSE (0x8848)
+#define RNP_DEV_ID_N10G (0x1000)
+
+struct rnp_eth_port {
+} __rte_cache_aligned;
+
+#endif /* __RNP_H__ */
diff --git a/drivers/net/rnp/rnp_ethdev.c b/drivers/net/rnp/rnp_ethdev.c
index 9ce3c0b497..390f2e7743 100644
--- a/drivers/net/rnp/rnp_ethdev.c
+++ b/drivers/net/rnp/rnp_ethdev.c
@@ -1,3 +1,86 @@
/* SPDX-License-Identifier: BSD-3-Clause
* Copyright(C) 2023 Mucse IC Design Ltd.
*/
+
+#include <ethdev_pci.h>
+#include <rte_io.h>
+#include <rte_malloc.h>
+
+#include "rnp.h"
+
+static int
+rnp_eth_dev_init(struct rte_eth_dev *eth_dev)
+{
+ RTE_SET_USED(eth_dev);
+
+ return -ENODEV;
+}
+
+static int
+rnp_eth_dev_uninit(struct rte_eth_dev *eth_dev)
+{
+ RTE_SET_USED(eth_dev);
+
+ return -ENODEV;
+}
+
+static int
+rnp_pci_remove(struct rte_pci_device *pci_dev)
+{
+ struct rte_eth_dev *eth_dev;
+ int rc;
+
+ eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
+
+ if (eth_dev) {
+ /* Cleanup eth dev */
+ rc = rte_eth_dev_pci_generic_remove(pci_dev,
+ rnp_eth_dev_uninit);
+ if (rc)
+ return rc;
+ }
+ /* Nothing to be done for secondary processes */
+ if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+ return 0;
+
+ return 0;
+}
+
+static int
+rnp_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev)
+{
+ int rc;
+
+ RTE_SET_USED(pci_drv);
+
+ rc = rte_eth_dev_pci_generic_probe(pci_dev, sizeof(struct rnp_eth_port),
+ rnp_eth_dev_init);
+
+ /* On error on secondary, recheck if port exists in primary or
+ * in mid of detach state.
+ */
+ if (rte_eal_process_type() != RTE_PROC_PRIMARY && rc)
+ if (!rte_eth_dev_allocated(pci_dev->device.name))
+ return 0;
+ return rc;
+}
+
+static const struct rte_pci_id pci_id_rnp_map[] = {
+ {
+ RTE_PCI_DEVICE(PCI_VENDOR_ID_MUCSE, RNP_DEV_ID_N10G)
+ },
+ {
+ .vendor_id = 0,
+ },
+};
+
+static struct rte_pci_driver rte_rnp_pmd = {
+ .id_table = pci_id_rnp_map,
+ .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
+ .probe = rnp_pci_probe,
+ .remove = rnp_pci_remove,
+};
+
+RTE_PMD_REGISTER_PCI(net_rnp, rte_rnp_pmd);
+RTE_PMD_REGISTER_PCI_TABLE(net_rnp, pci_id_rnp_map);
+RTE_PMD_REGISTER_KMOD_DEP(net_rnp, "igb_uio | uio_pci_generic | vfio-pci");
--
2.27.0
^ permalink raw reply related [flat|nested] 27+ messages in thread* Re: [PATCH v6 2/8] net/rnp: add ethdev probe and remove
2023-09-01 2:30 ` [PATCH v6 2/8] net/rnp: add ethdev probe and remove Wenbo Cao
@ 2023-09-05 15:36 ` Ferruh Yigit
2023-09-06 10:42 ` 11
0 siblings, 1 reply; 27+ messages in thread
From: Ferruh Yigit @ 2023-09-05 15:36 UTC (permalink / raw)
To: Wenbo Cao, Anatoly Burakov; +Cc: dev, thomas, andrew.rybchenko, yaojun
On 9/1/2023 3:30 AM, Wenbo Cao wrote:
> Add basic PCIe ethdev probe and remove.
>
> Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
> ---
> drivers/net/rnp/rnp.h | 13 ++++++
> drivers/net/rnp/rnp_ethdev.c | 83 ++++++++++++++++++++++++++++++++++++
> 2 files changed, 96 insertions(+)
> create mode 100644 drivers/net/rnp/rnp.h
>
> diff --git a/drivers/net/rnp/rnp.h b/drivers/net/rnp/rnp.h
> new file mode 100644
> index 0000000000..76d281cc0a
> --- /dev/null
> +++ b/drivers/net/rnp/rnp.h
> @@ -0,0 +1,13 @@
> +/* SPDX-License-Identifier: BSD-3-Clause
> + * Copyright(C) 2023 Mucse IC Design Ltd.
> + */
> +#ifndef __RNP_H__
> +#define __RNP_H__
> +
> +#define PCI_VENDOR_ID_MUCSE (0x8848)
> +#define RNP_DEV_ID_N10G (0x1000)
> +
> +struct rnp_eth_port {
> +} __rte_cache_aligned;
> +
Is the struct needs to be cache aligned?
> +#endif /* __RNP_H__ */
> diff --git a/drivers/net/rnp/rnp_ethdev.c b/drivers/net/rnp/rnp_ethdev.c
> index 9ce3c0b497..390f2e7743 100644
> --- a/drivers/net/rnp/rnp_ethdev.c
> +++ b/drivers/net/rnp/rnp_ethdev.c
> @@ -1,3 +1,86 @@
> /* SPDX-License-Identifier: BSD-3-Clause
> * Copyright(C) 2023 Mucse IC Design Ltd.
> */
> +
> +#include <ethdev_pci.h>
> +#include <rte_io.h>
> +#include <rte_malloc.h>
> +
> +#include "rnp.h"
> +
> +static int
> +rnp_eth_dev_init(struct rte_eth_dev *eth_dev)
> +{
> + RTE_SET_USED(eth_dev);
> +
> + return -ENODEV;
> +}
> +
> +static int
> +rnp_eth_dev_uninit(struct rte_eth_dev *eth_dev)
> +{
> + RTE_SET_USED(eth_dev);
> +
> + return -ENODEV;
> +}
> +
> +static int
> +rnp_pci_remove(struct rte_pci_device *pci_dev)
> +{
> + struct rte_eth_dev *eth_dev;
> + int rc;
> +
> + eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
> +
> + if (eth_dev) {
> + /* Cleanup eth dev */
> + rc = rte_eth_dev_pci_generic_remove(pci_dev,
> + rnp_eth_dev_uninit);
> + if (rc)
> + return rc;
> + }
> + /* Nothing to be done for secondary processes */
> + if (rte_eal_process_type() != RTE_PROC_PRIMARY)
> + return 0;
> +
Primary also don't do anything after this stage, I suggest adding this
code when primary & secondary diverges.
> + return 0;
> +}
> +
> +static int
> +rnp_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev)
> +{
> + int rc;
> +
> + RTE_SET_USED(pci_drv);
> +
> + rc = rte_eth_dev_pci_generic_probe(pci_dev, sizeof(struct rnp_eth_port),
> + rnp_eth_dev_init);
> +
> + /* On error on secondary, recheck if port exists in primary or
> + * in mid of detach state.
> + */
> + if (rte_eal_process_type() != RTE_PROC_PRIMARY && rc)
> + if (!rte_eth_dev_allocated(pci_dev->device.name))
> + return 0;
Why this additional secondary check is required?
'rte_eth_dev_pci_generic_probe()' should be dealing with
primary/secondary process case.
^ permalink raw reply [flat|nested] 27+ messages in thread* RE: [PATCH v6 2/8] net/rnp: add ethdev probe and remove
2023-09-05 15:36 ` Ferruh Yigit
@ 2023-09-06 10:42 ` 11
0 siblings, 0 replies; 27+ messages in thread
From: 11 @ 2023-09-06 10:42 UTC (permalink / raw)
To: 'Ferruh Yigit', 'Anatoly Burakov'
Cc: dev, thomas, andrew.rybchenko, yaojun
> -----Original Message-----
> From: Ferruh Yigit <ferruh.yigit@amd.com>
> Sent: 2023年9月5日 23:36
> To: Wenbo Cao <caowenbo@mucse.com>; Anatoly Burakov
> <anatoly.burakov@intel.com>
> Cc: dev@dpdk.org; thomas@monjalon.net; andrew.rybchenko@oktetlabs.ru;
> yaojun@mucse.com
> Subject: Re: [PATCH v6 2/8] net/rnp: add ethdev probe and remove
>
> On 9/1/2023 3:30 AM, Wenbo Cao wrote:
> > Add basic PCIe ethdev probe and remove.
> >
> > Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
> > ---
> > drivers/net/rnp/rnp.h | 13 ++++++
> > drivers/net/rnp/rnp_ethdev.c | 83
> > ++++++++++++++++++++++++++++++++++++
> > 2 files changed, 96 insertions(+)
> > create mode 100644 drivers/net/rnp/rnp.h
> >
> > diff --git a/drivers/net/rnp/rnp.h b/drivers/net/rnp/rnp.h new file
> > mode 100644 index 0000000000..76d281cc0a
> > --- /dev/null
> > +++ b/drivers/net/rnp/rnp.h
> > @@ -0,0 +1,13 @@
> > +/* SPDX-License-Identifier: BSD-3-Clause
> > + * Copyright(C) 2023 Mucse IC Design Ltd.
> > + */
> > +#ifndef __RNP_H__
> > +#define __RNP_H__
> > +
> > +#define PCI_VENDOR_ID_MUCSE (0x8848)
> > +#define RNP_DEV_ID_N10G (0x1000)
> > +
> > +struct rnp_eth_port {
> > +} __rte_cache_aligned;
> > +
>
> Is the struct needs to be cache aligned?
>
This struct may use frequently so I want it to fast to load.
> > +#endif /* __RNP_H__ */
> > diff --git a/drivers/net/rnp/rnp_ethdev.c
> > b/drivers/net/rnp/rnp_ethdev.c index 9ce3c0b497..390f2e7743 100644
> > --- a/drivers/net/rnp/rnp_ethdev.c
> > +++ b/drivers/net/rnp/rnp_ethdev.c
> > @@ -1,3 +1,86 @@
> > /* SPDX-License-Identifier: BSD-3-Clause
> > * Copyright(C) 2023 Mucse IC Design Ltd.
> > */
> > +
> > +#include <ethdev_pci.h>
> > +#include <rte_io.h>
> > +#include <rte_malloc.h>
> > +
> > +#include "rnp.h"
> > +
> > +static int
> > +rnp_eth_dev_init(struct rte_eth_dev *eth_dev) {
> > + RTE_SET_USED(eth_dev);
> > +
> > + return -ENODEV;
> > +}
> > +
> > +static int
> > +rnp_eth_dev_uninit(struct rte_eth_dev *eth_dev) {
> > + RTE_SET_USED(eth_dev);
> > +
> > + return -ENODEV;
> > +}
> > +
> > +static int
> > +rnp_pci_remove(struct rte_pci_device *pci_dev) {
> > + struct rte_eth_dev *eth_dev;
> > + int rc;
> > +
> > + eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
> > +
> > + if (eth_dev) {
> > + /* Cleanup eth dev */
> > + rc = rte_eth_dev_pci_generic_remove(pci_dev,
> > + rnp_eth_dev_uninit);
> > + if (rc)
> > + return rc;
> > + }
> > + /* Nothing to be done for secondary processes */
> > + if (rte_eal_process_type() != RTE_PROC_PRIMARY)
> > + return 0;
> > +
>
> Primary also don't do anything after this stage, I suggest adding this code when
> primary & secondary diverges.
>
Ok, I get it.
> > + return 0;
> > +}
> > +
> > +static int
> > +rnp_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device
> > +*pci_dev) {
> > + int rc;
> > +
> > + RTE_SET_USED(pci_drv);
> > +
> > + rc = rte_eth_dev_pci_generic_probe(pci_dev, sizeof(struct
> rnp_eth_port),
> > + rnp_eth_dev_init);
> > +
> > + /* On error on secondary, recheck if port exists in primary or
> > + * in mid of detach state.
> > + */
> > + if (rte_eal_process_type() != RTE_PROC_PRIMARY && rc)
> > + if (!rte_eth_dev_allocated(pci_dev->device.name))
> > + return 0;
>
> Why this additional secondary check is required?
> 'rte_eth_dev_pci_generic_probe()' should be dealing with primary/secondary
> process case.
>
Because of multiple port mode single pcie-bdf the other port
Alloc use xx:xx.0_0, xx:xx.0_1, xx:xx.0_2, xx:xx.0_0, use rnp_alloc_eth_port api.
^ permalink raw reply [flat|nested] 27+ messages in thread
* [PATCH v6 3/8] net/rnp: add device init and uninit
2023-09-01 2:30 [PATCH v6 0/8] [v6]drivers/net Add Support mucse N10 Pmd Driver Wenbo Cao
2023-09-01 2:30 ` [PATCH v6 1/8] net/rnp: add skeleton Wenbo Cao
2023-09-01 2:30 ` [PATCH v6 2/8] net/rnp: add ethdev probe and remove Wenbo Cao
@ 2023-09-01 2:30 ` Wenbo Cao
2023-09-05 15:44 ` Ferruh Yigit
2023-09-01 2:30 ` [PATCH v6 4/8] net/rnp: add mbx basic api feature Wenbo Cao
` (5 subsequent siblings)
8 siblings, 1 reply; 27+ messages in thread
From: Wenbo Cao @ 2023-09-01 2:30 UTC (permalink / raw)
To: Wenbo Cao, Anatoly Burakov
Cc: dev, ferruh.yigit, thomas, andrew.rybchenko, yaojun
Add basic init and uninit function
Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
---
drivers/net/rnp/base/rnp_hw.h | 19 ++++
drivers/net/rnp/meson.build | 1 +
drivers/net/rnp/rnp.h | 25 +++++
drivers/net/rnp/rnp_ethdev.c | 196 +++++++++++++++++++++++++++++++++-
drivers/net/rnp/rnp_logs.h | 34 ++++++
drivers/net/rnp/rnp_osdep.h | 30 ++++++
6 files changed, 300 insertions(+), 5 deletions(-)
create mode 100644 drivers/net/rnp/base/rnp_hw.h
create mode 100644 drivers/net/rnp/rnp_logs.h
create mode 100644 drivers/net/rnp/rnp_osdep.h
diff --git a/drivers/net/rnp/base/rnp_hw.h b/drivers/net/rnp/base/rnp_hw.h
new file mode 100644
index 0000000000..d80d23f4b4
--- /dev/null
+++ b/drivers/net/rnp/base/rnp_hw.h
@@ -0,0 +1,19 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(C) 2023 Mucse IC Design Ltd.
+ */
+#ifndef __RNP_HW_H__
+#define __RNP_HW_H__
+
+struct rnp_eth_adapter;
+struct rnp_hw {
+ struct rnp_eth_adapter *back;
+ void *iobar0;
+ uint32_t iobar0_len;
+ void *iobar4;
+ uint32_t iobar4_len;
+
+ uint16_t device_id;
+ uint16_t vendor_id;
+} __rte_cache_aligned;
+
+#endif /* __RNP_H__*/
diff --git a/drivers/net/rnp/meson.build b/drivers/net/rnp/meson.build
index 4f37c6b456..f85d597e68 100644
--- a/drivers/net/rnp/meson.build
+++ b/drivers/net/rnp/meson.build
@@ -9,3 +9,4 @@ endif
sources = files(
'rnp_ethdev.c',
)
+includes += include_directories('base')
diff --git a/drivers/net/rnp/rnp.h b/drivers/net/rnp/rnp.h
index 76d281cc0a..cab1b8e85d 100644
--- a/drivers/net/rnp/rnp.h
+++ b/drivers/net/rnp/rnp.h
@@ -4,10 +4,35 @@
#ifndef __RNP_H__
#define __RNP_H__
+#include "base/rnp_hw.h"
+
#define PCI_VENDOR_ID_MUCSE (0x8848)
#define RNP_DEV_ID_N10G (0x1000)
+#define RNP_MAX_PORT_OF_PF (4)
+#define RNP_CFG_BAR (4)
+#define RNP_PF_INFO_BAR (0)
struct rnp_eth_port {
+ struct rnp_eth_adapter *adapt;
+ struct rte_eth_dev *eth_dev;
+} __rte_cache_aligned;
+
+struct rnp_share_ops {
} __rte_cache_aligned;
+struct rnp_eth_adapter {
+ struct rnp_hw hw;
+ struct rte_pci_device *pdev;
+ struct rte_eth_dev *eth_dev; /* primary eth_dev */
+ struct rnp_eth_port *ports[RNP_MAX_PORT_OF_PF];
+ struct rnp_share_ops *share_priv;
+
+ uint8_t num_ports; /* Cur Pf Has physical Port Num */
+} __rte_cache_aligned;
+
+#define RNP_DEV_TO_PORT(eth_dev) \
+ (((struct rnp_eth_port *)((eth_dev)->data->dev_private)))
+#define RNP_DEV_TO_ADAPTER(eth_dev) \
+ ((struct rnp_eth_adapter *)(RNP_DEV_TO_PORT(eth_dev)->adapt))
+
#endif /* __RNP_H__ */
diff --git a/drivers/net/rnp/rnp_ethdev.c b/drivers/net/rnp/rnp_ethdev.c
index 390f2e7743..ae737643a7 100644
--- a/drivers/net/rnp/rnp_ethdev.c
+++ b/drivers/net/rnp/rnp_ethdev.c
@@ -5,23 +5,198 @@
#include <ethdev_pci.h>
#include <rte_io.h>
#include <rte_malloc.h>
+#include <ethdev_driver.h>
#include "rnp.h"
+#include "rnp_logs.h"
static int
-rnp_eth_dev_init(struct rte_eth_dev *eth_dev)
+rnp_mac_rx_disable(struct rte_eth_dev *dev)
{
- RTE_SET_USED(eth_dev);
+ RTE_SET_USED(dev);
- return -ENODEV;
+ return 0;
+}
+
+static int
+rnp_mac_tx_disable(struct rte_eth_dev *dev)
+{
+ RTE_SET_USED(dev);
+
+ return 0;
+}
+
+static int rnp_dev_close(struct rte_eth_dev *dev)
+{
+ RTE_SET_USED(dev);
+
+ return 0;
+}
+
+/* Features supported by this driver */
+static const struct eth_dev_ops rnp_eth_dev_ops = {
+};
+
+static int
+rnp_init_port_resource(struct rnp_eth_adapter *adapter,
+ struct rte_eth_dev *dev,
+ char *name,
+ uint8_t p_id)
+{
+ struct rnp_eth_port *port = RNP_DEV_TO_PORT(dev);
+
+ port->eth_dev = dev;
+ adapter->ports[p_id] = port;
+ dev->dev_ops = &rnp_eth_dev_ops;
+ RTE_SET_USED(name);
+
+ return 0;
+}
+
+static struct rte_eth_dev *
+rnp_alloc_eth_port(struct rte_pci_device *primary_pci, char *name)
+{
+ struct rnp_eth_port *port;
+ struct rte_eth_dev *eth_dev;
+
+ eth_dev = rte_eth_dev_allocate(name);
+ if (!eth_dev) {
+ RNP_PMD_DRV_LOG(ERR, "Could not allocate "
+ "eth_dev for %s\n", name);
+ return NULL;
+ }
+ port = rte_zmalloc_socket(name,
+ sizeof(*port),
+ RTE_CACHE_LINE_SIZE,
+ primary_pci->device.numa_node);
+ if (!port) {
+ RNP_PMD_DRV_LOG(ERR, "Could not allocate "
+ "rnp_eth_port for %s\n", name);
+ return NULL;
+ }
+ eth_dev->data->dev_private = port;
+ eth_dev->process_private = calloc(1, sizeof(struct rnp_share_ops));
+ if (!eth_dev->process_private) {
+ RNP_PMD_DRV_LOG(ERR, "Could not calloc "
+ "for Process_priv\n");
+ goto fail_calloc;
+ }
+ return eth_dev;
+fail_calloc:
+ rte_free(port);
+ rte_eth_dev_release_port(eth_dev);
+
+ return NULL;
+}
+
+static int
+rnp_eth_dev_init(struct rte_eth_dev *dev)
+{
+ struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+ struct rnp_eth_adapter *adapter = NULL;
+ char name[RTE_ETH_NAME_MAX_LEN] = " ";
+ struct rnp_eth_port *port = NULL;
+ struct rte_eth_dev *eth_dev;
+ struct rnp_hw *hw = NULL;
+ int32_t p_id;
+ int ret;
+
+ PMD_INIT_FUNC_TRACE();
+
+ if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+ return 0;
+ memset(name, 0, sizeof(name));
+ snprintf(name, sizeof(name), "rnp_adapter_%d", dev->data->port_id);
+ adapter = rte_zmalloc(name, sizeof(struct rnp_eth_adapter), 0);
+ if (!adapter) {
+ RNP_PMD_DRV_LOG(ERR, "zmalloc for adapter failed\n");
+ return -ENOMEM;
+ }
+ hw = &adapter->hw;
+ adapter->pdev = pci_dev;
+ adapter->eth_dev = dev;
+ adapter->num_ports = 1;
+ hw->back = adapter;
+ hw->iobar4 = pci_dev->mem_resource[RNP_CFG_BAR].addr;
+ hw->iobar0 = pci_dev->mem_resource[RNP_PF_INFO_BAR].addr;
+ hw->iobar4_len = pci_dev->mem_resource[RNP_CFG_BAR].len;
+ hw->iobar0_len = pci_dev->mem_resource[RNP_PF_INFO_BAR].len;
+ hw->device_id = pci_dev->id.device_id;
+ hw->vendor_id = pci_dev->id.vendor_id;
+ hw->device_id = pci_dev->id.device_id;
+ for (p_id = 0; p_id < adapter->num_ports; p_id++) {
+ /* port 0 resource has been allocated When Probe */
+ if (!p_id) {
+ eth_dev = dev;
+ } else {
+ snprintf(name, sizeof(name), "%s_%d",
+ adapter->pdev->device.name,
+ p_id);
+ eth_dev = rnp_alloc_eth_port(pci_dev, name);
+ if (eth_dev)
+ rte_memcpy(eth_dev->process_private,
+ adapter->share_priv,
+ sizeof(*adapter->share_priv));
+ if (!eth_dev) {
+ ret = -ENOMEM;
+ goto eth_alloc_error;
+ }
+ }
+ ret = rnp_init_port_resource(adapter, eth_dev, name, p_id);
+ if (ret)
+ goto eth_alloc_error;
+
+ rnp_mac_rx_disable(eth_dev);
+ rnp_mac_tx_disable(eth_dev);
+ }
+
+ return 0;
+eth_alloc_error:
+ for (p_id = 0; p_id < adapter->num_ports; p_id++) {
+ port = adapter->ports[p_id];
+ if (!port)
+ continue;
+ if (port->eth_dev) {
+ rnp_dev_close(port->eth_dev);
+ rte_eth_dev_release_port(port->eth_dev);
+ if (port->eth_dev->process_private)
+ free(port->eth_dev->process_private);
+ }
+ rte_free(port);
+ }
+ rte_free(adapter);
+
+ return 0;
}
static int
rnp_eth_dev_uninit(struct rte_eth_dev *eth_dev)
{
- RTE_SET_USED(eth_dev);
+ struct rnp_eth_adapter *adapter = RNP_DEV_TO_ADAPTER(eth_dev);
+ struct rnp_eth_port *port = NULL;
+ uint8_t p_id;
+
+ if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+ return 0;
- return -ENODEV;
+ if (adapter->eth_dev != eth_dev) {
+ RNP_PMD_DRV_LOG(ERR, "Input Argument ethdev "
+ "Isn't Primary Ethdev\n");
+ return -EINVAL;
+ }
+ for (p_id = 0; p_id < adapter->num_ports; p_id++) {
+ port = adapter->ports[p_id];
+ if (!port)
+ continue;
+ if (port->eth_dev) {
+ rnp_dev_close(port->eth_dev);
+ /* Just Release Not Primary Port Allocated By PMD */
+ if (p_id)
+ rte_eth_dev_release_port(port->eth_dev);
+ }
+ }
+
+ return 0;
}
static int
@@ -84,3 +259,14 @@ static struct rte_pci_driver rte_rnp_pmd = {
RTE_PMD_REGISTER_PCI(net_rnp, rte_rnp_pmd);
RTE_PMD_REGISTER_PCI_TABLE(net_rnp, pci_id_rnp_map);
RTE_PMD_REGISTER_KMOD_DEP(net_rnp, "igb_uio | uio_pci_generic | vfio-pci");
+
+RTE_LOG_REGISTER_SUFFIX(rnp_init_logtype, init, NOTICE);
+RTE_LOG_REGISTER_SUFFIX(rnp_drv_logtype, driver, NOTICE);
+
+#ifdef RTE_LIBRTE_RNP_DEBUG_RX
+ RTE_LOG_REGISTER_SUFFIX(rnp_rx_logtype, rx, DEBUG);
+#endif
+
+#ifdef RTE_LIBRTE_RNP_DEBUG_TX
+ RTE_LOG_REGISTER_SUFFIX(rnp_tx_logtype, tx, DEBUG);
+#endif
diff --git a/drivers/net/rnp/rnp_logs.h b/drivers/net/rnp/rnp_logs.h
new file mode 100644
index 0000000000..1b3ee33745
--- /dev/null
+++ b/drivers/net/rnp/rnp_logs.h
@@ -0,0 +1,34 @@
+#ifndef __RNP_LOGS_H__
+#define __RNP_LOGS_H__
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(C) 2023 Mucse IC Design Ltd.
+ */
+extern int rnp_init_logtype;
+
+#define RNP_PMD_INIT_LOG(level, fmt, args...) \
+ rte_log(RTE_LOG_##level, rnp_init_logtype, \
+ "%s() " fmt, __func__, ##args)
+#define PMD_INIT_FUNC_TRACE() RNP_PMD_INIT_LOG(DEBUG, " >>")
+extern int rnp_drv_logtype;
+#define RNP_PMD_DRV_LOG(level, fmt, args...) \
+ rte_log(RTE_LOG_##level, rnp_drv_logtype, \
+ "%s() " fmt, __func__, ##args)
+#ifdef RTE_LIBRTE_RNP_DEBUG_RX
+extern int rnp_rx_logtype;
+#define RNP_PMD_RX_LOG(level, fmt, args...) \
+ rte_log(RTE_LOG_ ## level, rnp_rx_logtype, \
+ "%s(): " fmt "\n", __func__, ##args)
+#else
+#define RNP_PMD_RX_LOG(level, fmt, args...) do { } while (0)
+#endif
+
+#ifdef RTE_LIBRTE_RNP_DEBUG_TX
+extern int rnp_tx_logtype;
+#define PMD_TX_LOG(level, fmt, args...) \
+ rte_log(RTE_LOG_ ## level, rnp_tx_logtype, \
+ "%s(): " fmt "\n", __func__, ##args)
+#else
+#define PMD_TX_LOG(level, fmt, args...) do { } while (0)
+#endif
+
+#endif /* __RNP_LOGS_H__ */
diff --git a/drivers/net/rnp/rnp_osdep.h b/drivers/net/rnp/rnp_osdep.h
new file mode 100644
index 0000000000..5685dd2404
--- /dev/null
+++ b/drivers/net/rnp/rnp_osdep.h
@@ -0,0 +1,30 @@
+#ifndef __RNP_OSDEP_H__
+#define __RNP_OSDEP_H__
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(C) 2023 Mucse IC Design Ltd.
+ */
+#include <stdint.h>
+
+#include <rte_byteorder.h>
+
+#define __iomem
+#define _RING_(off) ((off) + 0x000000)
+#define _DMA_(off) ((off))
+#define _GLB_(off) ((off) + 0x000000)
+#define _NIC_(off) ((off) + 0x000000)
+#define _ETH_(off) ((off))
+#define _MAC_(off) ((off))
+#define BIT(n) (1UL << (n))
+#define BIT64(n) (1ULL << (n))
+#define BITS_PER_LONG (__SIZEOF_LONG__ * 8)
+#define GENMASK(h, l) \
+ (((~0UL) << (l)) & (~0UL >> (BITS_PER_LONG - 1 - (h))))
+
+typedef uint8_t u8;
+typedef uint16_t u16;
+typedef uint32_t u32;
+typedef uint64_t u64;
+typedef int32_t s32;
+typedef int16_t s16;
+typedef int8_t s8;
+#endif /* __RNP_OSDEP_H__ */
--
2.27.0
^ permalink raw reply related [flat|nested] 27+ messages in thread* Re: [PATCH v6 3/8] net/rnp: add device init and uninit
2023-09-01 2:30 ` [PATCH v6 3/8] net/rnp: add device init and uninit Wenbo Cao
@ 2023-09-05 15:44 ` Ferruh Yigit
2023-09-06 11:03 ` 11
0 siblings, 1 reply; 27+ messages in thread
From: Ferruh Yigit @ 2023-09-05 15:44 UTC (permalink / raw)
To: Wenbo Cao, Anatoly Burakov; +Cc: dev, thomas, andrew.rybchenko, yaojun
On 9/1/2023 3:30 AM, Wenbo Cao wrote:
> Add basic init and uninit function
>
> Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
> ---
> drivers/net/rnp/base/rnp_hw.h | 19 ++++
> drivers/net/rnp/meson.build | 1 +
> drivers/net/rnp/rnp.h | 25 +++++
> drivers/net/rnp/rnp_ethdev.c | 196 +++++++++++++++++++++++++++++++++-
> drivers/net/rnp/rnp_logs.h | 34 ++++++
> drivers/net/rnp/rnp_osdep.h | 30 ++++++
> 6 files changed, 300 insertions(+), 5 deletions(-)
> create mode 100644 drivers/net/rnp/base/rnp_hw.h
> create mode 100644 drivers/net/rnp/rnp_logs.h
> create mode 100644 drivers/net/rnp/rnp_osdep.h
>
'rnp_osdep.h' not used at all in the patch, can you move it where it is
used?
> diff --git a/drivers/net/rnp/base/rnp_hw.h b/drivers/net/rnp/base/rnp_hw.h
> new file mode 100644
> index 0000000000..d80d23f4b4
> --- /dev/null
> +++ b/drivers/net/rnp/base/rnp_hw.h
> @@ -0,0 +1,19 @@
> +/* SPDX-License-Identifier: BSD-3-Clause
> + * Copyright(C) 2023 Mucse IC Design Ltd.
> + */
> +#ifndef __RNP_HW_H__
> +#define __RNP_HW_H__
> +
> +struct rnp_eth_adapter;
> +struct rnp_hw {
> + struct rnp_eth_adapter *back;
> + void *iobar0;
> + uint32_t iobar0_len;
> + void *iobar4;
> + uint32_t iobar4_len;
> +
> + uint16_t device_id;
> + uint16_t vendor_id;
> +} __rte_cache_aligned;
> +
All structs seems alligned to the cache, although it may not hurt, it is
better to align ones that is necessary for performance.
<...>
> +
> +static struct rte_eth_dev *
> +rnp_alloc_eth_port(struct rte_pci_device *primary_pci, char *name)
Why variable name is 'primary_pci', are there different pci device types?
> +{
> + struct rnp_eth_port *port;
> + struct rte_eth_dev *eth_dev;
> +
> + eth_dev = rte_eth_dev_allocate(name);
> + if (!eth_dev) {
> + RNP_PMD_DRV_LOG(ERR, "Could not allocate "
> + "eth_dev for %s\n", name);
Please don't split the log message, instead can do:
RNP_PMD_DRV_LOG(ERR,
"Could not allocate eth_dev for %s\n",
name);
Same for all log messages.
> + return NULL;
> + }
> + port = rte_zmalloc_socket(name,
> + sizeof(*port),
> + RTE_CACHE_LINE_SIZE,
> + primary_pci->device.numa_node);
> + if (!port) {
> + RNP_PMD_DRV_LOG(ERR, "Could not allocate "
> + "rnp_eth_port for %s\n", name);
> + return NULL;
Should 'eth_dev' released here?
> + }
> + eth_dev->data->dev_private = port;
> + eth_dev->process_private = calloc(1, sizeof(struct rnp_share_ops));
> + if (!eth_dev->process_private) {
> + RNP_PMD_DRV_LOG(ERR, "Could not calloc "
> + "for Process_priv\n");
> + goto fail_calloc;
> + }
> + return eth_dev;
> +fail_calloc:
> + rte_free(port);
> + rte_eth_dev_release_port(eth_dev);
> +
> + return NULL;
> +}
> +
> +static int
> +rnp_eth_dev_init(struct rte_eth_dev *dev)
> +{
> + struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
> + struct rnp_eth_adapter *adapter = NULL;
> + char name[RTE_ETH_NAME_MAX_LEN] = " ";
> + struct rnp_eth_port *port = NULL;
> + struct rte_eth_dev *eth_dev;
> + struct rnp_hw *hw = NULL;
> + int32_t p_id;
> + int ret;
> +
> + PMD_INIT_FUNC_TRACE();
> +
> + if (rte_eal_process_type() != RTE_PROC_PRIMARY)
> + return 0;
> + memset(name, 0, sizeof(name));
> + snprintf(name, sizeof(name), "rnp_adapter_%d", dev->data->port_id);
> + adapter = rte_zmalloc(name, sizeof(struct rnp_eth_adapter), 0);
> + if (!adapter) {
> + RNP_PMD_DRV_LOG(ERR, "zmalloc for adapter failed\n");
> + return -ENOMEM;
> + }
> + hw = &adapter->hw;
> + adapter->pdev = pci_dev;
> + adapter->eth_dev = dev;
> + adapter->num_ports = 1;
This is hardcoded value, so no need the loop below but perhaps it is for
future.
> + hw->back = adapter;
> + hw->iobar4 = pci_dev->mem_resource[RNP_CFG_BAR].addr;
> + hw->iobar0 = pci_dev->mem_resource[RNP_PF_INFO_BAR].addr;
> + hw->iobar4_len = pci_dev->mem_resource[RNP_CFG_BAR].len;
> + hw->iobar0_len = pci_dev->mem_resource[RNP_PF_INFO_BAR].len;
> + hw->device_id = pci_dev->id.device_id;
> + hw->vendor_id = pci_dev->id.vendor_id;
> + hw->device_id = pci_dev->id.device_id;
> + for (p_id = 0; p_id < adapter->num_ports; p_id++) {
> + /* port 0 resource has been allocated When Probe */
Why 'When' & 'Probe' starts with uppercase, similar usage exists many
places, although that is not an issue, just catches eye, can you please
fix them if there is no specific reason.
> + if (!p_id) {
> + eth_dev = dev;
> + } else {
> + snprintf(name, sizeof(name), "%s_%d",
> + adapter->pdev->device.name,
> + p_id);
> + eth_dev = rnp_alloc_eth_port(pci_dev, name);
> + if (eth_dev)
> + rte_memcpy(eth_dev->process_private,
> + adapter->share_priv,
> + sizeof(*adapter->share_priv));
> + if (!eth_dev) {
This can be 'else' leg of above branch.
> + ret = -ENOMEM;
> + goto eth_alloc_error;
> + }
> + }
> + ret = rnp_init_port_resource(adapter, eth_dev, name, p_id);
> + if (ret)
> + goto eth_alloc_error;
'rnp_init_port_resource()' can't return error but perhaps this check is
for future.
> +
> + rnp_mac_rx_disable(eth_dev);
> + rnp_mac_tx_disable(eth_dev);
> + }
> +
> + return 0;
> +eth_alloc_error:
> + for (p_id = 0; p_id < adapter->num_ports; p_id++) {
> + port = adapter->ports[p_id];
> + if (!port)
> + continue;
> + if (port->eth_dev) {
> + rnp_dev_close(port->eth_dev);
> + rte_eth_dev_release_port(port->eth_dev);
> + if (port->eth_dev->process_private)
This should crash, because 'port' is 'dev_private' and
'rte_eth_dev_release_port()' frees the dev_private, so can't access
'port->' here.
Also 'process_private' set to NULL in the 'rte_eth_dev_release_port()',
it should be freed in 'rnp_dev_close()', not here.
> + free(port->eth_dev->process_private);
> + }
> + rte_free(port);
> + }
> + rte_free(adapter);
> +
> + return 0;
> }
>
> static int
> rnp_eth_dev_uninit(struct rte_eth_dev *eth_dev)
> {
> - RTE_SET_USED(eth_dev);
> + struct rnp_eth_adapter *adapter = RNP_DEV_TO_ADAPTER(eth_dev);
> + struct rnp_eth_port *port = NULL;
> + uint8_t p_id;
> +
> + if (rte_eal_process_type() != RTE_PROC_PRIMARY)
> + return 0;
>
> - return -ENODEV;
> + if (adapter->eth_dev != eth_dev) {
> + RNP_PMD_DRV_LOG(ERR, "Input Argument ethdev "
> + "Isn't Primary Ethdev\n");
> + return -EINVAL;
> + }
> + for (p_id = 0; p_id < adapter->num_ports; p_id++) {
> + port = adapter->ports[p_id];
> + if (!port)
> + continue;
> + if (port->eth_dev) {
> + rnp_dev_close(port->eth_dev);
> + /* Just Release Not Primary Port Allocated By PMD */
> + if (p_id)
> + rte_eth_dev_release_port(port->eth_dev);
> + }
> + }
> +
> + return 0;
> }
>
> static int
> @@ -84,3 +259,14 @@ static struct rte_pci_driver rte_rnp_pmd = {
> RTE_PMD_REGISTER_PCI(net_rnp, rte_rnp_pmd);
> RTE_PMD_REGISTER_PCI_TABLE(net_rnp, pci_id_rnp_map);
> RTE_PMD_REGISTER_KMOD_DEP(net_rnp, "igb_uio | uio_pci_generic | vfio-pci");
> +
> +RTE_LOG_REGISTER_SUFFIX(rnp_init_logtype, init, NOTICE);
> +RTE_LOG_REGISTER_SUFFIX(rnp_drv_logtype, driver, NOTICE);
> +
Do you really need two different log types? What about reducing it to one?
<...>
> diff --git a/drivers/net/rnp/rnp_osdep.h b/drivers/net/rnp/rnp_osdep.h
> new file mode 100644
> index 0000000000..5685dd2404
> --- /dev/null
> +++ b/drivers/net/rnp/rnp_osdep.h
> @@ -0,0 +1,30 @@
> +#ifndef __RNP_OSDEP_H__
> +#define __RNP_OSDEP_H__
> +/* SPDX-License-Identifier: BSD-3-Clause
> + * Copyright(C) 2023 Mucse IC Design Ltd.
> + */
> +#include <stdint.h>
> +
> +#include <rte_byteorder.h>
> +
> +#define __iomem
> +#define _RING_(off) ((off) + 0x000000)
> +#define _DMA_(off) ((off))
> +#define _GLB_(off) ((off) + 0x000000)
> +#define _NIC_(off) ((off) + 0x000000)
> +#define _ETH_(off) ((off))
> +#define _MAC_(off) ((off))
> +#define BIT(n) (1UL << (n))
> +#define BIT64(n) (1ULL << (n))
DPDK has RTE_BIT64 / RTE_BIT32 macros that can be reused.
> +#define BITS_PER_LONG (__SIZEOF_LONG__ * 8)
> +#define GENMASK(h, l) \
> + (((~0UL) << (l)) & (~0UL >> (BITS_PER_LONG - 1 - (h))))
> +
> +typedef uint8_t u8;
> +typedef uint16_t u16;
> +typedef uint32_t u32;
> +typedef uint64_t u64;
> +typedef int32_t s32;
> +typedef int16_t s16;
> +typedef int8_t s8;
> +#endif /* __RNP_OSDEP_H__ */
^ permalink raw reply [flat|nested] 27+ messages in thread* RE: [PATCH v6 3/8] net/rnp: add device init and uninit
2023-09-05 15:44 ` Ferruh Yigit
@ 2023-09-06 11:03 ` 11
0 siblings, 0 replies; 27+ messages in thread
From: 11 @ 2023-09-06 11:03 UTC (permalink / raw)
To: 'Ferruh Yigit', 'Anatoly Burakov'
Cc: dev, thomas, andrew.rybchenko, yaojun
Hi Ferruh,
Thanks your kindly review, please see the below comment.
Regards Wenbo
> -----Original Message-----
> From: Ferruh Yigit <ferruh.yigit@amd.com>
> Sent: 2023年9月5日 23:44
> To: Wenbo Cao <caowenbo@mucse.com>; Anatoly Burakov
> <anatoly.burakov@intel.com>
> Cc: dev@dpdk.org; thomas@monjalon.net; andrew.rybchenko@oktetlabs.ru;
> yaojun@mucse.com
> Subject: Re: [PATCH v6 3/8] net/rnp: add device init and uninit
>
> On 9/1/2023 3:30 AM, Wenbo Cao wrote:
> > Add basic init and uninit function
> >
> > Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
> > ---
> > drivers/net/rnp/base/rnp_hw.h | 19 ++++
> > drivers/net/rnp/meson.build | 1 +
> > drivers/net/rnp/rnp.h | 25 +++++
> > drivers/net/rnp/rnp_ethdev.c | 196 +++++++++++++++++++++++++++++++++-
> > drivers/net/rnp/rnp_logs.h | 34 ++++++
> > drivers/net/rnp/rnp_osdep.h | 30 ++++++
> > 6 files changed, 300 insertions(+), 5 deletions(-) create mode
> > 100644 drivers/net/rnp/base/rnp_hw.h create mode 100644
> > drivers/net/rnp/rnp_logs.h create mode 100644
> > drivers/net/rnp/rnp_osdep.h
> >
>
> 'rnp_osdep.h' not used at all in the patch, can you move it where it is used?
>
I got it.
> > diff --git a/drivers/net/rnp/base/rnp_hw.h
> > b/drivers/net/rnp/base/rnp_hw.h new file mode 100644 index
> > 0000000000..d80d23f4b4
> > --- /dev/null
> > +++ b/drivers/net/rnp/base/rnp_hw.h
> > @@ -0,0 +1,19 @@
> > +/* SPDX-License-Identifier: BSD-3-Clause
> > + * Copyright(C) 2023 Mucse IC Design Ltd.
> > + */
> > +#ifndef __RNP_HW_H__
> > +#define __RNP_HW_H__
> > +
> > +struct rnp_eth_adapter;
> > +struct rnp_hw {
> > + struct rnp_eth_adapter *back;
> > + void *iobar0;
> > + uint32_t iobar0_len;
> > + void *iobar4;
> > + uint32_t iobar4_len;
> > +
> > + uint16_t device_id;
> > + uint16_t vendor_id;
> > +} __rte_cache_aligned;
> > +
>
> All structs seems alligned to the cache, although it may not hurt, it is better to
> align ones that is necessary for performance.
>
> <...>
>
For this I need to select a often used struct value?
> > +
> > +static struct rte_eth_dev *
> > +rnp_alloc_eth_port(struct rte_pci_device *primary_pci, char *name)
>
> Why variable name is 'primary_pci', are there different pci device types?
>
Because of the chip is just have two pcie-bdf, the can have max eight port,
So rte_eth_dev_pci_generic_probe alloc the first port, the other port is a fake string
Add for x:x.0_1, x:x.0_2, x:x.0_3, so the 'primary_pci' is to mark the real pci_addr
> > +{
> > + struct rnp_eth_port *port;
> > + struct rte_eth_dev *eth_dev;
> > +
> > + eth_dev = rte_eth_dev_allocate(name);
> > + if (!eth_dev) {
> > + RNP_PMD_DRV_LOG(ERR, "Could not allocate "
> > + "eth_dev for %s\n", name);
>
> Please don't split the log message, instead can do:
> RNP_PMD_DRV_LOG(ERR,
> "Could not allocate eth_dev for %s\n",
> name);
>
> Same for all log messages.
>
Ok, I got it.
> > + return NULL;
> > + }
> > + port = rte_zmalloc_socket(name,
> > + sizeof(*port),
> > + RTE_CACHE_LINE_SIZE,
> > + primary_pci->device.numa_node);
> > + if (!port) {
> > + RNP_PMD_DRV_LOG(ERR, "Could not allocate "
> > + "rnp_eth_port for %s\n", name);
> > + return NULL;
>
> Should 'eth_dev' released here?
>
Yes , this need to free resource.
> > + }
> > + eth_dev->data->dev_private = port;
> > + eth_dev->process_private = calloc(1, sizeof(struct rnp_share_ops));
> > + if (!eth_dev->process_private) {
> > + RNP_PMD_DRV_LOG(ERR, "Could not calloc "
> > + "for Process_priv\n");
> > + goto fail_calloc;
> > + }
> > + return eth_dev;
> > +fail_calloc:
> > + rte_free(port);
> > + rte_eth_dev_release_port(eth_dev);
> > +
> > + return NULL;
> > +}
> > +
> > +static int
> > +rnp_eth_dev_init(struct rte_eth_dev *dev) {
> > + struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
> > + struct rnp_eth_adapter *adapter = NULL;
> > + char name[RTE_ETH_NAME_MAX_LEN] = " ";
> > + struct rnp_eth_port *port = NULL;
> > + struct rte_eth_dev *eth_dev;
> > + struct rnp_hw *hw = NULL;
> > + int32_t p_id;
> > + int ret;
> > +
> > + PMD_INIT_FUNC_TRACE();
> > +
> > + if (rte_eal_process_type() != RTE_PROC_PRIMARY)
> > + return 0;
> > + memset(name, 0, sizeof(name));
> > + snprintf(name, sizeof(name), "rnp_adapter_%d", dev->data->port_id);
> > + adapter = rte_zmalloc(name, sizeof(struct rnp_eth_adapter), 0);
> > + if (!adapter) {
> > + RNP_PMD_DRV_LOG(ERR, "zmalloc for adapter failed\n");
> > + return -ENOMEM;
> > + }
> > + hw = &adapter->hw;
> > + adapter->pdev = pci_dev;
> > + adapter->eth_dev = dev;
> > + adapter->num_ports = 1;
>
> This is hardcoded value, so no need the loop below but perhaps it is for future.
>
I will change the commit sequence ,first achieve mailbox api get hw-info.
So adapter->num_ports info will get from firmware.
> > + hw->back = adapter;
> > + hw->iobar4 = pci_dev->mem_resource[RNP_CFG_BAR].addr;
> > + hw->iobar0 = pci_dev->mem_resource[RNP_PF_INFO_BAR].addr;
> > + hw->iobar4_len = pci_dev->mem_resource[RNP_CFG_BAR].len;
> > + hw->iobar0_len = pci_dev->mem_resource[RNP_PF_INFO_BAR].len;
> > + hw->device_id = pci_dev->id.device_id;
> > + hw->vendor_id = pci_dev->id.vendor_id;
> > + hw->device_id = pci_dev->id.device_id;
> > + for (p_id = 0; p_id < adapter->num_ports; p_id++) {
> > + /* port 0 resource has been allocated When Probe */
>
> Why 'When' & 'Probe' starts with uppercase, similar usage exists many places,
> although that is not an issue, just catches eye, can you please fix them if there is
> no specific reason.
>
There isn't special meaning, I will change it to low.
> > + if (!p_id) {
> > + eth_dev = dev;
> > + } else {
> > + snprintf(name, sizeof(name), "%s_%d",
> > + adapter->pdev->device.name,
> > + p_id);
> > + eth_dev = rnp_alloc_eth_port(pci_dev, name);
> > + if (eth_dev)
> > + rte_memcpy(eth_dev->process_private,
> > + adapter->share_priv,
> > + sizeof(*adapter->share_priv));
> > + if (!eth_dev) {
>
> This can be 'else' leg of above branch.
>
I will check the code.
> > + ret = -ENOMEM;
> > + goto eth_alloc_error;
> > + }
> > + }
> > + ret = rnp_init_port_resource(adapter, eth_dev, name, p_id);
> > + if (ret)
> > + goto eth_alloc_error;
>
> 'rnp_init_port_resource()' can't return error but perhaps this check is for future.
>
Thanks for you advise, I will consider it .
> > +
> > + rnp_mac_rx_disable(eth_dev);
> > + rnp_mac_tx_disable(eth_dev);
> > + }
> > +
> > + return 0;
> > +eth_alloc_error:
> > + for (p_id = 0; p_id < adapter->num_ports; p_id++) {
> > + port = adapter->ports[p_id];
> > + if (!port)
> > + continue;
> > + if (port->eth_dev) {
> > + rnp_dev_close(port->eth_dev);
> > + rte_eth_dev_release_port(port->eth_dev);
> > + if (port->eth_dev->process_private)
>
> This should crash, because 'port' is 'dev_private' and
> 'rte_eth_dev_release_port()' frees the dev_private, so can't access 'port->' here.
>
> > + for (p_id = 0; p_id < adapter->num_ports; p_id++) {
This adapter->num_ports need to use port init success num.
> Also 'process_private' set to NULL in the 'rte_eth_dev_release_port()', it should
> be freed in 'rnp_dev_close()', not here.
>
I have miss check the last dpdk version.
> > + free(port->eth_dev->process_private);
> > + }
> > + rte_free(port);
> > + }
> > + rte_free(adapter);
> > +
> > + return 0;
> > }
> >
> > static int
> > rnp_eth_dev_uninit(struct rte_eth_dev *eth_dev) {
> > - RTE_SET_USED(eth_dev);
> > + struct rnp_eth_adapter *adapter = RNP_DEV_TO_ADAPTER(eth_dev);
> > + struct rnp_eth_port *port = NULL;
> > + uint8_t p_id;
> > +
> > + if (rte_eal_process_type() != RTE_PROC_PRIMARY)
> > + return 0;
> >
> > - return -ENODEV;
> > + if (adapter->eth_dev != eth_dev) {
> > + RNP_PMD_DRV_LOG(ERR, "Input Argument ethdev "
> > + "Isn't Primary Ethdev\n");
> > + return -EINVAL;
> > + }
> > + for (p_id = 0; p_id < adapter->num_ports; p_id++) {
> > + port = adapter->ports[p_id];
> > + if (!port)
> > + continue;
> > + if (port->eth_dev) {
> > + rnp_dev_close(port->eth_dev);
> > + /* Just Release Not Primary Port Allocated By PMD */
> > + if (p_id)
> > + rte_eth_dev_release_port(port->eth_dev);
> > + }
> > + }
> > +
> > + return 0;
> > }
> >
> > static int
> > @@ -84,3 +259,14 @@ static struct rte_pci_driver rte_rnp_pmd = {
> > RTE_PMD_REGISTER_PCI(net_rnp, rte_rnp_pmd);
> > RTE_PMD_REGISTER_PCI_TABLE(net_rnp, pci_id_rnp_map);
> > RTE_PMD_REGISTER_KMOD_DEP(net_rnp, "igb_uio | uio_pci_generic |
> > vfio-pci");
> > +
> > +RTE_LOG_REGISTER_SUFFIX(rnp_init_logtype, init, NOTICE);
> > +RTE_LOG_REGISTER_SUFFIX(rnp_drv_logtype, driver, NOTICE);
> > +
>
> Do you really need two different log types? What about reducing it to one?
>
> <...>
>
This can reduce it .
> > diff --git a/drivers/net/rnp/rnp_osdep.h b/drivers/net/rnp/rnp_osdep.h
> > new file mode 100644 index 0000000000..5685dd2404
> > --- /dev/null
> > +++ b/drivers/net/rnp/rnp_osdep.h
> > @@ -0,0 +1,30 @@
> > +#ifndef __RNP_OSDEP_H__
> > +#define __RNP_OSDEP_H__
> > +/* SPDX-License-Identifier: BSD-3-Clause
> > + * Copyright(C) 2023 Mucse IC Design Ltd.
> > + */
> > +#include <stdint.h>
> > +
> > +#include <rte_byteorder.h>
> > +
> > +#define __iomem
> > +#define _RING_(off) ((off) + 0x000000)
> > +#define _DMA_(off) ((off))
> > +#define _GLB_(off) ((off) + 0x000000)
> > +#define _NIC_(off) ((off) + 0x000000)
> > +#define _ETH_(off) ((off))
> > +#define _MAC_(off) ((off))
> > +#define BIT(n) (1UL << (n))
> > +#define BIT64(n) (1ULL << (n))
>
> DPDK has RTE_BIT64 / RTE_BIT32 macros that can be reused.
>
Old DPDK version don't have the macro, this can be removed.
> > +#define BITS_PER_LONG (__SIZEOF_LONG__ * 8)
> > +#define GENMASK(h, l) \
> > + (((~0UL) << (l)) & (~0UL >> (BITS_PER_LONG - 1 - (h))))
> > +
> > +typedef uint8_t u8;
> > +typedef uint16_t u16;
> > +typedef uint32_t u32;
> > +typedef uint64_t u64;
> > +typedef int32_t s32;
> > +typedef int16_t s16;
> > +typedef int8_t s8;
> > +#endif /* __RNP_OSDEP_H__ */
>
^ permalink raw reply [flat|nested] 27+ messages in thread
* [PATCH v6 4/8] net/rnp: add mbx basic api feature
2023-09-01 2:30 [PATCH v6 0/8] [v6]drivers/net Add Support mucse N10 Pmd Driver Wenbo Cao
` (2 preceding siblings ...)
2023-09-01 2:30 ` [PATCH v6 3/8] net/rnp: add device init and uninit Wenbo Cao
@ 2023-09-01 2:30 ` Wenbo Cao
2023-09-05 15:45 ` Ferruh Yigit
2023-09-01 2:30 ` [PATCH v6 5/8] net/rnp add reset code for Chip Init process Wenbo Cao
` (4 subsequent siblings)
8 siblings, 1 reply; 27+ messages in thread
From: Wenbo Cao @ 2023-09-01 2:30 UTC (permalink / raw)
To: Wenbo Cao
Cc: dev, ferruh.yigit, thomas, andrew.rybchenko, yaojun, Stephen Hemminger
mbx base code is for communicate with the firmware
Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
Suggested-by: Stephen Hemminger <stephen@networkplumber.org>
---
drivers/net/rnp/base/rnp_api.c | 23 ++
drivers/net/rnp/base/rnp_api.h | 7 +
drivers/net/rnp/base/rnp_cfg.h | 7 +
drivers/net/rnp/base/rnp_dma_regs.h | 73 ++++
drivers/net/rnp/base/rnp_eth_regs.h | 124 +++++++
drivers/net/rnp/base/rnp_hw.h | 112 +++++-
drivers/net/rnp/meson.build | 1 +
drivers/net/rnp/rnp.h | 35 ++
drivers/net/rnp/rnp_ethdev.c | 70 +++-
drivers/net/rnp/rnp_logs.h | 9 +
drivers/net/rnp/rnp_mbx.c | 522 ++++++++++++++++++++++++++++
drivers/net/rnp/rnp_mbx.h | 139 ++++++++
drivers/net/rnp/rnp_mbx_fw.c | 271 +++++++++++++++
drivers/net/rnp/rnp_mbx_fw.h | 22 ++
14 files changed, 1412 insertions(+), 3 deletions(-)
create mode 100644 drivers/net/rnp/base/rnp_api.c
create mode 100644 drivers/net/rnp/base/rnp_api.h
create mode 100644 drivers/net/rnp/base/rnp_cfg.h
create mode 100644 drivers/net/rnp/base/rnp_dma_regs.h
create mode 100644 drivers/net/rnp/base/rnp_eth_regs.h
create mode 100644 drivers/net/rnp/rnp_mbx.c
create mode 100644 drivers/net/rnp/rnp_mbx.h
create mode 100644 drivers/net/rnp/rnp_mbx_fw.c
create mode 100644 drivers/net/rnp/rnp_mbx_fw.h
diff --git a/drivers/net/rnp/base/rnp_api.c b/drivers/net/rnp/base/rnp_api.c
new file mode 100644
index 0000000000..550da6217d
--- /dev/null
+++ b/drivers/net/rnp/base/rnp_api.c
@@ -0,0 +1,23 @@
+#include "rnp.h"
+#include "rnp_api.h"
+
+int
+rnp_init_hw(struct rte_eth_dev *dev)
+{
+ const struct rnp_mac_api *ops = RNP_DEV_TO_MAC_OPS(dev);
+ struct rnp_hw *hw = RNP_DEV_TO_HW(dev);
+
+ if (ops->init_hw)
+ return ops->init_hw(hw);
+ return -EOPNOTSUPP;
+}
+
+int
+rnp_reset_hw(struct rte_eth_dev *dev, struct rnp_hw *hw)
+{
+ const struct rnp_mac_api *ops = RNP_DEV_TO_MAC_OPS(dev);
+
+ if (ops->reset_hw)
+ return ops->reset_hw(hw);
+ return -EOPNOTSUPP;
+}
diff --git a/drivers/net/rnp/base/rnp_api.h b/drivers/net/rnp/base/rnp_api.h
new file mode 100644
index 0000000000..df574dab77
--- /dev/null
+++ b/drivers/net/rnp/base/rnp_api.h
@@ -0,0 +1,7 @@
+#ifndef __RNP_API_H__
+#define __RNP_API_H__
+int
+rnp_init_hw(struct rte_eth_dev *dev);
+int
+rnp_reset_hw(struct rte_eth_dev *dev, struct rnp_hw *hw);
+#endif /* __RNP_API_H__ */
diff --git a/drivers/net/rnp/base/rnp_cfg.h b/drivers/net/rnp/base/rnp_cfg.h
new file mode 100644
index 0000000000..90f25268ad
--- /dev/null
+++ b/drivers/net/rnp/base/rnp_cfg.h
@@ -0,0 +1,7 @@
+#ifndef __RNP_CFG_H__
+#define __RNP_CFG_H__
+#include "rnp_osdep.h"
+
+#define RNP_NIC_RESET _NIC_(0x0010)
+#define RNP_TX_QINQ_WORKAROUND _NIC_(0x801c)
+#endif /* __RNP_CFG_H__ */
diff --git a/drivers/net/rnp/base/rnp_dma_regs.h b/drivers/net/rnp/base/rnp_dma_regs.h
new file mode 100644
index 0000000000..bfe87e534d
--- /dev/null
+++ b/drivers/net/rnp/base/rnp_dma_regs.h
@@ -0,0 +1,73 @@
+#ifndef __RNP_REGS_H__
+#define __RNP_REGS_H__
+
+#include "rnp_osdep.h"
+
+/* mac address offset */
+#define RNP_DMA_CTRL (0x4)
+#define RNP_VEB_BYPASS_EN BIT(4)
+#define RNP_DMA_MEM_CFG_LE (0 << 5)
+#define TSNR10_DMA_MEM_CFG_BE (1 << 5)
+#define RNP_DMA_SCATTER_MEM_SHIFT (16)
+
+#define RNP_FIRMWARE_SYNC (0xc)
+#define RNP_FIRMWARE_SYNC_MASK GENMASK(31, 16)
+#define RNP_FIRMWARE_SYNC_MAGIC (0xa5a40000)
+#define RNP_DRIVER_REMOVE (0x5a000000)
+/* 1BIT <-> 16 bytes Dma Addr Size*/
+#define RNP_DMA_SCATTER_MEM_MASK GENMASK(31, 16)
+#define RNP_DMA_TX_MAP_MODE_SHIFT (12)
+#define RNP_DMA_TX_MAP_MODE_MASK GENMASK(15, 12)
+#define RNP_DMA_RX_MEM_PAD_EN BIT(8)
+/* === queue register ===== */
+/* enable */
+#define RNP_DMA_RXQ_START(qid) _RING_(0x0010 + 0x100 * (qid))
+#define RNP_DMA_RXQ_READY(qid) _RING_(0x0014 + 0x100 * (qid))
+#define RNP_DMA_TXQ_START(qid) _RING_(0x0018 + 0x100 * (qid))
+#define RNP_DMA_TXQ_READY(qid) _RING_(0x001c + 0x100 * (qid))
+
+#define RNP_DMA_INT_STAT(qid) _RING_(0x0020 + 0x100 * (qid))
+#define RNP_DMA_INT_MASK(qid) _RING_(0x0024 + 0x100 * (qid))
+#define RNP_TX_INT_MASK BIT(1)
+#define RNP_RX_INT_MASK BIT(0)
+#define RNP_DMA_INT_CLER(qid) _RING_(0x0028 + 0x100 * (qid))
+
+/* rx-queue */
+#define RNP_DMA_RXQ_BASE_ADDR_HI(qid) _RING_(0x0030 + 0x100 * (qid))
+#define RNP_DMA_RXQ_BASE_ADDR_LO(qid) _RING_(0x0034 + 0x100 * (qid))
+#define RNP_DMA_RXQ_LEN(qid) _RING_(0x0038 + 0x100 * (qid))
+#define RNP_DMA_RXQ_HEAD(qid) _RING_(0x003c + 0x100 * (qid))
+#define RNP_DMA_RXQ_TAIL(qid) _RING_(0x0040 + 0x100 * (qid))
+#define RNP_DMA_RXQ_DESC_FETCH_CTRL(qid) _RING_(0x0044 + 0x100 * (qid))
+#define RNP_DMA_RXQ_INT_DELAY_TIMER(qid) _RING_(0x0048 + 0x100 * (qid))
+#define RNP_DMA_RXQ_INT_DELAY_PKTCNT(qidx) _RING_(0x004c + 0x100 * (qid))
+#define RNP_DMA_RXQ_RX_PRI_LVL(qid) _RING_(0x0050 + 0x100 * (qid))
+#define RNP_DMA_RXQ_DROP_TIMEOUT_TH(qid) _RING_(0x0054 + 0x100 * (qid))
+/* tx-queue */
+#define RNP_DMA_TXQ_BASE_ADDR_HI(qid) _RING_(0x0060 + 0x100 * (qid))
+#define RNP_DMA_TXQ_BASE_ADDR_LO(qid) _RING_(0x0064 + 0x100 * (qid))
+#define RNP_DMA_TXQ_LEN(qid) _RING_(0x0068 + 0x100 * (qid))
+#define RNP_DMA_TXQ_HEAD(qid) _RING_(0x006c + 0x100 * (qid))
+#define RNP_DMA_TXQ_TAIL(qid) _RING_(0x0070 + 0x100 * (qid))
+#define RNP_DMA_TXQ_DESC_FETCH_CTRL(qid) _RING_(0x0074 + 0x100 * (qid))
+#define RNP_DMA_TXQ_INT_DELAY_TIMER(qid) _RING_(0x0078 + 0x100 * (qid))
+#define RNP_DMA_TXQ_INT_DELAY_PKTCNT(qid) _RING_(0x007c + 0x100 * (qid))
+
+#define RNP_DMA_TXQ_PRI_LVL(qid) _RING_(0x0080 + 0x100 * (qid))
+#define RNP_DMA_TXQ_RATE_CTRL_TH(qid) _RING_(0x0084 + 0x100 * (qid))
+#define RNP_DMA_TXQ_RATE_CTRL_TM(qid) _RING_(0x0088 + 0x100 * (qid))
+
+/* VEB Table Register */
+#define RNP_VBE_MAC_LO(port, nr) _RING_(0x00a0 + (4 * (port)) + \
+ (0x100 * (nr)))
+#define RNP_VBE_MAC_HI(port, nr) _RING_(0x00b0 + (4 * (port)) + \
+ (0x100 * (nr)))
+#define RNP_VEB_VID_CFG(port, nr) _RING_(0x00c0 + (4 * (port)) + \
+ (0x100 * (nr)))
+#define RNP_VEB_VF_RING(port, nr) _RING_(0x00d0 + (4 * (port)) + \
+ (0x100 * (nr)))
+#define RNP_MAX_VEB_TB (64)
+#define RNP_VEB_RING_CFG_OFFSET (8)
+#define RNP_VEB_SWITCH_VF_EN BIT(7)
+#define MAX_VEB_TABLES_NUM (4)
+#endif /* RNP_DMA_REGS_H_ */
diff --git a/drivers/net/rnp/base/rnp_eth_regs.h b/drivers/net/rnp/base/rnp_eth_regs.h
new file mode 100644
index 0000000000..88e8e1e552
--- /dev/null
+++ b/drivers/net/rnp/base/rnp_eth_regs.h
@@ -0,0 +1,124 @@
+#ifndef _RNP_ETH_REGS_H_
+#define _RNP_ETH_REGS_H_
+
+#include "rnp_osdep.h"
+
+/* PTP 1588 TM Offload */
+#define RNP_ETH_PTP_TX_STATUS(n) _ETH_(0x0400 + ((n) * 0x14))
+#define RNP_ETH_PTP_TX_HTIMES(n) _ETH_(0x0404 + ((n) * 0x14))
+#define RNP_ETH_PTP_TX_LTIMES(n) _ETH_(0x0408 + ((n) * 0x14))
+#define RNP_ETH_PTP_TX_TS_ST(n) _ETH_(0x040c + ((n) * 0x14))
+#define RNP_ETH_PTP_TX_CLEAR(n) _ETH_(0x0410 + ((n) * 0x14))
+
+#define RNP_ETH_ENGINE_BYPASS _ETH_(0x8000)
+#define RNP_EN_TUNNEL_VXLAN_PARSE _ETH_(0x8004)
+#define RNP_ETH_MAC_LOOPBACK _ETH_(0x8008)
+#define RNP_ETH_FIFO_CTRL _ETH_(0x800c)
+#define RNP_ETH_FOUR_FIFO BIT(0)
+#define RNP_ETH_TWO_FIFO BIT(1)
+#define RNP_ETH_ONE_FIFO BIT(2)
+#define RNP_FIFO_CFG_EN (0x1221)
+#define RNP_ETH_VXLAN_PORT_CTRL _ETH_(0x8010)
+#define RNP_ETH_VXLAN_DEF_PORT (4789)
+#define RNP_HOST_FILTER_EN _ETH_(0x801c)
+#define RNP_HW_SCTP_CKSUM_CTRL _ETH_(0x8038)
+#define RNP_HW_CHECK_ERR_CTRL _ETH_(0x8060)
+#define RNP_HW_ERR_HDR_LEN BIT(0)
+#define RNP_HW_ERR_PKTLEN BIT(1)
+#define RNP_HW_L3_CKSUM_ERR BIT(2)
+#define RNP_HW_L4_CKSUM_ERR BIT(3)
+#define RNP_HW_SCTP_CKSUM_ERR BIT(4)
+#define RNP_HW_INNER_L3_CKSUM_ERR BIT(5)
+#define RNP_HW_INNER_L4_CKSUM_ERR BIT(6)
+#define RNP_HW_CKSUM_ERR_MASK GENMASK(6, 2)
+#define RNP_HW_CHECK_ERR_MASK GENMASK(6, 0)
+#define RNP_HW_ERR_RX_ALL_MASK GENMASK(1, 0)
+
+#define RNP_REDIR_CTRL _ETH_(0x8030)
+#define RNP_VLAN_Q_STRIP_CTRL(n) _ETH_(0x8040 + 0x4 * ((n) / 32))
+/* This Just VLAN Master Switch */
+#define RNP_VLAN_TUNNEL_STRIP_EN _ETH_(0x8050)
+#define RNP_VLAN_TUNNEL_STRIP_MODE _ETH_(0x8054)
+#define RNP_VLAN_TUNNEL_STRIP_OUTER (0)
+#define RNP_VLAN_TUNNEL_STRIP_INNER (1)
+#define RNP_RSS_INNER_CTRL _ETH_(0x805c)
+#define RNP_INNER_RSS_EN (1)
+
+#define RNP_ETH_DEFAULT_RX_RING _ETH_(0x806c)
+#define RNP_RX_FC_HI_WATER(n) _ETH_(0x80c0 + ((n) * 0x8))
+#define RNP_RX_FC_LO_WATER(n) _ETH_(0x80c4 + ((n) * 0x8))
+
+#define RNP_RX_FIFO_FULL_THRETH(n) _ETH_(0x8070 + ((n) * 0x8))
+#define RNP_RX_WORKAROUND_VAL _ETH_(0x7ff)
+#define RNP_RX_DEFAULT_VAL _ETH_(0x270)
+
+#define RNP_MIN_FRAME_CTRL _ETH_(0x80f0)
+#define RNP_MAX_FRAME_CTRL _ETH_(0x80f4)
+
+#define RNP_RX_FC_ENABLE _ETH_(0x8520)
+#define RNP_RING_FC_EN(n) _ETH_(0x8524 + 0x4 * ((n) / 32))
+#define RNP_RING_FC_THRESH(n) _ETH_(0x8a00 + 0x4 * (n))
+
+/* Mac Host Filter */
+#define RNP_MAC_FCTRL _ETH_(0x9110)
+#define RNP_MAC_FCTRL_MPE BIT(8) /* Multicast Promiscuous En */
+#define RNP_MAC_FCTRL_UPE BIT(9) /* Unicast Promiscuous En */
+#define RNP_MAC_FCTRL_BAM BIT(10) /* Broadcast Accept Mode */
+#define RNP_MAC_FCTRL_BYPASS (RNP_MAC_FCTRL_MPE | \
+ RNP_MAC_FCTRL_UPE | \
+ RNP_MAC_FCTRL_BAM)
+/* MC UC Mac Hash Filter Ctrl */
+#define RNP_MAC_MCSTCTRL _ETH_(0x9114)
+#define RNP_MAC_HASH_MASK GENMASK(11, 0)
+#define RNP_MAC_MULTICASE_TBL_EN BIT(2)
+#define RNP_MAC_UNICASE_TBL_EN BIT(3)
+#define RNP_UC_HASH_TB(n) _ETH_(0xA800 + ((n) * 0x4))
+#define RNP_MC_HASH_TB(n) _ETH_(0xAC00 + ((n) * 0x4))
+
+#define RNP_VLAN_FILTER_CTRL _ETH_(0x9118)
+#define RNP_L2TYPE_FILTER_CTRL (RNP_VLAN_FILTER_CTRL)
+#define RNP_L2TYPE_FILTER_EN BIT(31)
+#define RNP_VLAN_FILTER_EN BIT(30)
+
+#define RNP_FC_PAUSE_FWD_ACT _ETH_(0x9280)
+#define RNP_FC_PAUSE_DROP BIT(31)
+#define RNP_FC_PAUSE_PASS (0)
+#define RNP_FC_PAUSE_TYPE _ETH_(0x9284)
+#define RNP_FC_PAUSE_POLICY_EN BIT(31)
+#define RNP_PAUSE_TYPE _ETH_(0x8808)
+
+#define RNP_INPUT_USE_CTRL _ETH_(0x91d0)
+#define RNP_INPUT_VALID_MASK (0xf)
+#define RNP_INPUT_POLICY(n) _ETH_(0x91e0 + ((n) * 0x4))
+/* RSS */
+#define RNP_RSS_MRQC_ADDR _ETH_(0x92a0)
+#define RNP_SRIOV_CTRL RNP_RSS_MRQC_ADDR
+#define RNP_SRIOV_ENABLE BIT(3)
+
+#define RNP_RSS_REDIR_TB(mac, idx) _ETH_(0xe000 + \
+ ((mac) * 0x200) + ((idx) * 0x4))
+#define RNP_RSS_KEY_TABLE(idx) _ETH_(0x92d0 + ((idx) * 0x4))
+/*=======================================================================
+ *HOST_MAC_ADDRESS_FILTER
+ *=======================================================================
+ */
+#define RNP_RAL_BASE_ADDR(vf_id) _ETH_(0xA000 + 0x04 * (vf_id))
+#define RNP_RAH_BASE_ADDR(vf_id) _ETH_(0xA400 + 0x04 * (vf_id))
+#define RNP_MAC_FILTER_EN BIT(31)
+
+/* ETH Statistic */
+#define RNP_ETH_RXTRANS_DROP(p_id) _ETH_((0x8904) + ((p_id) * (0x40)))
+#define RNP_ETH_RXTRANS_CAT_ERR(p_id) _ETH_((0x8928) + ((p_id) * (0x40)))
+#define RNP_ETH_TXTM_DROP _ETH_(0X0470)
+
+#define RNP_VFTA_BASE_ADDR _ETH_(0xB000)
+#define RNP_VFTA_HASH_TABLE(id) (RNP_VFTA_BASE_ADDR + 0x4 * (id))
+#define RNP_ETYPE_BASE_ADDR _ETH_(0xB300)
+#define RNP_MPSAR_BASE_ADDR(vf_id) _ETH_(0xB400 + 0x04 * (vf_id))
+#define RNP_PFVLVF_BASE_ADDR _ETH_(0xB600)
+#define RNP_PFVLVFB_BASE_ADDR _ETH_(0xB700)
+#define RNP_TUNNEL_PFVLVF_BASE_ADDR _ETH_(0xB800)
+#define RNP_TUNNEL_PFVLVFB_BASE_ADDR _ETH_(0xB900)
+
+#define RNP_TC_PORT_MAP_TB(port) _ETH_(0xe840 + 0x04 * (port))
+#endif /* RNP_ETH_REGS_H_ */
diff --git a/drivers/net/rnp/base/rnp_hw.h b/drivers/net/rnp/base/rnp_hw.h
index d80d23f4b4..1db966cf21 100644
--- a/drivers/net/rnp/base/rnp_hw.h
+++ b/drivers/net/rnp/base/rnp_hw.h
@@ -4,16 +4,126 @@
#ifndef __RNP_HW_H__
#define __RNP_HW_H__
+#include <rte_io.h>
+#include <ethdev_driver.h>
+
+#include "rnp_osdep.h"
+
+static inline unsigned int rnp_rd_reg(volatile void *addr)
+{
+ unsigned int v = rte_read32(addr);
+
+ return v;
+}
+
+static inline void rnp_wr_reg(volatile void *reg, int val)
+{
+ rte_write32_relaxed((val), (reg));
+}
+
+#define mbx_rd32(_hw, _off) \
+ rnp_rd_reg((uint8_t *)((_hw)->iobar4) + (_off))
+#define mbx_wr32(_hw, _off, _val) \
+ rnp_wr_reg((uint8_t *)((_hw)->iobar4) + (_off), (_val))
+#define rnp_io_rd(_base, _off) \
+ rnp_rd_reg((uint8_t *)(_base) + (_off))
+#define rnp_io_wr(_base, _off, _val) \
+ rnp_wr_reg((uint8_t *)(_base) + (_off), (_val))
+
+struct rnp_hw;
+/* Mbx Operate info */
+enum MBX_ID {
+ MBX_PF = 0,
+ MBX_VF,
+ MBX_CM3CPU,
+ MBX_FW = MBX_CM3CPU,
+ MBX_VFCNT
+};
+struct rnp_mbx_api {
+ void (*init_mbx)(struct rnp_hw *hw);
+ int32_t (*read)(struct rnp_hw *hw,
+ uint32_t *msg,
+ uint16_t size,
+ enum MBX_ID);
+ int32_t (*write)(struct rnp_hw *hw,
+ uint32_t *msg,
+ uint16_t size,
+ enum MBX_ID);
+ int32_t (*read_posted)(struct rte_eth_dev *dev,
+ uint32_t *msg,
+ uint16_t size,
+ enum MBX_ID);
+ int32_t (*write_posted)(struct rte_eth_dev *dev,
+ uint32_t *msg,
+ uint16_t size,
+ enum MBX_ID);
+ int32_t (*check_for_msg)(struct rnp_hw *hw, enum MBX_ID);
+ int32_t (*check_for_ack)(struct rnp_hw *hw, enum MBX_ID);
+ int32_t (*check_for_rst)(struct rnp_hw *hw, enum MBX_ID);
+};
+
+struct rnp_mbx_stats {
+ u32 msgs_tx;
+ u32 msgs_rx;
+
+ u32 acks;
+ u32 reqs;
+ u32 rsts;
+};
+
+struct rnp_mbx_info {
+ struct rnp_mbx_api ops;
+ uint32_t usec_delay; /* retry interval delay time */
+ uint32_t timeout; /* retry ops timeout limit */
+ uint16_t size; /* data buffer size*/
+ uint16_t vf_num; /* Virtual Function num */
+ uint16_t pf_num; /* Physical Function num */
+ uint16_t sriov_st; /* Sriov state */
+ bool irq_enabled;
+ union {
+ struct {
+ unsigned short pf_req;
+ unsigned short pf_ack;
+ };
+ struct {
+ unsigned short cpu_req;
+ unsigned short cpu_ack;
+ };
+ };
+ unsigned short vf_req[64];
+ unsigned short vf_ack[64];
+
+ struct rnp_mbx_stats stats;
+
+ rte_atomic16_t state;
+} __rte_cache_aligned;
+
struct rnp_eth_adapter;
+#define RNP_MAX_HW_PORT_PERR_PF (4)
struct rnp_hw {
struct rnp_eth_adapter *back;
void *iobar0;
uint32_t iobar0_len;
void *iobar4;
uint32_t iobar4_len;
+ void *link_sync;
+ void *dma_base;
+ void *eth_base;
+ void *veb_base;
+ void *mac_base[RNP_MAX_HW_PORT_PERR_PF];
+ void *msix_base;
+ /* === dma == */
+ void *dma_axi_en;
+ void *dma_axi_st;
uint16_t device_id;
uint16_t vendor_id;
-} __rte_cache_aligned;
+ uint16_t function;
+ uint16_t pf_vf_num;
+ uint16_t max_vfs;
+ void *cookie_pool;
+ char cookie_p_name[RTE_MEMZONE_NAMESIZE];
+ struct rnp_mbx_info mbx;
+} __rte_cache_aligned;
#endif /* __RNP_H__*/
diff --git a/drivers/net/rnp/meson.build b/drivers/net/rnp/meson.build
index f85d597e68..60bba486fc 100644
--- a/drivers/net/rnp/meson.build
+++ b/drivers/net/rnp/meson.build
@@ -8,5 +8,6 @@ endif
sources = files(
'rnp_ethdev.c',
+ 'rnp_mbx.c',
)
includes += include_directories('base')
diff --git a/drivers/net/rnp/rnp.h b/drivers/net/rnp/rnp.h
index cab1b8e85d..6e12885877 100644
--- a/drivers/net/rnp/rnp.h
+++ b/drivers/net/rnp/rnp.h
@@ -3,6 +3,7 @@
*/
#ifndef __RNP_H__
#define __RNP_H__
+#include <rte_log.h>
#include "base/rnp_hw.h"
@@ -14,14 +15,17 @@
struct rnp_eth_port {
struct rnp_eth_adapter *adapt;
+ struct rnp_hw *hw;
struct rte_eth_dev *eth_dev;
} __rte_cache_aligned;
struct rnp_share_ops {
+ const struct rnp_mbx_api *mbx_api;
} __rte_cache_aligned;
struct rnp_eth_adapter {
struct rnp_hw hw;
+ uint16_t max_vfs;
struct rte_pci_device *pdev;
struct rte_eth_dev *eth_dev; /* primary eth_dev */
struct rnp_eth_port *ports[RNP_MAX_PORT_OF_PF];
@@ -34,5 +38,36 @@ struct rnp_eth_adapter {
(((struct rnp_eth_port *)((eth_dev)->data->dev_private)))
#define RNP_DEV_TO_ADAPTER(eth_dev) \
((struct rnp_eth_adapter *)(RNP_DEV_TO_PORT(eth_dev)->adapt))
+#define RNP_DEV_TO_HW(eth_dev) \
+ (&((struct rnp_eth_adapter *)(RNP_DEV_TO_PORT((eth_dev))->adapt))->hw)
+#define RNP_DEV_PP_PRIV_TO_MBX_OPS(dev) \
+ (((struct rnp_share_ops *)(dev)->process_private)->mbx_api)
+#define RNP_DEV_TO_MBX_OPS(dev) RNP_DEV_PP_PRIV_TO_MBX_OPS(dev)
+static inline void rnp_reg_offset_init(struct rnp_hw *hw)
+{
+ uint16_t i;
+
+ if (hw->device_id == RNP_DEV_ID_N10G && hw->mbx.pf_num) {
+ hw->iobar4 = (void *)((uint8_t *)hw->iobar4 + 0x100000);
+ hw->msix_base = (void *)((uint8_t *)hw->iobar4 + 0xa4000);
+ hw->msix_base = (void *)((uint8_t *)hw->msix_base + 0x200);
+ } else {
+ hw->msix_base = (void *)((uint8_t *)hw->iobar4 + 0xa4000);
+ }
+ /* === dma status/config====== */
+ hw->link_sync = (void *)((uint8_t *)hw->iobar4 + 0x000c);
+ hw->dma_axi_en = (void *)((uint8_t *)hw->iobar4 + 0x0010);
+ hw->dma_axi_st = (void *)((uint8_t *)hw->iobar4 + 0x0014);
+
+ if (hw->mbx.pf_num)
+ hw->msix_base = (void *)((uint8_t *)0x200);
+ /* === queue registers === */
+ hw->dma_base = (void *)((uint8_t *)hw->iobar4 + 0x08000);
+ hw->veb_base = (void *)((uint8_t *)hw->iobar4 + 0x0);
+ hw->eth_base = (void *)((uint8_t *)hw->iobar4 + 0x10000);
+ /* mac */
+ for (i = 0; i < RNP_MAX_HW_PORT_PERR_PF; i++)
+ hw->mac_base[i] = (void *)((uint8_t *)hw->iobar4 + 0x60000 + 0x10000 * i);
+}
#endif /* __RNP_H__ */
diff --git a/drivers/net/rnp/rnp_ethdev.c b/drivers/net/rnp/rnp_ethdev.c
index ae737643a7..a2dc27548a 100644
--- a/drivers/net/rnp/rnp_ethdev.c
+++ b/drivers/net/rnp/rnp_ethdev.c
@@ -8,6 +8,7 @@
#include <ethdev_driver.h>
#include "rnp.h"
+#include "rnp_mbx.h"
#include "rnp_logs.h"
static int
@@ -89,6 +90,58 @@ rnp_alloc_eth_port(struct rte_pci_device *primary_pci, char *name)
return NULL;
}
+static void rnp_get_nic_attr(struct rnp_eth_adapter *adapter)
+{
+ RTE_SET_USED(adapter);
+}
+
+static int
+rnp_process_resource_init(struct rte_eth_dev *eth_dev)
+{
+ struct rnp_share_ops *share_priv;
+
+ /* allocate process_private memory this must can't
+ * belone to the dpdk mem resource manager
+ * such as from rte_malloc or rte_dma_zone..
+ */
+ /* use the process_prive point to resolve secondary process
+ * use point-func. This point is per process will be safe to cover.
+ * This will cause secondary process core-dump because of IPC
+ * Secondary will call primary process point func virt-address
+ * secondary process don't alloc user/pmd to alloc or free
+ * the memory of dpdk-mem resource it will cause hugepage
+ * mem exception
+ * be careful for secondary Process to use the share-mem of
+ * point correlation
+ */
+ share_priv = calloc(1, sizeof(*share_priv));
+ if (!share_priv) {
+ PMD_DRV_LOG(ERR, "calloc share_priv failed");
+ return -ENOMEM;
+ }
+ memset(share_priv, 0, sizeof(*share_priv));
+ eth_dev->process_private = share_priv;
+
+ return 0;
+}
+
+static void
+rnp_common_ops_init(struct rnp_eth_adapter *adapter)
+{
+ struct rnp_share_ops *share_priv;
+
+ share_priv = adapter->share_priv;
+ share_priv->mbx_api = &rnp_mbx_pf_ops;
+}
+
+static int
+rnp_special_ops_init(struct rte_eth_dev *eth_dev)
+{
+ RTE_SET_USED(eth_dev);
+
+ return 0;
+}
+
static int
rnp_eth_dev_init(struct rte_eth_dev *dev)
{
@@ -124,6 +177,20 @@ rnp_eth_dev_init(struct rte_eth_dev *dev)
hw->device_id = pci_dev->id.device_id;
hw->vendor_id = pci_dev->id.vendor_id;
hw->device_id = pci_dev->id.device_id;
+ adapter->max_vfs = pci_dev->max_vfs;
+ ret = rnp_process_resource_init(dev);
+ if (ret) {
+ PMD_DRV_LOG(ERR, "share prive resource init failed");
+ return ret;
+ }
+ adapter->share_priv = dev->process_private;
+ rnp_common_ops_init(adapter);
+ rnp_get_nic_attr(adapter);
+ /* We need Use Device Id To Change The Resource Mode */
+ rnp_special_ops_init(dev);
+ port->adapt = adapter;
+ port->hw = hw;
+ rnp_init_mbx_ops_pf(hw);
for (p_id = 0; p_id < adapter->num_ports; p_id++) {
/* port 0 resource has been allocated When Probe */
if (!p_id) {
@@ -158,11 +225,10 @@ rnp_eth_dev_init(struct rte_eth_dev *dev)
continue;
if (port->eth_dev) {
rnp_dev_close(port->eth_dev);
- rte_eth_dev_release_port(port->eth_dev);
if (port->eth_dev->process_private)
free(port->eth_dev->process_private);
+ rte_eth_dev_release_port(port->eth_dev);
}
- rte_free(port);
}
rte_free(adapter);
diff --git a/drivers/net/rnp/rnp_logs.h b/drivers/net/rnp/rnp_logs.h
index 1b3ee33745..f1648aabb5 100644
--- a/drivers/net/rnp/rnp_logs.h
+++ b/drivers/net/rnp/rnp_logs.h
@@ -13,6 +13,15 @@ extern int rnp_drv_logtype;
#define RNP_PMD_DRV_LOG(level, fmt, args...) \
rte_log(RTE_LOG_##level, rnp_drv_logtype, \
"%s() " fmt, __func__, ##args)
+#define PMD_DRV_LOG_RAW(level, fmt, args...) \
+ rte_log(RTE_LOG_ ## level, rnp_drv_logtype, "%s(): " fmt, \
+ __func__, ## args)
+#define PMD_DRV_LOG(level, fmt, args...) \
+ PMD_DRV_LOG_RAW(level, fmt "\n", ## args)
+
+#define RNP_PMD_LOG(level, fmt, args...) \
+ rte_log(RTE_LOG_##level, rnp_drv_logtype, \
+ "rnp_net: (%d) " fmt, __LINE__, ##args)
#ifdef RTE_LIBRTE_RNP_DEBUG_RX
extern int rnp_rx_logtype;
#define RNP_PMD_RX_LOG(level, fmt, args...) \
diff --git a/drivers/net/rnp/rnp_mbx.c b/drivers/net/rnp/rnp_mbx.c
new file mode 100644
index 0000000000..29aedc554b
--- /dev/null
+++ b/drivers/net/rnp/rnp_mbx.c
@@ -0,0 +1,522 @@
+#include <rte_cycles.h>
+#include <rte_log.h>
+
+#include "rnp.h"
+#include "rnp_hw.h"
+#include "rnp_mbx.h"
+#include "rnp_mbx_fw.h"
+#include "rnp_logs.h"
+
+#define RNP_MAX_VF_FUNCTIONS (64)
+/* == VEC == */
+#define VF2PF_MBOX_VEC(VF) (0xa5100 + 4 * (VF))
+#define CPU2PF_MBOX_VEC (0xa5300)
+
+/* == PF <--> VF mailbox ==== */
+#define SHARE_MEM_BYTES (64) /* 64bytes */
+/* for PF1 rtl will remap 6000 to 0xb000 */
+#define PF_VF_SHM(vf) ((0xa6000) + (64 * (vf)))
+#define PF2VF_COUNTER(vf) (PF_VF_SHM(vf) + 0)
+#define VF2PF_COUNTER(vf) (PF_VF_SHM(vf) + 4)
+#define PF_VF_SHM_DATA(vf) (PF_VF_SHM(vf) + 8)
+#define PF2VF_MBOX_CTRL(vf) ((0xa7100) + (4 * (vf)))
+#define PF_VF_MBOX_MASK_LO ((0xa7200))
+#define PF_VF_MBOX_MASK_HI ((0xa7300))
+
+/* === CPU <--> PF === */
+#define CPU_PF_SHM (0xaa000)
+#define CPU2PF_COUNTER (CPU_PF_SHM + 0)
+#define PF2CPU_COUNTER (CPU_PF_SHM + 4)
+#define CPU_PF_SHM_DATA (CPU_PF_SHM + 8)
+#define PF2CPU_MBOX_CTRL (0xaa100)
+#define CPU_PF_MBOX_MASK (0xaa300)
+
+/* === CPU <--> VF === */
+#define CPU_VF_SHM(vf) (0xa8000 + (64 * (vf)))
+#define CPU2VF_COUNTER(vf) (CPU_VF_SHM(vf) + 0)
+#define VF2CPU_COUNTER(vf) (CPU_VF_SHM(vf) + 4)
+#define CPU_VF_SHM_DATA(vf) (CPU_VF_SHM(vf) + 8)
+#define VF2CPU_MBOX_CTRL(vf) (0xa9000 + 64 * (vf))
+#define CPU_VF_MBOX_MASK_LO(vf) (0xa9200 + 64 * (vf))
+#define CPU_VF_MBOX_MASK_HI(vf) (0xa9300 + 64 * (vf))
+
+#define MBOX_CTRL_REQ (1 << 0) /* WO */
+/* VF:WR, PF:RO */
+#define MBOX_CTRL_PF_HOLD_SHM (1 << 3) /* VF:RO, PF:WR */
+
+#define MBOX_IRQ_EN (0)
+#define MBOX_IRQ_DISABLE (1)
+
+/****************************PF MBX OPS************************************/
+static inline u16 rnp_mbx_get_req(struct rnp_hw *hw, int reg)
+{
+ rte_mb();
+ return mbx_rd32(hw, reg) & 0xffff;
+}
+
+static inline u16 rnp_mbx_get_ack(struct rnp_hw *hw, int reg)
+{
+ rte_mb();
+ return (mbx_rd32(hw, reg) >> 16) & 0xffff;
+}
+
+static inline void rnp_mbx_inc_pf_req(struct rnp_hw *hw, enum MBX_ID mbx_id)
+{
+ int reg = (mbx_id == MBX_CM3CPU) ?
+ PF2CPU_COUNTER : PF2VF_COUNTER(mbx_id);
+ u32 v = mbx_rd32(hw, reg);
+ u16 req;
+
+ req = (v & 0xffff);
+ req++;
+ v &= ~(0x0000ffff);
+ v |= req;
+
+ rte_mb();
+ mbx_wr32(hw, reg, v);
+
+ /* update stats */
+ /* hw->mbx.stats.msgs_tx++; */
+}
+
+static inline void rnp_mbx_inc_pf_ack(struct rnp_hw *hw, enum MBX_ID mbx_id)
+{
+ int reg = (mbx_id == MBX_CM3CPU) ?
+ PF2CPU_COUNTER : PF2VF_COUNTER(mbx_id);
+ u32 v = mbx_rd32(hw, reg);
+ u16 ack;
+
+ ack = (v >> 16) & 0xffff;
+ ack++;
+ v &= ~(0xffff0000);
+ v |= (ack << 16);
+
+ rte_mb();
+ mbx_wr32(hw, reg, v);
+
+ /* update stats */
+ /* hw->mbx.stats.msgs_rx++; */
+}
+
+/**
+ * rnp_poll_for_msg - Wait for message notification
+ * @hw: pointer to the HW structure
+ * @mbx_id: id of mailbox to write
+ *
+ * returns SUCCESS if it successfully received a message notification
+ **/
+static int32_t rnp_poll_for_msg(struct rte_eth_dev *dev, enum MBX_ID mbx_id)
+{
+ const struct rnp_mbx_api *ops = RNP_DEV_TO_MBX_OPS(dev);
+ struct rnp_hw *hw = RNP_DEV_TO_HW(dev);
+ struct rnp_mbx_info *mbx = &hw->mbx;
+ int countdown = mbx->timeout;
+
+ if (!countdown || !ops->check_for_msg)
+ goto out;
+
+ while (countdown && ops->check_for_msg(hw, mbx_id)) {
+ countdown--;
+ if (!countdown)
+ break;
+ rte_delay_us_block(mbx->usec_delay);
+ }
+
+out:
+ return countdown ? 0 : -ETIMEDOUT;
+}
+
+/**
+ * rnp_poll_for_ack - Wait for message acknowledgment
+ * @hw: pointer to the HW structure
+ * @mbx_id: id of mailbox to write
+ *
+ * returns SUCCESS if it successfully received a message acknowledgment
+ **/
+static int32_t rnp_poll_for_ack(struct rte_eth_dev *dev, enum MBX_ID mbx_id)
+{
+ const struct rnp_mbx_api *ops = RNP_DEV_TO_MBX_OPS(dev);
+ struct rnp_hw *hw = RNP_DEV_TO_HW(dev);
+ struct rnp_mbx_info *mbx = &hw->mbx;
+ int countdown = mbx->timeout;
+
+ if (!countdown || !ops->check_for_ack)
+ goto out;
+
+ while (countdown && ops->check_for_ack(hw, mbx_id)) {
+ countdown--;
+ if (!countdown)
+ break;
+ rte_delay_us_block(mbx->usec_delay);
+ }
+
+out:
+ return countdown ? 0 : -ETIMEDOUT;
+}
+
+/**
+ * rnp_read_posted_mbx - Wait for message notification and receive message
+ * @hw: pointer to the HW structure
+ * @msg: The message buffer
+ * @size: Length of buffer
+ * @mbx_id: id of mailbox to write
+ *
+ * returns SUCCESS if it successfully received a message notification and
+ * copied it into the receive buffer.
+ **/
+static int32_t
+rnp_read_posted_mbx_pf(struct rte_eth_dev *dev, u32 *msg, u16 size,
+ enum MBX_ID mbx_id)
+{
+ const struct rnp_mbx_api *ops = RNP_DEV_TO_MBX_OPS(dev);
+ struct rnp_hw *hw = RNP_DEV_TO_HW(dev);
+ struct rnp_mbx_info *mbx = &hw->mbx;
+ int countdown = mbx->timeout;
+ int32_t ret_val = -ETIMEDOUT;
+
+ if (!ops->read || !countdown)
+ return -EOPNOTSUPP;
+
+ ret_val = rnp_poll_for_msg(dev, mbx_id);
+
+ /* if ack received read message, otherwise we timed out */
+ if (!ret_val)
+ return ops->read(hw, msg, size, mbx_id);
+ return ret_val;
+}
+
+/**
+ * rnp_write_posted_mbx - Write a message to the mailbox, wait for ack
+ * @hw: pointer to the HW structure
+ * @msg: The message buffer
+ * @size: Length of buffer
+ * @mbx_id: id of mailbox to write
+ *
+ * returns SUCCESS if it successfully copied message into the buffer and
+ * received an ack to that message within delay * timeout period
+ **/
+static int32_t
+rnp_write_posted_mbx_pf(struct rte_eth_dev *dev, u32 *msg, u16 size,
+ enum MBX_ID mbx_id)
+{
+ const struct rnp_mbx_api *ops = RNP_DEV_TO_MBX_OPS(dev);
+ struct rnp_hw *hw = RNP_DEV_TO_HW(dev);
+ struct rnp_mbx_info *mbx = &hw->mbx;
+ int32_t ret_val = -ETIMEDOUT;
+
+ /* exit if either we can't write or there isn't a defined timeout */
+ if (!ops->write || !mbx->timeout)
+ goto out;
+
+ /* send msg and hold buffer lock */
+ if (ops->write)
+ ret_val = ops->write(hw, msg, size, mbx_id);
+
+ /* if msg sent wait until we receive an ack */
+ if (!ret_val)
+ ret_val = rnp_poll_for_ack(dev, mbx_id);
+out:
+ return ret_val;
+}
+
+/**
+ * rnp_check_for_msg_pf - checks to see if the VF has sent mail
+ * @hw: pointer to the HW structure
+ * @vf_number: the VF index
+ *
+ * returns SUCCESS if the VF has set the Status bit or else ERR_MBX
+ **/
+static int32_t rnp_check_for_msg_pf(struct rnp_hw *hw, enum MBX_ID mbx_id)
+{
+ int32_t ret_val = -ETIMEDOUT;
+
+ if (mbx_id == MBX_CM3CPU) {
+ if (rnp_mbx_get_req(hw, CPU2PF_COUNTER) != hw->mbx.cpu_req) {
+ ret_val = 0;
+ /* hw->mbx.stats.reqs++; */
+ }
+ } else {
+ if (rnp_mbx_get_req(hw, VF2PF_COUNTER(mbx_id)) !=
+ hw->mbx.vf_req[mbx_id]) {
+ ret_val = 0;
+ /* hw->mbx.stats.reqs++; */
+ }
+ }
+
+ return ret_val;
+}
+
+/**
+ * rnp_check_for_ack_pf - checks to see if the VF has ACKed
+ * @hw: pointer to the HW structure
+ * @vf_number: the VF index
+ *
+ * returns SUCCESS if the VF has set the Status bit or else ERR_MBX
+ **/
+static int32_t rnp_check_for_ack_pf(struct rnp_hw *hw, enum MBX_ID mbx_id)
+{
+ int32_t ret_val = -ETIMEDOUT;
+
+ if (mbx_id == MBX_CM3CPU) {
+ if (rnp_mbx_get_ack(hw, CPU2PF_COUNTER) != hw->mbx.cpu_ack) {
+ ret_val = 0;
+ /* hw->mbx.stats.acks++; */
+ }
+ } else {
+ if (rnp_mbx_get_ack(hw, VF2PF_COUNTER(mbx_id)) != hw->mbx.vf_ack[mbx_id]) {
+ ret_val = 0;
+ /* hw->mbx.stats.acks++; */
+ }
+ }
+
+ return ret_val;
+}
+
+/**
+ * rnp_obtain_mbx_lock_pf - obtain mailbox lock
+ * @hw: pointer to the HW structure
+ * @mbx_id: the VF index or CPU
+ *
+ * return SUCCESS if we obtained the mailbox lock
+ **/
+static int32_t rnp_obtain_mbx_lock_pf(struct rnp_hw *hw, enum MBX_ID mbx_id)
+{
+ int32_t ret_val = -ETIMEDOUT;
+ int try_cnt = 5000; /* 500ms */
+ u32 CTRL_REG = (mbx_id == MBX_CM3CPU) ?
+ PF2CPU_MBOX_CTRL : PF2VF_MBOX_CTRL(mbx_id);
+
+ while (try_cnt-- > 0) {
+ /* Take ownership of the buffer */
+ mbx_wr32(hw, CTRL_REG, MBOX_CTRL_PF_HOLD_SHM);
+
+ /* reserve mailbox for cm3 use */
+ if (mbx_rd32(hw, CTRL_REG) & MBOX_CTRL_PF_HOLD_SHM)
+ return 0;
+ rte_delay_us_block(100);
+ }
+
+ RNP_PMD_LOG(WARNING, "%s: failed to get:%d lock\n",
+ __func__, mbx_id);
+ return ret_val;
+}
+
+/**
+ * rnp_write_mbx_pf - Places a message in the mailbox
+ * @hw: pointer to the HW structure
+ * @msg: The message buffer
+ * @size: Length of buffer
+ * @mbx_id: the VF index
+ *
+ * returns SUCCESS if it successfully copied message into the buffer
+ **/
+static int32_t rnp_write_mbx_pf(struct rnp_hw *hw, u32 *msg,
+ u16 size, enum MBX_ID mbx_id)
+{
+ u32 DATA_REG = (mbx_id == MBX_CM3CPU) ?
+ CPU_PF_SHM_DATA : PF_VF_SHM_DATA(mbx_id);
+ u32 CTRL_REG = (mbx_id == MBX_CM3CPU) ?
+ PF2CPU_MBOX_CTRL : PF2VF_MBOX_CTRL(mbx_id);
+ int32_t ret_val = 0;
+ u32 stat __rte_unused;
+ u16 i;
+
+ if (size > RNP_VFMAILBOX_SIZE) {
+ RNP_PMD_LOG(ERR, "%s: size:%d should <%d\n", __func__,
+ size, RNP_VFMAILBOX_SIZE);
+ return -EINVAL;
+ }
+
+ /* lock the mailbox to prevent pf/vf/cpu race condition */
+ ret_val = rnp_obtain_mbx_lock_pf(hw, mbx_id);
+ if (ret_val) {
+ RNP_PMD_LOG(WARNING, "PF[%d] Can't Get Mbx-Lock Try Again\n",
+ hw->function);
+ return ret_val;
+ }
+
+ /* copy the caller specified message to the mailbox memory buffer */
+ for (i = 0; i < size; i++) {
+#ifdef MBX_WR_DEBUG
+ mbx_pwr32(hw, DATA_REG + i * 4, msg[i]);
+#else
+ mbx_wr32(hw, DATA_REG + i * 4, msg[i]);
+#endif
+ }
+
+ /* flush msg and acks as we are overwriting the message buffer */
+ if (mbx_id == MBX_CM3CPU)
+ hw->mbx.cpu_ack = rnp_mbx_get_ack(hw, CPU2PF_COUNTER);
+ else
+ hw->mbx.vf_ack[mbx_id] = rnp_mbx_get_ack(hw, VF2PF_COUNTER(mbx_id));
+
+ rnp_mbx_inc_pf_req(hw, mbx_id);
+ rte_mb();
+
+ rte_delay_us(300);
+
+ /* Interrupt VF/CM3 to tell it a message
+ * has been sent and release buffer
+ */
+ mbx_wr32(hw, CTRL_REG, MBOX_CTRL_REQ);
+
+ return 0;
+}
+
+/**
+ * rnp_read_mbx_pf - Read a message from the mailbox
+ * @hw: pointer to the HW structure
+ * @msg: The message buffer
+ * @size: Length of buffer
+ * @vf_number: the VF index
+ *
+ * This function copies a message from the mailbox buffer to the caller's
+ * memory buffer. The presumption is that the caller knows that there was
+ * a message due to a VF/CPU request so no polling for message is needed.
+ **/
+static int32_t rnp_read_mbx_pf(struct rnp_hw *hw, u32 *msg,
+ u16 size, enum MBX_ID mbx_id)
+{
+ u32 BUF_REG = (mbx_id == MBX_CM3CPU) ?
+ CPU_PF_SHM_DATA : PF_VF_SHM_DATA(mbx_id);
+ u32 CTRL_REG = (mbx_id == MBX_CM3CPU) ?
+ PF2CPU_MBOX_CTRL : PF2VF_MBOX_CTRL(mbx_id);
+ int32_t ret_val = -EIO;
+ u32 stat __rte_unused, i;
+ if (size > RNP_VFMAILBOX_SIZE) {
+ RNP_PMD_LOG(ERR, "%s: size:%d should <%d\n", __func__,
+ size, RNP_VFMAILBOX_SIZE);
+ return -EINVAL;
+ }
+ /* lock the mailbox to prevent pf/vf race condition */
+ ret_val = rnp_obtain_mbx_lock_pf(hw, mbx_id);
+ if (ret_val)
+ goto out_no_read;
+
+ /* copy the message from the mailbox memory buffer */
+ for (i = 0; i < size; i++) {
+#ifdef MBX_RD_DEBUG
+ msg[i] = mbx_prd32(hw, BUF_REG + 4 * i);
+#else
+ msg[i] = mbx_rd32(hw, BUF_REG + 4 * i);
+#endif
+ }
+ mbx_wr32(hw, BUF_REG, 0);
+
+ /* update req. used by rnpvf_check_for_msg_vf */
+ if (mbx_id == MBX_CM3CPU)
+ hw->mbx.cpu_req = rnp_mbx_get_req(hw, CPU2PF_COUNTER);
+ else
+ hw->mbx.vf_req[mbx_id] = rnp_mbx_get_req(hw, VF2PF_COUNTER(mbx_id));
+
+ /* this ack maybe too earier? */
+ /* Acknowledge receipt and release mailbox, then we're done */
+ rnp_mbx_inc_pf_ack(hw, mbx_id);
+
+ rte_mb();
+
+ /* free ownership of the buffer */
+ mbx_wr32(hw, CTRL_REG, 0);
+
+out_no_read:
+
+ return ret_val;
+}
+
+static void rnp_mbx_reset_pf(struct rnp_hw *hw)
+{
+ int v;
+
+ /* reset pf->cm3 status */
+ v = mbx_rd32(hw, CPU2PF_COUNTER);
+ hw->mbx.cpu_req = v & 0xffff;
+ hw->mbx.cpu_ack = (v >> 16) & 0xffff;
+ /* release pf->cm3 buffer lock */
+ mbx_wr32(hw, PF2CPU_MBOX_CTRL, 0);
+
+ rte_mb();
+ /* enable irq to fw */
+ mbx_wr32(hw, CPU_PF_MBOX_MASK, 0);
+}
+
+static int get_pfvfnum(struct rnp_hw *hw)
+{
+ uint32_t addr_mask;
+ uint32_t offset;
+ uint32_t val;
+#define RNP_PF_NUM_REG (0x75f000)
+#define RNP_PFVF_SHIFT (4)
+#define RNP_PF_SHIFT (6)
+#define RNP_PF_BIT_MASK BIT(6)
+ addr_mask = hw->iobar0_len - 1;
+ offset = RNP_PF_NUM_REG & addr_mask;
+ val = rnp_io_rd(hw->iobar0, offset);
+
+ return val >> RNP_PFVF_SHIFT;
+}
+
+const struct rnp_mbx_api rnp_mbx_pf_ops = {
+ .read = rnp_read_mbx_pf,
+ .write = rnp_write_mbx_pf,
+ .read_posted = rnp_read_posted_mbx_pf,
+ .write_posted = rnp_write_posted_mbx_pf,
+ .check_for_msg = rnp_check_for_msg_pf,
+ .check_for_ack = rnp_check_for_ack_pf,
+};
+
+void *rnp_memzone_reserve(const char *name, unsigned int size)
+{
+#define NO_FLAGS 0
+ const struct rte_memzone *mz = NULL;
+
+ if (name) {
+ if (size) {
+ mz = rte_memzone_reserve(name, size,
+ rte_socket_id(), NO_FLAGS);
+ if (mz)
+ memset(mz->addr, 0, size);
+ } else {
+ mz = rte_memzone_lookup(name);
+ }
+ return mz ? mz->addr : NULL;
+ }
+ return NULL;
+}
+
+void rnp_init_mbx_ops_pf(struct rnp_hw *hw)
+{
+ struct rnp_eth_adapter *adapter = hw->back;
+ struct rnp_mbx_info *mbx = &hw->mbx;
+ struct mbx_req_cookie *cookie;
+ uint32_t vf_isolat_off;
+
+ mbx->size = RNP_VFMAILBOX_SIZE;
+ mbx->usec_delay = RNP_MBX_DELAY_US;
+ mbx->timeout = (RNP_MBX_TIMEOUT_SECONDS * 1000 * 1000) /
+ mbx->usec_delay;
+ if (hw->device_id == RNP_DEV_ID_N10G) {
+ vf_isolat_off = RNP_VF_ISOLATE_CTRL &
+ (hw->iobar0_len - 1);
+ rnp_io_wr(hw->iobar0, vf_isolat_off, 0);
+ }
+ mbx->sriov_st = 0;
+ hw->pf_vf_num = get_pfvfnum(hw);
+ mbx->vf_num = UINT16_MAX;
+ mbx->pf_num = (hw->pf_vf_num & RNP_PF_BIT_MASK) >> RNP_PF_SHIFT;
+ hw->function = mbx->pf_num;
+ /* Retrieving and storing the HW base address of device */
+ rnp_reg_offset_init(hw);
+ snprintf(hw->cookie_p_name, RTE_MEMZONE_NAMESIZE, "mbx_req_cookie%d_%d",
+ hw->function, adapter->eth_dev->data->port_id);
+ hw->cookie_pool = rnp_memzone_reserve(hw->cookie_p_name,
+ sizeof(struct mbx_req_cookie));
+
+ cookie = (struct mbx_req_cookie *)hw->cookie_pool;
+ if (cookie) {
+ cookie->timeout_ms = 1000;
+ cookie->magic = COOKIE_MAGIC;
+ cookie->priv_len = RNP_MAX_SHARE_MEM;
+ }
+
+ rnp_mbx_reset_pf(hw);
+}
diff --git a/drivers/net/rnp/rnp_mbx.h b/drivers/net/rnp/rnp_mbx.h
new file mode 100644
index 0000000000..87949c1726
--- /dev/null
+++ b/drivers/net/rnp/rnp_mbx.h
@@ -0,0 +1,139 @@
+#ifndef __TSRN10_MBX_H__
+#define __TSRN10_MBX_H__
+
+#define VF_NUM_MASK_TEMP (0xff0)
+#define VF_NUM_OFF (4)
+#define RNP_VF_NUM (0x75f000)
+#define RNP_VF_NB_MASK (0x3f)
+#define RNP_PF_NB_MASK (0x40)
+#define RNP_VF_ISOLATE_CTRL (0x7982fc)
+#define RNP_IS_SRIOV BIT(7)
+#define RNP_SRIOV_ST_SHIFT (24)
+#define RNP_VF_DEFAULT_PORT (0)
+
+/* Mbx Ctrl state */
+#define RNP_VFMAILBOX_SIZE (14) /* 16 32 bit words - 64 bytes */
+#define TSRN10_VFMBX_SIZE (RNP_VFMAILBOX_SIZE)
+#define RNP_VT_MSGTYPE_ACK (0x80000000)
+
+#define RNP_VT_MSGTYPE_NACK (0x40000000)
+/* Messages below or'd with * this are the NACK */
+#define RNP_VT_MSGTYPE_CTS (0x20000000)
+/* Indicates that VF is still
+ *clear to send requests
+ */
+#define RNP_VT_MSGINFO_SHIFT (16)
+
+#define RNP_VT_MSGINFO_MASK (0xFF << RNP_VT_MSGINFO_SHIFT)
+/* The mailbox memory size is 64 bytes accessed by 32-bit registers */
+#define RNP_VLVF_VIEN (0x80000000) /* filter is valid */
+#define RNP_VLVF_ENTRIES (64)
+#define RNP_VLVF_VLANID_MASK (0x00000FFF)
+/* Every VF own 64 bytes mem for communitate accessed by 32-bit */
+
+#define RNP_VF_RESET (0x01) /* VF requests reset */
+#define RNP_VF_SET_MAC_ADDR (0x02) /* VF requests PF to set MAC addr */
+#define RNP_VF_SET_MULTICAST (0x03) /* VF requests PF to set MC addr */
+#define RNP_VF_SET_VLAN (0x04) /* VF requests PF to set VLAN */
+
+#define RNP_VF_SET_LPE (0x05) /* VF requests PF to set VMOLR.LPE */
+#define RNP_VF_SET_MACVLAN (0x06) /* VF requests PF for unicast filter */
+#define RNP_VF_GET_MACVLAN (0x07) /* VF requests mac */
+#define RNP_VF_API_NEGOTIATE (0x08) /* negotiate API version */
+#define RNP_VF_GET_QUEUES (0x09) /* get queue configuration */
+#define RNP_VF_GET_LINK (0x10) /* get link status */
+
+#define RNP_VF_SET_VLAN_STRIP (0x0a) /* VF Requests PF to set VLAN STRIP */
+#define RNP_VF_REG_RD (0x0b) /* VF Read Reg */
+#define RNP_VF_GET_MAX_MTU (0x0c) /* VF Get Max Mtu */
+#define RNP_VF_SET_MTU (0x0d) /* VF Set Mtu */
+#define RNP_VF_GET_FW (0x0e) /* VF Get Firmware Version */
+
+#define RNP_PF_VFNUM_MASK GENMASK(26, 21)
+
+#define RNP_PF_SET_FCS (0x10) /* PF set fcs status */
+#define RNP_PF_SET_PAUSE (0x11) /* PF set pause status */
+#define RNP_PF_SET_FT_PADDING (0x12) /* PF set ft padding status */
+#define RNP_PF_SET_VLAN_FILTER (0x13) /* PF set ntuple status */
+#define RNP_PF_SET_VLAN (0x14)
+#define RNP_PF_SET_LINK (0x15)
+#define RNP_PF_SET_SPEED_40G BIT(8)
+#define RNP_PF_SET_SPEED_10G BIT(7)
+#define RNP_PF_SET_SPEED_1G BIT(5)
+#define RNP_PF_SET_SPEED_100M BIT(3)
+
+#define RNP_PF_SET_MTU (0x16)
+#define RNP_PF_SET_RESET (0x17)
+#define RNP_PF_LINK_UP BIT(31)
+#define RNP_PF_SPEED_MASK GENMASK(15, 0)
+
+/* Define mailbox register bits */
+#define RNP_PF_REMOVE (0x0f)
+
+/* Mailbox API ID VF Request */
+/* length of permanent address message returned from PF */
+#define RNP_VF_PERMADDR_MSG_LEN (11)
+#define RNP_VF_TX_QUEUES (1) /* number of Tx queues supported */
+#define RNP_VF_RX_QUEUES (2) /* number of Rx queues supported */
+#define RNP_VF_TRANS_VLAN (3) /* Indication of port vlan */
+#define RNP_VF_DEF_QUEUE (4) /* Default queue offset */
+/* word in permanent address message with the current multicast type */
+#define RNP_VF_VLAN_WORD (5)
+#define RNP_VF_PHY_TYPE_WORD (6)
+#define RNP_VF_FW_VERSION_WORD (7)
+#define RNP_VF_LINK_STATUS_WORD (8)
+#define RNP_VF_AXI_MHZ (9)
+#define RNP_VF_RNP_VF_FEATURE (10)
+#define RNP_VF_RNP_VF_FILTER_EN BIT(0)
+
+#define RNP_LINK_SPEED_UNKNOWN 0
+#define RNP_LINK_SPEED_10_FULL BIT(2)
+#define RNP_LINK_SPEED_100_FULL BIT(3)
+#define RNP_LINK_SPEED_1GB_FULL BIT(4)
+#define RNP_LINK_SPEED_10GB_FULL BIT(5)
+#define RNP_LINK_SPEED_40GB_FULL BIT(6)
+#define RNP_LINK_SPEED_25GB_FULL BIT(7)
+#define RNP_LINK_SPEED_50GB_FULL BIT(8)
+#define RNP_LINK_SPEED_100GB_FULL BIT(9)
+#define RNP_LINK_SPEED_10_HALF BIT(10)
+#define RNP_LINK_SPEED_100_HALF BIT(11)
+#define RNP_LINK_SPEED_1GB_HALF BIT(12)
+
+/* Mailbox API ID PF Request */
+#define RNP_VF_MC_TYPE_WORD (3)
+#define RNP_VF_DMA_VERSION_WORD (4)
+/* Get Queue write-back reference value */
+#define RNP_PF_CONTROL_PRING_MSG (0x0100) /* PF control message */
+
+#define TSRN10_MBX_VECTOR_ID (0)
+#define TSRN10_PF2VF_MBX_VEC_CTR(n) (0xa5000 + 0x4 * (n))
+
+#define RNP_VF_INIT_TIMEOUT (200) /* Number of retries to clear RSTI */
+#define RNP_VF_MBX_INIT_TIMEOUT (2000) /* number of retries on mailbox */
+
+#define MBOX_CTRL_REQ (1 << 0) /* WO */
+#define MBOX_CTRL_VF_HOLD_SHM (1 << 2) /* VF:WR, PF:RO */
+#define VF_NUM_MASK 0x3f
+#define VFNUM(num) ((num) & VF_NUM_MASK)
+
+#define PF_VF_SHM(vf) \
+ ((0xa6000) + (64 * (vf))) /* for PF1 rtl will remap 6000 to 0xb000 */
+#define PF2VF_COUNTER(vf) (PF_VF_SHM(vf) + 0)
+#define VF2PF_COUNTER(vf) (PF_VF_SHM(vf) + 4)
+#define PF_VF_SHM_DATA(vf) (PF_VF_SHM(vf) + 8)
+#define VF2PF_MBOX_CTRL(vf) ((0xa7000) + (4 * (vf)))
+
+/* Error Codes */
+#define RNP_ERR_INVALID_MAC_ADDR (-1)
+#define RNP_ERR_MBX (-100)
+
+#define RNP_MBX_DELAY_US (100) /* Delay us for Retry */
+/* Max Retry Time */
+#define RNP_MBX_TIMEOUT_SECONDS (2) /* Max Retry Time 2s */
+#define RNP_ARRAY_OPCODE_OFFSET (0)
+#define RNP_ARRAY_CTRL_OFFSET (1)
+
+void rnp_init_mbx_ops_pf(struct rnp_hw *hw);
+extern const struct rnp_mbx_api rnp_mbx_pf_ops;
+void *rnp_memzone_reserve(const char *name, unsigned int size);
+#endif
diff --git a/drivers/net/rnp/rnp_mbx_fw.c b/drivers/net/rnp/rnp_mbx_fw.c
new file mode 100644
index 0000000000..6fe008351b
--- /dev/null
+++ b/drivers/net/rnp/rnp_mbx_fw.c
@@ -0,0 +1,271 @@
+#include <stdio.h>
+
+#include <rte_version.h>
+#include <ethdev_pci.h>
+#include <rte_malloc.h>
+#include <rte_alarm.h>
+
+#include "rnp.h"
+#include "rnp_mbx.h"
+#include "rnp_mbx_fw.h"
+#include "rnp_logs.h"
+
+static int
+rnp_fw_send_cmd_wait(struct rte_eth_dev *dev, struct mbx_fw_cmd_req *req,
+ struct mbx_fw_cmd_reply *reply)
+{
+ const struct rnp_mbx_api *ops = RNP_DEV_TO_MBX_OPS(dev);
+ struct rnp_hw *hw = RNP_DEV_TO_HW(dev);
+ int err;
+
+ rte_spinlock_lock(&hw->fw_lock);
+
+ err = ops->write_posted(dev, (u32 *)req,
+ (req->datalen + MBX_REQ_HDR_LEN) / 4, MBX_FW);
+ if (err) {
+ RNP_PMD_LOG(ERR, "%s: write_posted failed! err:0x%x\n",
+ __func__, err);
+ rte_spinlock_unlock(&hw->fw_lock);
+ return err;
+ }
+
+ err = ops->read_posted(dev, (u32 *)reply, sizeof(*reply) / 4, MBX_FW);
+ rte_spinlock_unlock(&hw->fw_lock);
+ if (err) {
+ RNP_PMD_LOG(ERR,
+ "%s: read_posted failed! err:0x%x. "
+ "req-op:0x%x\n",
+ __func__,
+ err,
+ req->opcode);
+ goto err_quit;
+ }
+
+ if (reply->error_code) {
+ RNP_PMD_LOG(ERR,
+ "%s: reply err:0x%x. req-op:0x%x\n",
+ __func__,
+ reply->error_code,
+ req->opcode);
+ err = -reply->error_code;
+ goto err_quit;
+ }
+
+ return 0;
+err_quit:
+ RNP_PMD_LOG(ERR,
+ "%s:PF[%d]: req:%08x_%08x_%08x_%08x "
+ "reply:%08x_%08x_%08x_%08x\n",
+ __func__,
+ hw->function,
+ ((int *)req)[0],
+ ((int *)req)[1],
+ ((int *)req)[2],
+ ((int *)req)[3],
+ ((int *)reply)[0],
+ ((int *)reply)[1],
+ ((int *)reply)[2],
+ ((int *)reply)[3]);
+
+ return err;
+}
+
+static int rnp_mbx_fw_post_req(struct rte_eth_dev *dev,
+ struct mbx_fw_cmd_req *req,
+ struct mbx_req_cookie *cookie)
+{
+ const struct rnp_mbx_api *ops = RNP_DEV_TO_MBX_OPS(dev);
+ struct rnp_hw *hw = RNP_DEV_TO_HW(dev);
+ int err = 0;
+ int timeout_cnt;
+#define WAIT_MS 10
+
+ cookie->done = 0;
+
+ rte_spinlock_lock(&hw->fw_lock);
+
+ /* down_interruptible(&pf_cpu_lock); */
+ err = ops->write(hw, (u32 *)req,
+ (req->datalen + MBX_REQ_HDR_LEN) / 4, MBX_FW);
+ if (err) {
+ RNP_PMD_LOG(ERR, "rnp_write_mbx failed!\n");
+ goto quit;
+ }
+
+ timeout_cnt = cookie->timeout_ms / WAIT_MS;
+ while (timeout_cnt > 0) {
+ rte_delay_ms(WAIT_MS);
+ timeout_cnt--;
+ if (cookie->done)
+ break;
+ }
+
+quit:
+ rte_spinlock_unlock(&hw->fw_lock);
+ return err;
+}
+
+static int rnp_fw_get_capablity(struct rte_eth_dev *dev,
+ struct phy_abilities *abil)
+{
+ struct mbx_fw_cmd_reply reply;
+ struct mbx_fw_cmd_req req;
+ int err;
+
+ memset(&req, 0, sizeof(req));
+ memset(&reply, 0, sizeof(reply));
+
+ build_phy_abalities_req(&req, &req);
+
+ err = rnp_fw_send_cmd_wait(dev, &req, &reply);
+ if (err)
+ return err;
+
+ memcpy(abil, &reply.phy_abilities, sizeof(*abil));
+
+ return 0;
+}
+
+#define RNP_MBX_API_MAX_RETRY (10)
+int rnp_mbx_get_capability(struct rte_eth_dev *dev,
+ int *lane_mask,
+ int *nic_mode)
+{
+ struct rnp_hw *hw = RNP_DEV_TO_HW(dev);
+ struct phy_abilities ablity;
+ uint16_t temp_lmask;
+ uint16_t lane_bit = 0;
+ uint16_t retry = 0;
+ int lane_cnt = 0;
+ uint8_t lane_idx;
+ int err = -EIO;
+ uint8_t idx;
+
+ memset(&ablity, 0, sizeof(ablity));
+
+ /* enable CM3CPU to PF MBX IRQ */
+ do {
+ err = rnp_fw_get_capablity(dev, &ablity);
+ if (retry > RNP_MBX_API_MAX_RETRY)
+ break;
+ retry++;
+ } while (err);
+ if (!err) {
+ hw->lane_mask = ablity.lane_mask;
+ hw->nic_mode = ablity.nic_mode;
+ hw->pfvfnum = ablity.pfnum;
+ hw->fw_version = ablity.fw_version;
+ hw->axi_mhz = ablity.axi_mhz;
+ hw->fw_uid = ablity.fw_uid;
+ if (ablity.phy_type == PHY_TYPE_SGMII) {
+ hw->is_sgmii = 1;
+ hw->sgmii_phy_id = ablity.phy_id;
+ }
+
+ if (ablity.ext_ablity != 0xffffffff && ablity.e.valid) {
+ hw->ncsi_en = (ablity.e.ncsi_en == 1);
+ hw->ncsi_rar_entries = 1;
+ }
+
+ if (hw->nic_mode == RNP_SINGLE_10G &&
+ hw->fw_version >= 0x00050201 &&
+ ablity.speed == RTE_ETH_SPEED_NUM_10G) {
+ hw->force_speed_stat = FORCE_SPEED_STAT_DISABLED;
+ hw->force_10g_1g_speed_ablity = 1;
+ }
+
+ if (lane_mask)
+ *lane_mask = hw->lane_mask;
+ if (nic_mode)
+ *nic_mode = hw->nic_mode;
+
+ lane_cnt = __builtin_popcount(hw->lane_mask);
+ temp_lmask = hw->lane_mask;
+ for (idx = 0; idx < lane_cnt; idx++) {
+ hw->phy_port_ids[idx] = ablity.port_ids[idx];
+ lane_bit = ffs(temp_lmask) - 1;
+ lane_idx = ablity.port_ids[idx] % lane_cnt;
+ hw->lane_of_port[lane_idx] = lane_bit;
+ temp_lmask &= ~BIT(lane_bit);
+ }
+ hw->max_port_num = lane_cnt;
+ }
+
+ RNP_PMD_LOG(INFO,
+ "%s: nic-mode:%d lane_cnt:%d lane_mask:0x%x "
+ "pfvfnum:0x%x, fw_version:0x%08x, ports:%d-%d-%d-%d ncsi:en:%d\n",
+ __func__,
+ hw->nic_mode,
+ lane_cnt,
+ hw->lane_mask,
+ hw->pfvfnum,
+ ablity.fw_version,
+ ablity.port_ids[0],
+ ablity.port_ids[1],
+ ablity.port_ids[2],
+ ablity.port_ids[3],
+ hw->ncsi_en);
+
+ if (lane_cnt <= 0 || lane_cnt > 4)
+ return -EIO;
+
+ return err;
+}
+
+int rnp_mbx_link_event_enable(struct rte_eth_dev *dev, int enable)
+{
+ const struct rnp_mbx_api *ops = RNP_DEV_TO_MBX_OPS(dev);
+ struct rnp_hw *hw = RNP_DEV_TO_HW(dev);
+ struct mbx_fw_cmd_reply reply;
+ struct mbx_fw_cmd_req req;
+ int err, v;
+
+ memset(&req, 0, sizeof(req));
+ memset(&reply, 0, sizeof(reply));
+
+ rte_spinlock_lock(&hw->fw_lock);
+ if (enable) {
+ v = rnp_rd_reg(hw->link_sync);
+ v &= ~RNP_FIRMWARE_SYNC_MASK;
+ v |= RNP_FIRMWARE_SYNC_MAGIC;
+ rnp_wr_reg(hw->link_sync, v);
+ } else {
+ rnp_wr_reg(hw->link_sync, 0);
+ }
+ rte_spinlock_unlock(&hw->fw_lock);
+
+ build_link_set_event_mask(&req, BIT(EVT_LINK_UP),
+ (enable & 1) << EVT_LINK_UP, &req);
+
+ rte_spinlock_lock(&hw->fw_lock);
+ err = ops->write_posted(dev, (u32 *)&req,
+ (req.datalen + MBX_REQ_HDR_LEN) / 4, MBX_FW);
+ rte_spinlock_unlock(&hw->fw_lock);
+
+ rte_delay_ms(200);
+
+ return err;
+}
+
+int rnp_mbx_fw_reset_phy(struct rte_eth_dev *dev)
+{
+ struct rnp_hw *hw = RNP_DEV_TO_HW(dev);
+ struct mbx_fw_cmd_reply reply;
+ struct mbx_req_cookie *cookie;
+ struct mbx_fw_cmd_req req;
+
+ memset(&req, 0, sizeof(req));
+ memset(&reply, 0, sizeof(reply));
+
+ if (hw->mbx.irq_enabled) {
+ cookie = rnp_memzone_reserve(hw->cookie_p_name, 0);
+ if (!cookie)
+ return -ENOMEM;
+ memset(cookie->priv, 0, cookie->priv_len);
+ build_reset_phy_req(&req, cookie);
+ return rnp_mbx_fw_post_req(dev, &req, cookie);
+ }
+ build_reset_phy_req(&req, &req);
+
+ return rnp_fw_send_cmd_wait(dev, &req, &reply);
+}
diff --git a/drivers/net/rnp/rnp_mbx_fw.h b/drivers/net/rnp/rnp_mbx_fw.h
new file mode 100644
index 0000000000..439090b5a3
--- /dev/null
+++ b/drivers/net/rnp/rnp_mbx_fw.h
@@ -0,0 +1,22 @@
+#ifndef __RNP_MBX_FW_H__
+#define __RNP_MBX_FW_H__
+
+struct mbx_fw_cmd_reply;
+typedef void (*cookie_cb)(struct mbx_fw_cmd_reply *reply, void *priv);
+#define RNP_MAX_SHARE_MEM (8 * 8)
+struct mbx_req_cookie {
+ int magic;
+#define COOKIE_MAGIC 0xCE
+ cookie_cb cb;
+ int timeout_ms;
+ int errcode;
+
+ /* wait_queue_head_t wait; */
+ volatile int done;
+ int priv_len;
+ char priv[RNP_MAX_SHARE_MEM];
+};
+struct mbx_fw_cmd_reply {
+} __rte_cache_aligned;
+
+#endif /* __RNP_MBX_FW_H__*/
--
2.27.0
^ permalink raw reply related [flat|nested] 27+ messages in thread* Re: [PATCH v6 4/8] net/rnp: add mbx basic api feature
2023-09-01 2:30 ` [PATCH v6 4/8] net/rnp: add mbx basic api feature Wenbo Cao
@ 2023-09-05 15:45 ` Ferruh Yigit
2023-09-06 10:32 ` 11
0 siblings, 1 reply; 27+ messages in thread
From: Ferruh Yigit @ 2023-09-05 15:45 UTC (permalink / raw)
To: Wenbo Cao; +Cc: dev, thomas, andrew.rybchenko, yaojun, Stephen Hemminger
On 9/1/2023 3:30 AM, Wenbo Cao wrote:
> mbx base code is for communicate with the firmware
>
I assume mbx is mailbox, can you please use full name in the patch title
and commit log.
How the parties get notified about new messages, is the interrupt
support enabled or polling, can you please elabore the support in the
commit log?
> Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
> Suggested-by: Stephen Hemminger <stephen@networkplumber.org>
> ---
> drivers/net/rnp/base/rnp_api.c | 23 ++
> drivers/net/rnp/base/rnp_api.h | 7 +
> drivers/net/rnp/base/rnp_cfg.h | 7 +
> drivers/net/rnp/base/rnp_dma_regs.h | 73 ++++
> drivers/net/rnp/base/rnp_eth_regs.h | 124 +++++++
> drivers/net/rnp/base/rnp_hw.h | 112 +++++-
Why there is no meson file in base folder?
> drivers/net/rnp/meson.build | 1 +
> drivers/net/rnp/rnp.h | 35 ++
> drivers/net/rnp/rnp_ethdev.c | 70 +++-
> drivers/net/rnp/rnp_logs.h | 9 +
> drivers/net/rnp/rnp_mbx.c | 522 ++++++++++++++++++++++++++++
> drivers/net/rnp/rnp_mbx.h | 139 ++++++++
> drivers/net/rnp/rnp_mbx_fw.c | 271 +++++++++++++++
> drivers/net/rnp/rnp_mbx_fw.h | 22 ++
Should 'rnp_mbx*' files also go under base folder? I can see they use
some 'u8' like typedef, making me think they are shared between other
platforms, if so base folder suits better.
> 14 files changed, 1412 insertions(+), 3 deletions(-)
> create mode 100644 drivers/net/rnp/base/rnp_api.c
> create mode 100644 drivers/net/rnp/base/rnp_api.h
> create mode 100644 drivers/net/rnp/base/rnp_cfg.h
> create mode 100644 drivers/net/rnp/base/rnp_dma_regs.h
> create mode 100644 drivers/net/rnp/base/rnp_eth_regs.h
> create mode 100644 drivers/net/rnp/rnp_mbx.c
> create mode 100644 drivers/net/rnp/rnp_mbx.h
> create mode 100644 drivers/net/rnp/rnp_mbx_fw.c
> create mode 100644 drivers/net/rnp/rnp_mbx_fw.h
>
> diff --git a/drivers/net/rnp/base/rnp_api.c b/drivers/net/rnp/base/rnp_api.c
> new file mode 100644
> index 0000000000..550da6217d
> --- /dev/null
> +++ b/drivers/net/rnp/base/rnp_api.c
> @@ -0,0 +1,23 @@
> +#include "rnp.h"
> +#include "rnp_api.h"
> +
> +int
> +rnp_init_hw(struct rte_eth_dev *dev)
> +{
>
Base code is mostly for works as HAL and you don't need to pass "struct
rte_eth_dev" in this level, but mostly adapte (struct rnp_eth_adapter)
or hw (struct rnp_hw) should be sufficient.
I assume you are passing "struct rte_eth_dev" because "struct
rnp_mac_api" is stored in 'process_private', this part I am not clear.
Why need 'mac_ops' pointers for secondary process?
Primary process should be doing all hw initialization, otherwise both
primary and secondar(y|ies) trying to configure the MAC can cause
unexpected results.
Seconday process is only for datapath, which can access to the queues
and do packet processing.
Needing "mac_ops" in the secondary can be bad design desicion, can you
please double check it?
<...>
> diff --git a/drivers/net/rnp/rnp.h b/drivers/net/rnp/rnp.h
> index cab1b8e85d..6e12885877 100644
> --- a/drivers/net/rnp/rnp.h
> +++ b/drivers/net/rnp/rnp.h
> @@ -3,6 +3,7 @@
> */
> #ifndef __RNP_H__
> #define __RNP_H__
> +#include <rte_log.h>
>
> #include "base/rnp_hw.h"
>
> @@ -14,14 +15,17 @@
>
> struct rnp_eth_port {
> struct rnp_eth_adapter *adapt;
> + struct rnp_hw *hw;
adapter already has the 'hw', is there a specific reason not to access
it via 'port->adapt->hw'
> struct rte_eth_dev *eth_dev;
> } __rte_cache_aligned;
>
> struct rnp_share_ops {
> + const struct rnp_mbx_api *mbx_api;
> } __rte_cache_aligned;
>
> struct rnp_eth_adapter {
> struct rnp_hw hw;
> + uint16_t max_vfs;
> struct rte_pci_device *pdev;
> struct rte_eth_dev *eth_dev; /* primary eth_dev */
> struct rnp_eth_port *ports[RNP_MAX_PORT_OF_PF];
> @@ -34,5 +38,36 @@ struct rnp_eth_adapter {
> (((struct rnp_eth_port *)((eth_dev)->data->dev_private)))
> #define RNP_DEV_TO_ADAPTER(eth_dev) \
> ((struct rnp_eth_adapter *)(RNP_DEV_TO_PORT(eth_dev)->adapt))
> +#define RNP_DEV_TO_HW(eth_dev) \
> + (&((struct rnp_eth_adapter *)(RNP_DEV_TO_PORT((eth_dev))->adapt))->hw)
> +#define RNP_DEV_PP_PRIV_TO_MBX_OPS(dev) \
> + (((struct rnp_share_ops *)(dev)->process_private)->mbx_api)
> +#define RNP_DEV_TO_MBX_OPS(dev) RNP_DEV_PP_PRIV_TO_MBX_OPS(dev)
>
> +static inline void rnp_reg_offset_init(struct rnp_hw *hw)
> +{
> + uint16_t i;
> +
> + if (hw->device_id == RNP_DEV_ID_N10G && hw->mbx.pf_num) {
> + hw->iobar4 = (void *)((uint8_t *)hw->iobar4 + 0x100000);
> + hw->msix_base = (void *)((uint8_t *)hw->iobar4 + 0xa4000);
> + hw->msix_base = (void *)((uint8_t *)hw->msix_base + 0x200);
> + } else {
> + hw->msix_base = (void *)((uint8_t *)hw->iobar4 + 0xa4000);
> + }
> + /* === dma status/config====== */
> + hw->link_sync = (void *)((uint8_t *)hw->iobar4 + 0x000c);
> + hw->dma_axi_en = (void *)((uint8_t *)hw->iobar4 + 0x0010);
> + hw->dma_axi_st = (void *)((uint8_t *)hw->iobar4 + 0x0014);
> +
> + if (hw->mbx.pf_num)
> + hw->msix_base = (void *)((uint8_t *)0x200);
> + /* === queue registers === */
> + hw->dma_base = (void *)((uint8_t *)hw->iobar4 + 0x08000);
> + hw->veb_base = (void *)((uint8_t *)hw->iobar4 + 0x0);
> + hw->eth_base = (void *)((uint8_t *)hw->iobar4 + 0x10000);
There are lots of hardcoded values in this function, can you make them
macros?
> + /* mac */
> + for (i = 0; i < RNP_MAX_HW_PORT_PERR_PF; i++)
> + hw->mac_base[i] = (void *)((uint8_t *)hw->iobar4 + 0x60000 + 0x10000 * i);
> +}
> #endif /* __RNP_H__ */
> diff --git a/drivers/net/rnp/rnp_ethdev.c b/drivers/net/rnp/rnp_ethdev.c
> index ae737643a7..a2dc27548a 100644
> --- a/drivers/net/rnp/rnp_ethdev.c
> +++ b/drivers/net/rnp/rnp_ethdev.c
> @@ -8,6 +8,7 @@
> #include <ethdev_driver.h>
>
> #include "rnp.h"
> +#include "rnp_mbx.h"
> #include "rnp_logs.h"
>
> static int
> @@ -89,6 +90,58 @@ rnp_alloc_eth_port(struct rte_pci_device *primary_pci, char *name)
> return NULL;
> }
>
> +static void rnp_get_nic_attr(struct rnp_eth_adapter *adapter)
> +{
> + RTE_SET_USED(adapter);
> +}
> +
> +static int
> +rnp_process_resource_init(struct rte_eth_dev *eth_dev)
> +{
> + struct rnp_share_ops *share_priv;
> +
> + /* allocate process_private memory this must can't
> + * belone to the dpdk mem resource manager
> + * such as from rte_malloc or rte_dma_zone..
> + */
> + /* use the process_prive point to resolve secondary process
> + * use point-func. This point is per process will be safe to cover.
> + * This will cause secondary process core-dump because of IPC
> + * Secondary will call primary process point func virt-address
> + * secondary process don't alloc user/pmd to alloc or free
> + * the memory of dpdk-mem resource it will cause hugepage
> + * mem exception
> + * be careful for secondary Process to use the share-mem of
> + * point correlation
> + */
As commented above, I am not sure why you need this, it should be
sufficient to have mac_ops in primary proccess.
> + share_priv = calloc(1, sizeof(*share_priv));
> + if (!share_priv) {
> + PMD_DRV_LOG(ERR, "calloc share_priv failed");
> + return -ENOMEM;
> + }
> + memset(share_priv, 0, sizeof(*share_priv));
> + eth_dev->process_private = share_priv;
> +
> + return 0;
> +}
> +
> +static void
> +rnp_common_ops_init(struct rnp_eth_adapter *adapter)
> +{
> + struct rnp_share_ops *share_priv;
> +
> + share_priv = adapter->share_priv;
> + share_priv->mbx_api = &rnp_mbx_pf_ops;
> +}
> +
> +static int
> +rnp_special_ops_init(struct rte_eth_dev *eth_dev)
> +{
> + RTE_SET_USED(eth_dev);
> +
> + return 0;
> +}
> +
> static int
> rnp_eth_dev_init(struct rte_eth_dev *dev)
> {
> @@ -124,6 +177,20 @@ rnp_eth_dev_init(struct rte_eth_dev *dev)
> hw->device_id = pci_dev->id.device_id;
> hw->vendor_id = pci_dev->id.vendor_id;
> hw->device_id = pci_dev->id.device_id;
> + adapter->max_vfs = pci_dev->max_vfs;
> + ret = rnp_process_resource_init(dev);
> + if (ret) {
> + PMD_DRV_LOG(ERR, "share prive resource init failed");
> + return ret;
> + }
> + adapter->share_priv = dev->process_private;
Isn't 'adapter' shared betwen primary and secondary ('port' is
dev_private and 'port->adapt' is 'adapter'), so updating
"adapter->share_priv" defeats the purpose of process_private, no?
> + rnp_common_ops_init(adapter);
> + rnp_get_nic_attr(adapter);
> + /* We need Use Device Id To Change The Resource Mode */
> + rnp_special_ops_init(dev);
>
You can add this when it is needed, so in same patch we can see where it
is called and how it is implemented, right now it is not possible that
what 'special' init does.
> + port->adapt = adapter;
> + port->hw = hw;
At this stage port is NULL, probably above needs to go to other patch
that sets 'port'.
> + rnp_init_mbx_ops_pf(hw);
> for (p_id = 0; p_id < adapter->num_ports; p_id++) {
> /* port 0 resource has been allocated When Probe */
> if (!p_id) {
> @@ -158,11 +225,10 @@ rnp_eth_dev_init(struct rte_eth_dev *dev)
> continue;
> if (port->eth_dev) {
> rnp_dev_close(port->eth_dev);
> - rte_eth_dev_release_port(port->eth_dev);
> if (port->eth_dev->process_private)
> free(port->eth_dev->process_private);
> + rte_eth_dev_release_port(port->eth_dev);
> }
> - rte_free(port);
> }
> rte_free(adapter);
>
> diff --git a/drivers/net/rnp/rnp_logs.h b/drivers/net/rnp/rnp_logs.h
> index 1b3ee33745..f1648aabb5 100644
> --- a/drivers/net/rnp/rnp_logs.h
> +++ b/drivers/net/rnp/rnp_logs.h
> @@ -13,6 +13,15 @@ extern int rnp_drv_logtype;
> #define RNP_PMD_DRV_LOG(level, fmt, args...) \
> rte_log(RTE_LOG_##level, rnp_drv_logtype, \
> "%s() " fmt, __func__, ##args)
> +#define PMD_DRV_LOG_RAW(level, fmt, args...) \
> + rte_log(RTE_LOG_ ## level, rnp_drv_logtype, "%s(): " fmt, \
> + __func__, ## args)
> +#define PMD_DRV_LOG(level, fmt, args...) \
> + PMD_DRV_LOG_RAW(level, fmt "\n", ## args)
> +
> +#define RNP_PMD_LOG(level, fmt, args...) \
> + rte_log(RTE_LOG_##level, rnp_drv_logtype, \
> + "rnp_net: (%d) " fmt, __LINE__, ##args)
With these there are three different macros used to log in
'rnp_drv_logtype' type:
RNP_PMD_LOG
PMD_DRV_LOG
RNP_PMD_DRV_LOG
'RNP_PMD_DRV_LOG' and 'PMD_DRV_LOG' looks exact same, do you need all
these macros, why not stick to one?
If more than one required, can you please comment what is used for what,
and perhaps rename macros in a way that it is clear which one to use
where (as I assume one is specific to base files).
<...>
> +/**
> + * rnp_write_mbx_pf - Places a message in the mailbox
> + * @hw: pointer to the HW structure
> + * @msg: The message buffer
> + * @size: Length of buffer
> + * @mbx_id: the VF index
> + *
> + * returns SUCCESS if it successfully copied message into the buffer
> + **/
> +static int32_t rnp_write_mbx_pf(struct rnp_hw *hw, u32 *msg,
> + u16 size, enum MBX_ID mbx_id)
> +{
> + u32 DATA_REG = (mbx_id == MBX_CM3CPU) ?
> + CPU_PF_SHM_DATA : PF_VF_SHM_DATA(mbx_id);
> + u32 CTRL_REG = (mbx_id == MBX_CM3CPU) ?
> + PF2CPU_MBOX_CTRL : PF2VF_MBOX_CTRL(mbx_id);
> + int32_t ret_val = 0;
> + u32 stat __rte_unused;
> + u16 i;
> +
> + if (size > RNP_VFMAILBOX_SIZE) {
> + RNP_PMD_LOG(ERR, "%s: size:%d should <%d\n", __func__,
> + size, RNP_VFMAILBOX_SIZE);
> + return -EINVAL;
> + }
> +
> + /* lock the mailbox to prevent pf/vf/cpu race condition */
> + ret_val = rnp_obtain_mbx_lock_pf(hw, mbx_id);
> + if (ret_val) {
> + RNP_PMD_LOG(WARNING, "PF[%d] Can't Get Mbx-Lock Try Again\n",
> + hw->function);
> + return ret_val;
> + }
> +
> + /* copy the caller specified message to the mailbox memory buffer */
> + for (i = 0; i < size; i++) {
> +#ifdef MBX_WR_DEBUG
> + mbx_pwr32(hw, DATA_REG + i * 4, msg[i]);
> +#else
> + mbx_wr32(hw, DATA_REG + i * 4, msg[i]);
> +#endif
Who sets this define, if not used please remove it.
^ permalink raw reply [flat|nested] 27+ messages in thread* RE: [PATCH v6 4/8] net/rnp: add mbx basic api feature
2023-09-05 15:45 ` Ferruh Yigit
@ 2023-09-06 10:32 ` 11
0 siblings, 0 replies; 27+ messages in thread
From: 11 @ 2023-09-06 10:32 UTC (permalink / raw)
To: 'Ferruh Yigit'
Cc: dev, thomas, andrew.rybchenko, yaojun, 'Stephen Hemminger'
Hi Ferruh,
Please see the below comment :)
Regards Wenbo
> -----Original Message-----
> From: Ferruh Yigit <ferruh.yigit@amd.com>
> Sent: 2023年9月5日 23:45
> To: Wenbo Cao <caowenbo@mucse.com>
> Cc: dev@dpdk.org; thomas@monjalon.net; andrew.rybchenko@oktetlabs.ru;
> yaojun@mucse.com; Stephen Hemminger <stephen@networkplumber.org>
> Subject: Re: [PATCH v6 4/8] net/rnp: add mbx basic api feature
>
> On 9/1/2023 3:30 AM, Wenbo Cao wrote:
> > mbx base code is for communicate with the firmware
> >
>
> I assume mbx is mailbox, can you please use full name in the patch title and
> commit log.
>
Yes, mbx is mailbox.
I will change the commit log
> How the parties get notified about new messages, is the interrupt support
> enabled or polling, can you please elabore the support in the commit log?
>
> > Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
> > Suggested-by: Stephen Hemminger <stephen@networkplumber.org>
> > ---
> > drivers/net/rnp/base/rnp_api.c | 23 ++
> > drivers/net/rnp/base/rnp_api.h | 7 +
> > drivers/net/rnp/base/rnp_cfg.h | 7 +
> > drivers/net/rnp/base/rnp_dma_regs.h | 73 ++++
> > drivers/net/rnp/base/rnp_eth_regs.h | 124 +++++++
> > drivers/net/rnp/base/rnp_hw.h | 112 +++++-
>
> Why there is no meson file in base folder?
>
For base folder also need meson.build ?
I have add the base/.c in the rnp/meson.build.
Do I need to split it to the base/meson.build ?
> > drivers/net/rnp/meson.build | 1 +
> > drivers/net/rnp/rnp.h | 35 ++
> > drivers/net/rnp/rnp_ethdev.c | 70 +++-
> > drivers/net/rnp/rnp_logs.h | 9 +
> > drivers/net/rnp/rnp_mbx.c | 522 ++++++++++++++++++++++++++++
> > drivers/net/rnp/rnp_mbx.h | 139 ++++++++
> > drivers/net/rnp/rnp_mbx_fw.c | 271 +++++++++++++++
> > drivers/net/rnp/rnp_mbx_fw.h | 22 ++
>
> Should 'rnp_mbx*' files also go under base folder? I can see they use some 'u8'
> like typedef, making me think they are shared between other platforms, if so base
> folder suits better.
>
Yes, the mbx.c or rnp_mbx_fw.c is share code, need change to the base folder.
> > 14 files changed, 1412 insertions(+), 3 deletions(-) create mode
> > 100644 drivers/net/rnp/base/rnp_api.c create mode 100644
> > drivers/net/rnp/base/rnp_api.h create mode 100644
> > drivers/net/rnp/base/rnp_cfg.h create mode 100644
> > drivers/net/rnp/base/rnp_dma_regs.h
> > create mode 100644 drivers/net/rnp/base/rnp_eth_regs.h
> > create mode 100644 drivers/net/rnp/rnp_mbx.c create mode 100644
> > drivers/net/rnp/rnp_mbx.h create mode 100644
> > drivers/net/rnp/rnp_mbx_fw.c create mode 100644
> > drivers/net/rnp/rnp_mbx_fw.h
> >
> > diff --git a/drivers/net/rnp/base/rnp_api.c
> > b/drivers/net/rnp/base/rnp_api.c new file mode 100644 index
> > 0000000000..550da6217d
> > --- /dev/null
> > +++ b/drivers/net/rnp/base/rnp_api.c
> > @@ -0,0 +1,23 @@
> > +#include "rnp.h"
> > +#include "rnp_api.h"
> > +
> > +int
> > +rnp_init_hw(struct rte_eth_dev *dev)
> > +{
> >
>
> Base code is mostly for works as HAL and you don't need to pass "struct
> rte_eth_dev" in this level, but mostly adapte (struct rnp_eth_adapter) or hw
> (struct rnp_hw) should be sufficient.
>
Because of one-pcie-bdf have multiple-port, the hw only manage by the adapter struct
But the adapter struct and hw struct don't have port info(0,1,2,3).
For this problem I can change 'rte_eth_dev' to struct rnp_eth_port'
> I assume you are passing "struct rte_eth_dev" because "struct rnp_mac_api" is
> stored in 'process_private', this part I am not clear.
>
> Why need 'mac_ops' pointers for secondary process?
> Primary process should be doing all hw initialization, otherwise both primary and
> secondar(y|ies) trying to configure the MAC can cause unexpected results.
>
For secondary process support to set, I find refer driver don't have limit for
'dev->dev_ops' set in SECONDARY mode.
.promiscuous_enable
.promiscuous_disable
.allmulticast_enable
.allmulticast_disable
.fw_version_get
> Seconday process is only for datapath, which can access to the queues and do
> packet processing.
>
For secondary process support to set, I find refer driver don't have limit for
'dev->dev_ops' set in SECONDARY mode.
User can set api , if 'dev->dev_ops' set don’t limit.
> Needing "mac_ops" in the secondary can be bad design desicion, can you please
> double check it?
>
> <...>
>
I refer to some exist driver, it is not special limit for SECONDARY process for 'dev->dev_ops' set.
So when I test some code fond that some api will be cause crash such as 'get_module_eeprom'
And I find this is because of function point api, the secondary process use primary function point api address.
The address for secondary is can't reach. So I try to move the mac_ops, mbx_ops to the process_prive to
resolved this problem. Primary and secondary will have it own function point api address.
> > diff --git a/drivers/net/rnp/rnp.h b/drivers/net/rnp/rnp.h index
> > cab1b8e85d..6e12885877 100644
> > --- a/drivers/net/rnp/rnp.h
> > +++ b/drivers/net/rnp/rnp.h
> > @@ -3,6 +3,7 @@
> > */
> > #ifndef __RNP_H__
> > #define __RNP_H__
> > +#include <rte_log.h>
> >
> > #include "base/rnp_hw.h"
> >
> > @@ -14,14 +15,17 @@
> >
> > struct rnp_eth_port {
> > struct rnp_eth_adapter *adapt;
> > + struct rnp_hw *hw;
>
> adapter already has the 'hw', is there a specific reason not to access it via 'port-
> >adapt->hw'
>
> > struct rte_eth_dev *eth_dev;
> > } __rte_cache_aligned;
> >
> > struct rnp_share_ops {
> > + const struct rnp_mbx_api *mbx_api;
> > } __rte_cache_aligned;
> >
> > struct rnp_eth_adapter {
> > struct rnp_hw hw;
> > + uint16_t max_vfs;
> > struct rte_pci_device *pdev;
> > struct rte_eth_dev *eth_dev; /* primary eth_dev */
> > struct rnp_eth_port *ports[RNP_MAX_PORT_OF_PF]; @@ -34,5 +38,36
> @@
> > struct rnp_eth_adapter {
> > (((struct rnp_eth_port *)((eth_dev)->data->dev_private)))
> > #define RNP_DEV_TO_ADAPTER(eth_dev) \
> > ((struct rnp_eth_adapter *)(RNP_DEV_TO_PORT(eth_dev)->adapt))
> > +#define RNP_DEV_TO_HW(eth_dev) \
> > + (&((struct rnp_eth_adapter
> > +*)(RNP_DEV_TO_PORT((eth_dev))->adapt))->hw)
> > +#define RNP_DEV_PP_PRIV_TO_MBX_OPS(dev) \
> > + (((struct rnp_share_ops *)(dev)->process_private)->mbx_api)
> > +#define RNP_DEV_TO_MBX_OPS(dev) RNP_DEV_PP_PRIV_TO_MBX_OPS(dev)
> >
> > +static inline void rnp_reg_offset_init(struct rnp_hw *hw) {
> > + uint16_t i;
> > +
> > + if (hw->device_id == RNP_DEV_ID_N10G && hw->mbx.pf_num) {
> > + hw->iobar4 = (void *)((uint8_t *)hw->iobar4 + 0x100000);
> > + hw->msix_base = (void *)((uint8_t *)hw->iobar4 + 0xa4000);
> > + hw->msix_base = (void *)((uint8_t *)hw->msix_base + 0x200);
> > + } else {
> > + hw->msix_base = (void *)((uint8_t *)hw->iobar4 + 0xa4000);
> > + }
> > + /* === dma status/config====== */
> > + hw->link_sync = (void *)((uint8_t *)hw->iobar4 + 0x000c);
> > + hw->dma_axi_en = (void *)((uint8_t *)hw->iobar4 + 0x0010);
> > + hw->dma_axi_st = (void *)((uint8_t *)hw->iobar4 + 0x0014);
> > +
> > + if (hw->mbx.pf_num)
> > + hw->msix_base = (void *)((uint8_t *)0x200);
> > + /* === queue registers === */
> > + hw->dma_base = (void *)((uint8_t *)hw->iobar4 + 0x08000);
> > + hw->veb_base = (void *)((uint8_t *)hw->iobar4 + 0x0);
> > + hw->eth_base = (void *)((uint8_t *)hw->iobar4 + 0x10000);
>
> There are lots of hardcoded values in this function, can you make them macros?
>
Yes, I can change it to macros.
> > + /* mac */
> > + for (i = 0; i < RNP_MAX_HW_PORT_PERR_PF; i++)
> > + hw->mac_base[i] = (void *)((uint8_t *)hw->iobar4 + 0x60000 +
> > +0x10000 * i); }
> > #endif /* __RNP_H__ */
> > diff --git a/drivers/net/rnp/rnp_ethdev.c
> > b/drivers/net/rnp/rnp_ethdev.c index ae737643a7..a2dc27548a 100644
> > --- a/drivers/net/rnp/rnp_ethdev.c
> > +++ b/drivers/net/rnp/rnp_ethdev.c
> > @@ -8,6 +8,7 @@
> > #include <ethdev_driver.h>
> >
> > #include "rnp.h"
> > +#include "rnp_mbx.h"
> > #include "rnp_logs.h"
> >
> > static int
> > @@ -89,6 +90,58 @@ rnp_alloc_eth_port(struct rte_pci_device *primary_pci,
> char *name)
> > return NULL;
> > }
> >
> > +static void rnp_get_nic_attr(struct rnp_eth_adapter *adapter) {
> > + RTE_SET_USED(adapter);
> > +}
> > +
> > +static int
> > +rnp_process_resource_init(struct rte_eth_dev *eth_dev) {
> > + struct rnp_share_ops *share_priv;
> > +
> > + /* allocate process_private memory this must can't
> > + * belone to the dpdk mem resource manager
> > + * such as from rte_malloc or rte_dma_zone..
> > + */
> > + /* use the process_prive point to resolve secondary process
> > + * use point-func. This point is per process will be safe to cover.
> > + * This will cause secondary process core-dump because of IPC
> > + * Secondary will call primary process point func virt-address
> > + * secondary process don't alloc user/pmd to alloc or free
> > + * the memory of dpdk-mem resource it will cause hugepage
> > + * mem exception
> > + * be careful for secondary Process to use the share-mem of
> > + * point correlation
> > + */
>
> As commented above, I am not sure why you need this, it should be sufficient to
> have mac_ops in primary proccess.
>
This is because of supporting
.promiscuous_enable
.promiscuous_disable
.allmulticast_enable
.allmulticast_disable
.fw_version_get
In the secondary process
> > + share_priv = calloc(1, sizeof(*share_priv));
> > + if (!share_priv) {
> > + PMD_DRV_LOG(ERR, "calloc share_priv failed");
> > + return -ENOMEM;
> > + }
> > + memset(share_priv, 0, sizeof(*share_priv));
> > + eth_dev->process_private = share_priv;
> > +
> > + return 0;
> > +}
> > +
> > +static void
> > +rnp_common_ops_init(struct rnp_eth_adapter *adapter) {
> > + struct rnp_share_ops *share_priv;
> > +
> > + share_priv = adapter->share_priv;
> > + share_priv->mbx_api = &rnp_mbx_pf_ops; }
> > +
> > +static int
> > +rnp_special_ops_init(struct rte_eth_dev *eth_dev) {
> > + RTE_SET_USED(eth_dev);
> > +
> > + return 0;
> > +}
> > +
> > static int
> > rnp_eth_dev_init(struct rte_eth_dev *dev) { @@ -124,6 +177,20 @@
> > rnp_eth_dev_init(struct rte_eth_dev *dev)
> > hw->device_id = pci_dev->id.device_id;
> > hw->vendor_id = pci_dev->id.vendor_id;
> > hw->device_id = pci_dev->id.device_id;
> > + adapter->max_vfs = pci_dev->max_vfs;
> > + ret = rnp_process_resource_init(dev);
> > + if (ret) {
> > + PMD_DRV_LOG(ERR, "share prive resource init failed");
> > + return ret;
> > + }
> > + adapter->share_priv = dev->process_private;
>
> Isn't 'adapter' shared betwen primary and secondary ('port' is dev_private and
> 'port->adapt' is 'adapter'), so updating "adapter->share_priv" defeats the
> purpose of process_private, no?
>
Yes just process private.
dev->process_private; this dev is alloc by 'rte_eth_dev_pci_generic_probe'
So we must record the first port dev->process_private,
when the secondary port of the same PF, alloc by rnp_alloc_eth_port
we can use lase record info 'adapter->share_priv' turn to
'dev->process_private' (alloc by rnp_alloc_eth_port).
> > + rnp_common_ops_init(adapter);
> > + rnp_get_nic_attr(adapter);
> > + /* We need Use Device Id To Change The Resource Mode */
> > + rnp_special_ops_init(dev);
> >
>
> You can add this when it is needed, so in same patch we can see where it is
> called and how it is implemented, right now it is not possible that what 'special'
> init does.
>
>
Yes, it is used to for multiple-port mode and one pcie-bdf select api.
For now, I can remove the code for one pcie-bdf-one-port mode.
> > + port->adapt = adapter;
> > + port->hw = hw;
>
> At this stage port is NULL, probably above needs to go to other patch that sets
> 'port'.
>
Port(dev_private) is get from the first port (bdf) alloc by rte_eth_dev_pci_generic_probe
> > + rnp_init_mbx_ops_pf(hw);
> > for (p_id = 0; p_id < adapter->num_ports; p_id++) {
> > /* port 0 resource has been allocated When Probe */
> > if (!p_id) {
> > @@ -158,11 +225,10 @@ rnp_eth_dev_init(struct rte_eth_dev *dev)
> > continue;
> > if (port->eth_dev) {
> > rnp_dev_close(port->eth_dev);
> > - rte_eth_dev_release_port(port->eth_dev);
> > if (port->eth_dev->process_private)
> > free(port->eth_dev->process_private);
> > + rte_eth_dev_release_port(port->eth_dev);
> > }
> > - rte_free(port);
> > }
> > rte_free(adapter);
> >
> > diff --git a/drivers/net/rnp/rnp_logs.h b/drivers/net/rnp/rnp_logs.h
> > index 1b3ee33745..f1648aabb5 100644
> > --- a/drivers/net/rnp/rnp_logs.h
> > +++ b/drivers/net/rnp/rnp_logs.h
> > @@ -13,6 +13,15 @@ extern int rnp_drv_logtype; #define
> > RNP_PMD_DRV_LOG(level, fmt, args...) \
> > rte_log(RTE_LOG_##level, rnp_drv_logtype, \
> > "%s() " fmt, __func__, ##args)
> > +#define PMD_DRV_LOG_RAW(level, fmt, args...) \
> > + rte_log(RTE_LOG_ ## level, rnp_drv_logtype, "%s(): " fmt, \
> > + __func__, ## args)
> > +#define PMD_DRV_LOG(level, fmt, args...) \
> > + PMD_DRV_LOG_RAW(level, fmt "\n", ## args)
> > +
> > +#define RNP_PMD_LOG(level, fmt, args...) \
> > + rte_log(RTE_LOG_##level, rnp_drv_logtype, \
> > + "rnp_net: (%d) " fmt, __LINE__, ##args)
>
> With these there are three different macros used to log in 'rnp_drv_logtype' type:
>
> RNP_PMD_LOG
> PMD_DRV_LOG
> RNP_PMD_DRV_LOG
>
> 'RNP_PMD_DRV_LOG' and 'PMD_DRV_LOG' looks exact same, do you need all
> these macros, why not stick to one?
>
> PMD_DRV_LOG
This need to remove it.
> If more than one required, can you please comment what is used for what, and
> perhaps rename macros in a way that it is clear which one to use where (as I
> assume one is specific to base files).
>
> <...>
>
Thanks a lot, I will add comment for the macros before it.
> > +/**
> > + * rnp_write_mbx_pf - Places a message in the mailbox
> > + * @hw: pointer to the HW structure
> > + * @msg: The message buffer
> > + * @size: Length of buffer
> > + * @mbx_id: the VF index
> > + *
> > + * returns SUCCESS if it successfully copied message into the buffer
> > +**/ static int32_t rnp_write_mbx_pf(struct rnp_hw *hw, u32 *msg,
> > + u16 size, enum MBX_ID mbx_id)
> > +{
> > + u32 DATA_REG = (mbx_id == MBX_CM3CPU) ?
> > + CPU_PF_SHM_DATA : PF_VF_SHM_DATA(mbx_id);
> > + u32 CTRL_REG = (mbx_id == MBX_CM3CPU) ?
> > + PF2CPU_MBOX_CTRL : PF2VF_MBOX_CTRL(mbx_id);
> > + int32_t ret_val = 0;
> > + u32 stat __rte_unused;
> > + u16 i;
> > +
> > + if (size > RNP_VFMAILBOX_SIZE) {
> > + RNP_PMD_LOG(ERR, "%s: size:%d should <%d\n", __func__,
> > + size, RNP_VFMAILBOX_SIZE);
> > + return -EINVAL;
> > + }
> > +
> > + /* lock the mailbox to prevent pf/vf/cpu race condition */
> > + ret_val = rnp_obtain_mbx_lock_pf(hw, mbx_id);
> > + if (ret_val) {
> > + RNP_PMD_LOG(WARNING, "PF[%d] Can't Get Mbx-Lock Try
> Again\n",
> > + hw->function);
> > + return ret_val;
> > + }
> > +
> > + /* copy the caller specified message to the mailbox memory buffer */
> > + for (i = 0; i < size; i++) {
> > +#ifdef MBX_WR_DEBUG
> > + mbx_pwr32(hw, DATA_REG + i * 4, msg[i]); #else
> > + mbx_wr32(hw, DATA_REG + i * 4, msg[i]); #endif
>
> Who sets this define, if not used please remove it.
>
This need to remove 😊.
^ permalink raw reply [flat|nested] 27+ messages in thread
* [PATCH v6 5/8] net/rnp add reset code for Chip Init process
2023-09-01 2:30 [PATCH v6 0/8] [v6]drivers/net Add Support mucse N10 Pmd Driver Wenbo Cao
` (3 preceding siblings ...)
2023-09-01 2:30 ` [PATCH v6 4/8] net/rnp: add mbx basic api feature Wenbo Cao
@ 2023-09-01 2:30 ` Wenbo Cao
2023-09-05 15:46 ` Ferruh Yigit
2023-09-01 2:30 ` [PATCH v6 6/8] net/rnp add port info resource init Wenbo Cao
` (3 subsequent siblings)
8 siblings, 1 reply; 27+ messages in thread
From: Wenbo Cao @ 2023-09-01 2:30 UTC (permalink / raw)
To: Wenbo Cao; +Cc: dev, ferruh.yigit, thomas, andrew.rybchenko, yaojun
we must get the shape info of nic from Firmware for
reset. so the related codes is first get firmware info
and then reset the chip
Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
---
drivers/net/rnp/base/rnp_hw.h | 56 +++++++++++-
drivers/net/rnp/meson.build | 3 +
drivers/net/rnp/rnp.h | 27 ++++++
drivers/net/rnp/rnp_ethdev.c | 93 ++++++++++++++++++-
drivers/net/rnp/rnp_mbx_fw.h | 163 +++++++++++++++++++++++++++++++++-
5 files changed, 339 insertions(+), 3 deletions(-)
diff --git a/drivers/net/rnp/base/rnp_hw.h b/drivers/net/rnp/base/rnp_hw.h
index 1db966cf21..57b7dc75a0 100644
--- a/drivers/net/rnp/base/rnp_hw.h
+++ b/drivers/net/rnp/base/rnp_hw.h
@@ -8,6 +8,9 @@
#include <ethdev_driver.h>
#include "rnp_osdep.h"
+#include "rnp_dma_regs.h"
+#include "rnp_eth_regs.h"
+#include "rnp_cfg.h"
static inline unsigned int rnp_rd_reg(volatile void *addr)
{
@@ -29,7 +32,18 @@ static inline void rnp_wr_reg(volatile void *reg, int val)
rnp_rd_reg((uint8_t *)(_base) + (_off))
#define rnp_io_wr(_base, _off, _val) \
rnp_wr_reg((uint8_t *)(_base) + (_off), (_val))
-
+#define rnp_eth_rd(_hw, _off) \
+ rnp_rd_reg((uint8_t *)((_hw)->eth_base) + (_off))
+#define rnp_eth_wr(_hw, _off, _val) \
+ rnp_wr_reg((uint8_t *)((_hw)->eth_base) + (_off), (_val))
+#define rnp_dma_rd(_hw, _off) \
+ rnp_rd_reg((uint8_t *)((_hw)->dma_base) + (_off))
+#define rnp_dma_wr(_hw, _off, _val) \
+ rnp_wr_reg((uint8_t *)((_hw)->dma_base) + (_off), (_val))
+#define rnp_top_rd(_hw, _off) \
+ rnp_rd_reg((uint8_t *)((_hw)->comm_reg_base) + (_off))
+#define rnp_top_wr(_hw, _off, _val) \
+ rnp_wr_reg((uint8_t *)((_hw)->comm_reg_base) + (_off), (_val))
struct rnp_hw;
/* Mbx Operate info */
enum MBX_ID {
@@ -98,6 +112,17 @@ struct rnp_mbx_info {
rte_atomic16_t state;
} __rte_cache_aligned;
+struct rnp_mac_api {
+ int32_t (*init_hw)(struct rnp_hw *hw);
+ int32_t (*reset_hw)(struct rnp_hw *hw);
+};
+
+struct rnp_mac_info {
+ uint8_t assign_addr[RTE_ETHER_ADDR_LEN];
+ uint8_t set_addr[RTE_ETHER_ADDR_LEN];
+ struct rnp_mac_api ops;
+} __rte_cache_aligned;
+
struct rnp_eth_adapter;
#define RNP_MAX_HW_PORT_PERR_PF (4)
struct rnp_hw {
@@ -111,8 +136,10 @@ struct rnp_hw {
void *eth_base;
void *veb_base;
void *mac_base[RNP_MAX_HW_PORT_PERR_PF];
+ void *comm_reg_base;
void *msix_base;
/* === dma == */
+ void *dev_version;
void *dma_axi_en;
void *dma_axi_st;
@@ -120,10 +147,37 @@ struct rnp_hw {
uint16_t vendor_id;
uint16_t function;
uint16_t pf_vf_num;
+ int pfvfnum;
uint16_t max_vfs;
+
+ bool ncsi_en;
+ uint8_t ncsi_rar_entries;
+
+ int sgmii_phy_id;
+ int is_sgmii;
+ u16 phy_type;
+ uint8_t force_10g_1g_speed_ablity;
+ uint8_t force_speed_stat;
+#define FORCE_SPEED_STAT_DISABLED (0)
+#define FORCE_SPEED_STAT_1G (1)
+#define FORCE_SPEED_STAT_10G (2)
+ uint32_t speed;
+ unsigned int axi_mhz;
+
+ int fw_version; /* Primary FW Version */
+ uint32_t fw_uid; /* Subclass Fw Version */
+
+ int nic_mode;
+ unsigned char lane_mask;
+ int lane_of_port[4];
+ char phy_port_ids[4]; /* port id: for lane0~3: value: 0 ~ 7 */
+ uint8_t max_port_num; /* Max Port Num This PF Have */
+
void *cookie_pool;
char cookie_p_name[RTE_MEMZONE_NAMESIZE];
+ struct rnp_mac_info mac;
struct rnp_mbx_info mbx;
+ rte_spinlock_t fw_lock;
} __rte_cache_aligned;
#endif /* __RNP_H__*/
diff --git a/drivers/net/rnp/meson.build b/drivers/net/rnp/meson.build
index 60bba486fc..855c894032 100644
--- a/drivers/net/rnp/meson.build
+++ b/drivers/net/rnp/meson.build
@@ -9,5 +9,8 @@ endif
sources = files(
'rnp_ethdev.c',
'rnp_mbx.c',
+ 'rnp_mbx_fw.c',
+ 'base/rnp_api.c',
)
+
includes += include_directories('base')
diff --git a/drivers/net/rnp/rnp.h b/drivers/net/rnp/rnp.h
index 6e12885877..45638aae5b 100644
--- a/drivers/net/rnp/rnp.h
+++ b/drivers/net/rnp/rnp.h
@@ -13,6 +13,20 @@
#define RNP_CFG_BAR (4)
#define RNP_PF_INFO_BAR (0)
+enum rnp_resource_share_m {
+ RNP_SHARE_CORPORATE = 0,
+ RNP_SHARE_INDEPENDENT,
+};
+/*
+ * Structure to store private data for each driver instance (for each port).
+ */
+enum rnp_work_mode {
+ RNP_SINGLE_40G = 0,
+ RNP_SINGLE_10G = 1,
+ RNP_DUAL_10G = 2,
+ RNP_QUAD_10G = 3,
+};
+
struct rnp_eth_port {
struct rnp_eth_adapter *adapt;
struct rnp_hw *hw;
@@ -21,9 +35,12 @@ struct rnp_eth_port {
struct rnp_share_ops {
const struct rnp_mbx_api *mbx_api;
+ const struct rnp_mac_api *mac_api;
} __rte_cache_aligned;
struct rnp_eth_adapter {
+ enum rnp_work_mode mode;
+ enum rnp_resource_share_m s_mode; /* Port Resource Share Policy */
struct rnp_hw hw;
uint16_t max_vfs;
struct rte_pci_device *pdev;
@@ -31,7 +48,9 @@ struct rnp_eth_adapter {
struct rnp_eth_port *ports[RNP_MAX_PORT_OF_PF];
struct rnp_share_ops *share_priv;
+ int max_link_speed;
uint8_t num_ports; /* Cur Pf Has physical Port Num */
+ uint8_t lane_mask;
} __rte_cache_aligned;
#define RNP_DEV_TO_PORT(eth_dev) \
@@ -40,9 +59,14 @@ struct rnp_eth_adapter {
((struct rnp_eth_adapter *)(RNP_DEV_TO_PORT(eth_dev)->adapt))
#define RNP_DEV_TO_HW(eth_dev) \
(&((struct rnp_eth_adapter *)(RNP_DEV_TO_PORT((eth_dev))->adapt))->hw)
+#define RNP_HW_TO_ADAPTER(hw) \
+ ((struct rnp_eth_adapter *)((hw)->back))
#define RNP_DEV_PP_PRIV_TO_MBX_OPS(dev) \
(((struct rnp_share_ops *)(dev)->process_private)->mbx_api)
#define RNP_DEV_TO_MBX_OPS(dev) RNP_DEV_PP_PRIV_TO_MBX_OPS(dev)
+#define RNP_DEV_PP_PRIV_TO_MAC_OPS(dev) \
+ (((struct rnp_share_ops *)(dev)->process_private)->mac_api)
+#define RNP_DEV_TO_MAC_OPS(dev) RNP_DEV_PP_PRIV_TO_MAC_OPS(dev)
static inline void rnp_reg_offset_init(struct rnp_hw *hw)
{
@@ -56,6 +80,7 @@ static inline void rnp_reg_offset_init(struct rnp_hw *hw)
hw->msix_base = (void *)((uint8_t *)hw->iobar4 + 0xa4000);
}
/* === dma status/config====== */
+ hw->dev_version = (void *)((uint8_t *)hw->iobar4 + 0x0000);
hw->link_sync = (void *)((uint8_t *)hw->iobar4 + 0x000c);
hw->dma_axi_en = (void *)((uint8_t *)hw->iobar4 + 0x0010);
hw->dma_axi_st = (void *)((uint8_t *)hw->iobar4 + 0x0014);
@@ -69,5 +94,7 @@ static inline void rnp_reg_offset_init(struct rnp_hw *hw)
/* mac */
for (i = 0; i < RNP_MAX_HW_PORT_PERR_PF; i++)
hw->mac_base[i] = (void *)((uint8_t *)hw->iobar4 + 0x60000 + 0x10000 * i);
+ /* === top reg === */
+ hw->comm_reg_base = (void *)((uint8_t *)hw->iobar4 + 0x30000);
}
#endif /* __RNP_H__ */
diff --git a/drivers/net/rnp/rnp_ethdev.c b/drivers/net/rnp/rnp_ethdev.c
index a2dc27548a..8bb4fd5963 100644
--- a/drivers/net/rnp/rnp_ethdev.c
+++ b/drivers/net/rnp/rnp_ethdev.c
@@ -8,7 +8,9 @@
#include <ethdev_driver.h>
#include "rnp.h"
+#include "rnp_api.h"
#include "rnp_mbx.h"
+#include "rnp_mbx_fw.h"
#include "rnp_logs.h"
static int
@@ -92,7 +94,30 @@ rnp_alloc_eth_port(struct rte_pci_device *primary_pci, char *name)
static void rnp_get_nic_attr(struct rnp_eth_adapter *adapter)
{
- RTE_SET_USED(adapter);
+ struct rnp_hw *hw = &adapter->hw;
+ int lane_mask = 0, err, mode = 0;
+
+ rnp_mbx_link_event_enable(adapter->eth_dev, false);
+
+ err = rnp_mbx_get_capability(adapter->eth_dev, &lane_mask, &mode);
+ if (err < 0 || !lane_mask) {
+ PMD_DRV_LOG(ERR, "%s: mbx_get_capability error! errcode=%d\n",
+ __func__, hw->speed);
+ return;
+ }
+
+ adapter->num_ports = __builtin_popcount(lane_mask);
+ adapter->max_link_speed = hw->speed;
+ adapter->lane_mask = lane_mask;
+ adapter->mode = hw->nic_mode;
+
+ PMD_DRV_LOG(INFO, "max link speed:%d lane_mask:0x%x nic-mode:0x%x\n",
+ (int)adapter->max_link_speed,
+ (int)adapter->num_ports, adapter->mode);
+ if (adapter->num_ports && adapter->num_ports == 1)
+ adapter->s_mode = RNP_SHARE_CORPORATE;
+ else
+ adapter->s_mode = RNP_SHARE_INDEPENDENT;
}
static int
@@ -125,6 +150,72 @@ rnp_process_resource_init(struct rte_eth_dev *eth_dev)
return 0;
}
+static int32_t rnp_init_hw_pf(struct rnp_hw *hw)
+{
+ struct rnp_eth_adapter *adapter = RNP_HW_TO_ADAPTER(hw);
+ uint32_t version;
+ uint32_t reg;
+
+ PMD_INIT_FUNC_TRACE();
+ version = rnp_rd_reg(hw->dev_version);
+ PMD_DRV_LOG(INFO, "NIC HW Version:0x%.2x\n", version);
+
+ /* Disable Rx/Tx Dma */
+ rnp_wr_reg(hw->dma_axi_en, false);
+ /* Check Dma Chanle Status */
+ while (rnp_rd_reg(hw->dma_axi_st) == 0)
+ ;
+
+ /* Reset Nic All Hardware */
+ if (rnp_reset_hw(adapter->eth_dev, hw))
+ return -EPERM;
+
+ /* Rx Proto Offload No-BYPASS */
+ rnp_eth_wr(hw, RNP_ETH_ENGINE_BYPASS, false);
+ /* Enable Flow Filter Engine */
+ rnp_eth_wr(hw, RNP_HOST_FILTER_EN, true);
+ /* Enable VXLAN Parse */
+ rnp_eth_wr(hw, RNP_EN_TUNNEL_VXLAN_PARSE, true);
+ /* Enabled REDIR ACTION */
+ rnp_eth_wr(hw, RNP_REDIR_CTRL, true);
+
+ /* Setup Scatter DMA Mem Size */
+ reg = ((RTE_ETHER_MAX_LEN / 16) << RNP_DMA_SCATTER_MEM_SHIFT);
+ rnp_dma_wr(hw, RNP_DMA_CTRL, reg);
+#ifdef PHYTIUM_SUPPORT
+#define RNP_DMA_PADDING (1 << 8)
+ reg = rnp_dma_rd(hw, RNP_DMA_CTRL);
+ reg |= RNP_DMA_PADDING;
+ rnp_dma_wr(hw, RNP_DMA_CTRL, reg);
+#endif
+ /* Enable Rx/Tx Dma */
+ rnp_wr_reg(hw->dma_axi_en, 0b1111);
+
+ rnp_top_wr(hw, RNP_TX_QINQ_WORKAROUND, 1);
+
+ return 0;
+}
+
+static int32_t rnp_reset_hw_pf(struct rnp_hw *hw)
+{
+ struct rnp_eth_adapter *adapter = hw->back;
+
+ rnp_top_wr(hw, RNP_NIC_RESET, 0);
+ rte_wmb();
+ rnp_top_wr(hw, RNP_NIC_RESET, 1);
+
+ rnp_mbx_fw_reset_phy(adapter->eth_dev);
+
+ PMD_DRV_LOG(INFO, "PF[%d] reset nic finish\n",
+ hw->function);
+ return 0;
+}
+
+const struct rnp_mac_api rnp_mac_ops = {
+ .reset_hw = rnp_reset_hw_pf,
+ .init_hw = rnp_init_hw_pf
+};
+
static void
rnp_common_ops_init(struct rnp_eth_adapter *adapter)
{
diff --git a/drivers/net/rnp/rnp_mbx_fw.h b/drivers/net/rnp/rnp_mbx_fw.h
index 439090b5a3..f842639c86 100644
--- a/drivers/net/rnp/rnp_mbx_fw.h
+++ b/drivers/net/rnp/rnp_mbx_fw.h
@@ -16,7 +16,168 @@ struct mbx_req_cookie {
int priv_len;
char priv[RNP_MAX_SHARE_MEM];
};
+enum GENERIC_CMD {
+ /* link configuration admin commands */
+ GET_PHY_ABALITY = 0x0601,
+ RESET_PHY = 0x0603,
+ SET_EVENT_MASK = 0x0613,
+};
+
+enum link_event_mask {
+ EVT_LINK_UP = 1,
+ EVT_NO_MEDIA = 2,
+ EVT_LINK_FAULT = 3,
+ EVT_PHY_TEMP_ALARM = 4,
+ EVT_EXCESSIVE_ERRORS = 5,
+ EVT_SIGNAL_DETECT = 6,
+ EVT_AUTO_NEGOTIATION_DONE = 7,
+ EVT_MODULE_QUALIFICATION_FAILED = 8,
+ EVT_PORT_TX_SUSPEND = 9,
+};
+
+enum pma_type {
+ PHY_TYPE_NONE = 0,
+ PHY_TYPE_1G_BASE_KX,
+ PHY_TYPE_SGMII,
+ PHY_TYPE_10G_BASE_KR,
+ PHY_TYPE_25G_BASE_KR,
+ PHY_TYPE_40G_BASE_KR4,
+ PHY_TYPE_10G_BASE_SR,
+ PHY_TYPE_40G_BASE_SR4,
+ PHY_TYPE_40G_BASE_CR4,
+ PHY_TYPE_40G_BASE_LR4,
+ PHY_TYPE_10G_BASE_LR,
+ PHY_TYPE_10G_BASE_ER,
+};
+
+struct phy_abilities {
+ unsigned char link_stat;
+ unsigned char lane_mask;
+
+ int speed;
+ short phy_type;
+ short nic_mode;
+ short pfnum;
+ unsigned int fw_version;
+ unsigned int axi_mhz;
+ uint8_t port_ids[4];
+ uint32_t fw_uid;
+ uint32_t phy_id;
+
+ int wol_status;
+
+ union {
+ unsigned int ext_ablity;
+ struct {
+ unsigned int valid : 1;
+ unsigned int wol_en : 1;
+ unsigned int pci_preset_runtime_en : 1;
+ unsigned int smbus_en : 1;
+ unsigned int ncsi_en : 1;
+ unsigned int rpu_en : 1;
+ unsigned int v2 : 1;
+ unsigned int pxe_en : 1;
+ unsigned int mctp_en : 1;
+ } e;
+ };
+} __rte_packed __rte_aligned(4);
+
+/* firmware -> driver */
struct mbx_fw_cmd_reply {
-} __rte_cache_aligned;
+ /* fw must set: DD, CMP, Error(if error), copy value */
+ unsigned short flags;
+ /* from command: LB,RD,VFC,BUF,SI,EI,FE */
+ unsigned short opcode; /* 2-3: copy from req */
+ unsigned short error_code; /* 4-5: 0 if no error */
+ unsigned short datalen; /* 6-7: */
+ union {
+ struct {
+ unsigned int cookie_lo; /* 8-11: */
+ unsigned int cookie_hi; /* 12-15: */
+ };
+ void *cookie;
+ };
+ /* ===== data ==== [16-64] */
+ union {
+ struct phy_abilities phy_abilities;
+ };
+} __rte_packed __rte_aligned(4);
+
+#define MBX_REQ_HDR_LEN 24
+/* driver -> firmware */
+struct mbx_fw_cmd_req {
+ unsigned short flags; /* 0-1 */
+ unsigned short opcode; /* 2-3 enum LINK_ADM_CMD */
+ unsigned short datalen; /* 4-5 */
+ unsigned short ret_value; /* 6-7 */
+ union {
+ struct {
+ unsigned int cookie_lo; /* 8-11 */
+ unsigned int cookie_hi; /* 12-15 */
+ };
+ void *cookie;
+ };
+ unsigned int reply_lo; /* 16-19 5dw */
+ unsigned int reply_hi; /* 20-23 */
+ /* === data === [24-64] 7dw */
+ union {
+ struct {
+ int requester;
+#define REQUEST_BY_DPDK 0xa1
+#define REQUEST_BY_DRV 0xa2
+#define REQUEST_BY_PXE 0xa3
+ } get_phy_ablity;
+
+ struct {
+ unsigned short enable_stat;
+ unsigned short event_mask; /* enum link_event_mask */
+ } stat_event_mask;
+ };
+} __rte_packed __rte_aligned(4);
+
+static inline void
+build_phy_abalities_req(struct mbx_fw_cmd_req *req, void *cookie)
+{
+ req->flags = 0;
+ req->opcode = GET_PHY_ABALITY;
+ req->datalen = 0;
+ req->reply_lo = 0;
+ req->reply_hi = 0;
+ req->cookie = cookie;
+}
+
+/* enum link_event_mask or */
+static inline void
+build_link_set_event_mask(struct mbx_fw_cmd_req *req,
+ unsigned short event_mask,
+ unsigned short enable,
+ void *cookie)
+{
+ req->flags = 0;
+ req->opcode = SET_EVENT_MASK;
+ req->datalen = sizeof(req->stat_event_mask);
+ req->cookie = cookie;
+ req->reply_lo = 0;
+ req->reply_hi = 0;
+ req->stat_event_mask.event_mask = event_mask;
+ req->stat_event_mask.enable_stat = enable;
+}
+
+static inline void
+build_reset_phy_req(struct mbx_fw_cmd_req *req,
+ void *cookie)
+{
+ req->flags = 0;
+ req->opcode = RESET_PHY;
+ req->datalen = 0;
+ req->reply_lo = 0;
+ req->reply_hi = 0;
+ req->cookie = cookie;
+}
+int rnp_mbx_get_capability(struct rte_eth_dev *dev,
+ int *lane_mask,
+ int *nic_mode);
+int rnp_mbx_link_event_enable(struct rte_eth_dev *dev, int enable);
+int rnp_mbx_fw_reset_phy(struct rte_eth_dev *dev);
#endif /* __RNP_MBX_FW_H__*/
--
2.27.0
^ permalink raw reply related [flat|nested] 27+ messages in thread* Re: [PATCH v6 5/8] net/rnp add reset code for Chip Init process
2023-09-01 2:30 ` [PATCH v6 5/8] net/rnp add reset code for Chip Init process Wenbo Cao
@ 2023-09-05 15:46 ` Ferruh Yigit
2023-09-06 9:23 ` 11
0 siblings, 1 reply; 27+ messages in thread
From: Ferruh Yigit @ 2023-09-05 15:46 UTC (permalink / raw)
To: Wenbo Cao; +Cc: dev, thomas, andrew.rybchenko, yaojun
On 9/1/2023 3:30 AM, Wenbo Cao wrote:
> we must get the shape info of nic from Firmware for
What is "shape info"?
> reset. so the related codes is first get firmware info
> and then reset the chip
>
Is there a specific reason to have "Chip Init" start with uppercase, if
not please make all lowercase, also ':' after domain, like "net/rnp:'.
^ permalink raw reply [flat|nested] 27+ messages in thread
* RE: [PATCH v6 5/8] net/rnp add reset code for Chip Init process
2023-09-05 15:46 ` Ferruh Yigit
@ 2023-09-06 9:23 ` 11
0 siblings, 0 replies; 27+ messages in thread
From: 11 @ 2023-09-06 9:23 UTC (permalink / raw)
To: 'Ferruh Yigit'; +Cc: dev, thomas, andrew.rybchenko, yaojun
Hi Ferruh,
Please see the below comment.
Regards Wenbo.
> -----Original Message-----
> From: Ferruh Yigit <ferruh.yigit@amd.com>
> Sent: 2023年9月5日 23:46
> To: Wenbo Cao <caowenbo@mucse.com>
> Cc: dev@dpdk.org; thomas@monjalon.net; andrew.rybchenko@oktetlabs.ru;
> yaojun@mucse.com
> Subject: Re: [PATCH v6 5/8] net/rnp add reset code for Chip Init process
>
> On 9/1/2023 3:30 AM, Wenbo Cao wrote:
> > we must get the shape info of nic from Firmware for
>
> What is "shape info"?
>
For the nic 4/8 port mode or one-pcie-one port, media type fiber or copper, 10G/1G
"shape info" words may be confused.
> > reset. so the related codes is first get firmware info and then reset
> > the chip
> >
>
> Is there a specific reason to have "Chip Init" start with uppercase, if not please
> make all lowercase, also ':' after domain, like "net/rnp:'.
"Chip Init" is don't have any special meaning, will change the to lowercase.
^ permalink raw reply [flat|nested] 27+ messages in thread
* [PATCH v6 6/8] net/rnp add port info resource init
2023-09-01 2:30 [PATCH v6 0/8] [v6]drivers/net Add Support mucse N10 Pmd Driver Wenbo Cao
` (4 preceding siblings ...)
2023-09-01 2:30 ` [PATCH v6 5/8] net/rnp add reset code for Chip Init process Wenbo Cao
@ 2023-09-01 2:30 ` Wenbo Cao
2023-09-05 16:56 ` Ferruh Yigit
2023-09-01 2:30 ` [PATCH v6 7/8] net/rnp add devargs runtime parsing functions Wenbo Cao
` (2 subsequent siblings)
8 siblings, 1 reply; 27+ messages in thread
From: Wenbo Cao @ 2023-09-01 2:30 UTC (permalink / raw)
To: Wenbo Cao; +Cc: dev, ferruh.yigit, thomas, andrew.rybchenko, yaojun
Add Api For FW Mac Info, Port Resoucre info init Code
For Different Shape Of Nic
Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
---
drivers/net/rnp/base/rnp_api.c | 48 +++++++
drivers/net/rnp/base/rnp_api.h | 10 ++
drivers/net/rnp/base/rnp_hw.h | 18 +++
drivers/net/rnp/meson.build | 1 +
drivers/net/rnp/rnp.h | 88 +++++++++++++
drivers/net/rnp/rnp_ethdev.c | 224 +++++++++++++++++++++++++++++++--
drivers/net/rnp/rnp_mbx_fw.c | 112 +++++++++++++++++
drivers/net/rnp/rnp_mbx_fw.h | 115 +++++++++++++++++
drivers/net/rnp/rnp_rxtx.c | 82 ++++++++++++
drivers/net/rnp/rnp_rxtx.h | 14 +++
10 files changed, 705 insertions(+), 7 deletions(-)
create mode 100644 drivers/net/rnp/rnp_rxtx.c
create mode 100644 drivers/net/rnp/rnp_rxtx.h
diff --git a/drivers/net/rnp/base/rnp_api.c b/drivers/net/rnp/base/rnp_api.c
index 550da6217d..cf74769fb6 100644
--- a/drivers/net/rnp/base/rnp_api.c
+++ b/drivers/net/rnp/base/rnp_api.c
@@ -21,3 +21,51 @@ rnp_reset_hw(struct rte_eth_dev *dev, struct rnp_hw *hw)
return ops->reset_hw(hw);
return -EOPNOTSUPP;
}
+
+int
+rnp_get_mac_addr(struct rte_eth_dev *dev, uint8_t *macaddr)
+{
+ struct rnp_eth_port *port = RNP_DEV_TO_PORT(dev);
+ const struct rnp_mac_api *ops = RNP_DEV_TO_MAC_OPS(dev);
+
+ if (!macaddr)
+ return -EINVAL;
+ if (ops->get_mac_addr)
+ return ops->get_mac_addr(port, port->attr.nr_lane, macaddr);
+ return -EOPNOTSUPP;
+}
+
+int
+rnp_set_default_mac(struct rte_eth_dev *dev, uint8_t *mac_addr)
+{
+ const struct rnp_mac_api *ops = RNP_DEV_TO_MAC_OPS(dev);
+ struct rnp_eth_port *port = RNP_DEV_TO_PORT(dev);
+
+ if (ops->set_default_mac)
+ return ops->set_default_mac(port, mac_addr);
+ return -EOPNOTSUPP;
+}
+
+int
+rnp_set_rafb(struct rte_eth_dev *dev, uint8_t *addr,
+ uint8_t vm_pool, uint8_t index)
+{
+ const struct rnp_mac_api *ops = RNP_DEV_TO_MAC_OPS(dev);
+ struct rnp_eth_port *port = RNP_DEV_TO_PORT(dev);
+
+ if (ops->set_rafb)
+ return ops->set_rafb(port, addr, vm_pool, index);
+ return -EOPNOTSUPP;
+}
+
+int
+rnp_clear_rafb(struct rte_eth_dev *dev,
+ uint8_t vm_pool, uint8_t index)
+{
+ const struct rnp_mac_api *ops = RNP_DEV_TO_MAC_OPS(dev);
+ struct rnp_eth_port *port = RNP_DEV_TO_PORT(dev);
+
+ if (ops->clear_rafb)
+ return ops->clear_rafb(port, vm_pool, index);
+ return -EOPNOTSUPP;
+}
diff --git a/drivers/net/rnp/base/rnp_api.h b/drivers/net/rnp/base/rnp_api.h
index df574dab77..b998b11237 100644
--- a/drivers/net/rnp/base/rnp_api.h
+++ b/drivers/net/rnp/base/rnp_api.h
@@ -4,4 +4,14 @@ int
rnp_init_hw(struct rte_eth_dev *dev);
int
rnp_reset_hw(struct rte_eth_dev *dev, struct rnp_hw *hw);
+int
+rnp_get_mac_addr(struct rte_eth_dev *dev, uint8_t *macaddr);
+int
+rnp_set_default_mac(struct rte_eth_dev *dev, uint8_t *mac_addr);
+int
+rnp_set_rafb(struct rte_eth_dev *dev, uint8_t *addr,
+ uint8_t vm_pool, uint8_t index);
+int
+rnp_clear_rafb(struct rte_eth_dev *dev,
+ uint8_t vm_pool, uint8_t index);
#endif /* __RNP_API_H__ */
diff --git a/drivers/net/rnp/base/rnp_hw.h b/drivers/net/rnp/base/rnp_hw.h
index 57b7dc75a0..395b9d5c71 100644
--- a/drivers/net/rnp/base/rnp_hw.h
+++ b/drivers/net/rnp/base/rnp_hw.h
@@ -44,6 +44,10 @@ static inline void rnp_wr_reg(volatile void *reg, int val)
rnp_rd_reg((uint8_t *)((_hw)->comm_reg_base) + (_off))
#define rnp_top_wr(_hw, _off, _val) \
rnp_wr_reg((uint8_t *)((_hw)->comm_reg_base) + (_off), (_val))
+#define RNP_MACADDR_UPDATE_LO(hw, hw_idx, val) \
+ rnp_eth_wr(hw, RNP_RAL_BASE_ADDR(hw_idx), val)
+#define RNP_MACADDR_UPDATE_HI(hw, hw_idx, val) \
+ rnp_eth_wr(hw, RNP_RAH_BASE_ADDR(hw_idx), val)
struct rnp_hw;
/* Mbx Operate info */
enum MBX_ID {
@@ -112,9 +116,23 @@ struct rnp_mbx_info {
rte_atomic16_t state;
} __rte_cache_aligned;
+struct rnp_eth_port;
struct rnp_mac_api {
int32_t (*init_hw)(struct rnp_hw *hw);
int32_t (*reset_hw)(struct rnp_hw *hw);
+ /* MAC Address */
+ int32_t (*get_mac_addr)(struct rnp_eth_port *port,
+ uint8_t lane,
+ uint8_t *macaddr);
+ int32_t (*set_default_mac)(struct rnp_eth_port *port, uint8_t *mac);
+ /* Receive Address Filter Table */
+ int32_t (*set_rafb)(struct rnp_eth_port *port,
+ uint8_t *mac,
+ uint8_t vm_pool,
+ uint8_t index);
+ int32_t (*clear_rafb)(struct rnp_eth_port *port,
+ uint8_t vm_pool,
+ uint8_t index);
};
struct rnp_mac_info {
diff --git a/drivers/net/rnp/meson.build b/drivers/net/rnp/meson.build
index 855c894032..f72815b396 100644
--- a/drivers/net/rnp/meson.build
+++ b/drivers/net/rnp/meson.build
@@ -10,6 +10,7 @@ sources = files(
'rnp_ethdev.c',
'rnp_mbx.c',
'rnp_mbx_fw.c',
+ 'rnp_rxtx.c',
'base/rnp_api.c',
)
diff --git a/drivers/net/rnp/rnp.h b/drivers/net/rnp/rnp.h
index 45638aae5b..9ae801414b 100644
--- a/drivers/net/rnp/rnp.h
+++ b/drivers/net/rnp/rnp.h
@@ -9,14 +9,90 @@
#define PCI_VENDOR_ID_MUCSE (0x8848)
#define RNP_DEV_ID_N10G (0x1000)
+#define RNP_DEV_ID_N400L_X4 (0x1021)
#define RNP_MAX_PORT_OF_PF (4)
#define RNP_CFG_BAR (4)
#define RNP_PF_INFO_BAR (0)
+/* Peer Port Own Independent Resource */
+#define RNP_PORT_MAX_MACADDR (32)
+#define RNP_PORT_MAX_UC_MAC_SIZE (256)
+#define RNP_PORT_MAX_VLAN_HASH (12)
+#define RNP_PORT_MAX_UC_HASH_TB (8)
+
+/* Hardware Resource info */
+#define RNP_MAX_RX_QUEUE_NUM (128)
+#define RNP_MAX_TX_QUEUE_NUM (128)
+#define RNP_N400_MAX_RX_QUEUE_NUM (8)
+#define RNP_N400_MAX_TX_QUEUE_NUM (8)
+#define RNP_MAX_HASH_KEY_SIZE (10)
+#define RNP_MAX_MAC_ADDRS (128)
+#define RNP_MAX_SUPPORT_VF_NUM (64)
+#define RNP_MAX_VFTA_SIZE (128)
+#define RNP_MAX_TC_SUPPORT (4)
+
+#define RNP_MAX_UC_MAC_SIZE (4096) /* Max Num of Unicast MAC addr */
+#define RNP_MAX_UC_HASH_TB (128)
+#define RNP_MAX_MC_MAC_SIZE (4096) /* Max Num of Multicast MAC addr */
+#define RNP_MAC_MC_HASH_TB (128)
+#define RNP_MAX_VLAN_HASH_TB_SIZE (4096)
+
+#define RNP_MAX_UC_HASH_TABLE (128)
+#define RNP_MAC_MC_HASH_TABLE (128)
+#define RNP_UTA_BIT_SHIFT (5)
+
enum rnp_resource_share_m {
RNP_SHARE_CORPORATE = 0,
RNP_SHARE_INDEPENDENT,
};
+
+/* media type */
+enum rnp_media_type {
+ RNP_MEDIA_TYPE_UNKNOWN,
+ RNP_MEDIA_TYPE_FIBER,
+ RNP_MEDIA_TYPE_COPPER,
+ RNP_MEDIA_TYPE_BACKPLANE,
+ RNP_MEDIA_TYPE_NONE,
+};
+
+struct rnp_phy_meta {
+ uint16_t phy_type;
+ uint32_t speed_cap;
+ uint32_t supported_link;
+ uint16_t link_duplex;
+ uint16_t link_autoneg;
+ uint8_t media_type;
+ bool is_sgmii;
+ bool is_backplane;
+ bool fec;
+ uint32_t phy_identifier;
+};
+
+struct rnp_port_attr {
+ uint16_t max_mac_addrs; /* Max Support Mac Address */
+ uint16_t uc_hash_tb_size; /* Unicast Hash Table Size */
+ uint16_t max_uc_mac_hash; /* Max Num of hash MAC addr for UC */
+ uint16_t mc_hash_tb_size; /* Multicast Hash Table Size */
+ uint16_t max_mc_mac_hash; /* Max Num Of Hash Mac addr For MC */
+ uint16_t max_vlan_hash; /* Max Num Of Hash For Vlan ID*/
+ uint32_t hash_table_shift;
+ uint16_t rte_pid; /* Dpdk Manage Port Sequence Id */
+ uint8_t max_rx_queues; /* Belong To This Port Rxq Resource */
+ uint8_t max_tx_queues; /* Belong To This Port Rxq Resource */
+ uint8_t queue_ring_base;
+ uint8_t port_offset; /* Use For Redir Table Dma Ring Offset Of Port */
+ union {
+ uint8_t nr_lane; /* phy lane of This PF:0~3 */
+ uint8_t nr_port; /* phy lane of This PF:0~3 */
+ };
+ struct rnp_phy_meta phy_meta;
+ bool link_ready;
+ bool pre_link;
+ uint32_t speed;
+ uint16_t max_rx_pktlen; /* Current Port Max Support Packet Len */
+ uint16_t max_mtu;
+};
+
/*
* Structure to store private data for each driver instance (for each port).
*/
@@ -29,8 +105,16 @@ enum rnp_work_mode {
struct rnp_eth_port {
struct rnp_eth_adapter *adapt;
+ uint8_t mac_addr[RTE_ETHER_ADDR_LEN];
struct rnp_hw *hw;
struct rte_eth_dev *eth_dev;
+ struct rnp_port_attr attr;
+ /* Recvice Mac Address Record Table */
+ uint8_t mac_use_tb[RNP_MAX_MAC_ADDRS];
+ uint8_t use_num_mac;
+ bool port_stopped;
+ bool port_closed;
+ enum rnp_resource_share_m s_mode; /* Independent Port Resource */
} __rte_cache_aligned;
struct rnp_share_ops {
@@ -61,6 +145,10 @@ struct rnp_eth_adapter {
(&((struct rnp_eth_adapter *)(RNP_DEV_TO_PORT((eth_dev))->adapt))->hw)
#define RNP_HW_TO_ADAPTER(hw) \
((struct rnp_eth_adapter *)((hw)->back))
+#define RNP_PORT_TO_HW(port) \
+ (&(((struct rnp_eth_adapter *)(port)->adapt)->hw))
+#define RNP_PORT_TO_ADAPTER(port) \
+ ((struct rnp_eth_adapter *)((port)->adapt))
#define RNP_DEV_PP_PRIV_TO_MBX_OPS(dev) \
(((struct rnp_share_ops *)(dev)->process_private)->mbx_api)
#define RNP_DEV_TO_MBX_OPS(dev) RNP_DEV_PP_PRIV_TO_MBX_OPS(dev)
diff --git a/drivers/net/rnp/rnp_ethdev.c b/drivers/net/rnp/rnp_ethdev.c
index 8bb4fd5963..f0ef2ef329 100644
--- a/drivers/net/rnp/rnp_ethdev.c
+++ b/drivers/net/rnp/rnp_ethdev.c
@@ -11,6 +11,7 @@
#include "rnp_api.h"
#include "rnp_mbx.h"
#include "rnp_mbx_fw.h"
+#include "rnp_rxtx.h"
#include "rnp_logs.h"
static int
@@ -40,6 +41,62 @@ static int rnp_dev_close(struct rte_eth_dev *dev)
static const struct eth_dev_ops rnp_eth_dev_ops = {
};
+static void
+rnp_setup_port_attr(struct rnp_eth_port *port,
+ struct rte_eth_dev *dev,
+ uint8_t num_ports,
+ uint8_t p_id)
+{
+ struct rnp_port_attr *attr = &port->attr;
+ struct rnp_hw *hw = RNP_DEV_TO_HW(dev);
+ uint32_t lane_bit;
+
+ if (port->s_mode == RNP_SHARE_INDEPENDENT) {
+ attr->max_mac_addrs = RNP_PORT_MAX_MACADDR;
+ attr->max_uc_mac_hash = RNP_PORT_MAX_UC_MAC_SIZE;
+ attr->uc_hash_tb_size = RNP_PORT_MAX_UC_HASH_TB;
+ attr->max_mc_mac_hash = RNP_PORT_MAX_MACADDR;
+ attr->max_vlan_hash = RNP_PORT_MAX_VLAN_HASH;
+ attr->hash_table_shift = 26 - (attr->max_uc_mac_hash >> 7);
+ } else {
+ attr->max_mac_addrs = RNP_MAX_MAC_ADDRS / num_ports;
+ attr->max_uc_mac_hash = RNP_MAX_UC_MAC_SIZE / num_ports;
+ attr->uc_hash_tb_size = RNP_MAX_UC_HASH_TB;
+ attr->max_mc_mac_hash = RNP_MAX_MC_MAC_SIZE / num_ports;
+ attr->mc_hash_tb_size = RNP_MAC_MC_HASH_TB;
+ attr->max_vlan_hash = RNP_MAX_VLAN_HASH_TB_SIZE / num_ports;
+ attr->hash_table_shift = RNP_UTA_BIT_SHIFT;
+ }
+ if (hw->ncsi_en)
+ attr->uc_hash_tb_size -= hw->ncsi_rar_entries;
+ if (hw->device_id == RNP_DEV_ID_N400L_X4) {
+ attr->max_rx_queues = RNP_N400_MAX_RX_QUEUE_NUM;
+ attr->max_tx_queues = RNP_N400_MAX_TX_QUEUE_NUM;
+ } else {
+ attr->max_rx_queues = RNP_MAX_RX_QUEUE_NUM / num_ports;
+ attr->max_tx_queues = RNP_MAX_TX_QUEUE_NUM / num_ports;
+ }
+
+ attr->rte_pid = dev->data->port_id;
+ lane_bit = hw->phy_port_ids[p_id] & (hw->max_port_num - 1);
+
+ attr->nr_port = lane_bit;
+ attr->port_offset = rnp_eth_rd(hw, RNP_TC_PORT_MAP_TB(attr->nr_port));
+
+ rnp_mbx_get_lane_stat(dev);
+
+ PMD_DRV_LOG(INFO, "PF[%d] SW-ETH-PORT[%d]<->PHY_LANE[%d]\n",
+ hw->function, p_id, lane_bit);
+}
+
+static void
+rnp_init_filter_setup(struct rnp_eth_port *port,
+ uint8_t num_ports)
+{
+ RTE_SET_USED(port);
+ RTE_SET_USED(num_ports);
+}
+
static int
rnp_init_port_resource(struct rnp_eth_adapter *adapter,
struct rte_eth_dev *dev,
@@ -47,11 +104,53 @@ rnp_init_port_resource(struct rnp_eth_adapter *adapter,
uint8_t p_id)
{
struct rnp_eth_port *port = RNP_DEV_TO_PORT(dev);
+ struct rte_pci_device *pci_dev = adapter->pdev;
+ struct rnp_hw *hw = &adapter->hw;
+ port->adapt = adapter;
+ port->s_mode = adapter->s_mode;
+ port->port_stopped = 1;
+ port->hw = hw;
port->eth_dev = dev;
- adapter->ports[p_id] = port;
+
+ dev->device = &pci_dev->device;
+ rte_eth_copy_pci_info(dev, pci_dev);
dev->dev_ops = &rnp_eth_dev_ops;
- RTE_SET_USED(name);
+ dev->rx_queue_count = rnp_dev_rx_queue_count;
+ dev->rx_descriptor_status = rnp_dev_rx_descriptor_status;
+ dev->tx_descriptor_status = rnp_dev_tx_descriptor_status;
+ dev->rx_pkt_burst = rnp_recv_pkts;
+ dev->tx_pkt_burst = rnp_xmit_pkts;
+ dev->tx_pkt_prepare = rnp_prep_pkts;
+
+ rnp_setup_port_attr(port, dev, adapter->num_ports, p_id);
+ rnp_init_filter_setup(port, adapter->num_ports);
+ rnp_get_mac_addr(dev, port->mac_addr);
+ dev->data->mac_addrs = rte_zmalloc(name, sizeof(struct rte_ether_addr) *
+ port->attr.max_mac_addrs, 0);
+ if (!dev->data->mac_addrs) {
+ RNP_PMD_DRV_LOG(ERR, "Memory allocation "
+ "for MAC failed! Exiting.\n");
+ return -ENOMEM;
+ }
+ /* Allocate memory for storing hash filter MAC addresses */
+ dev->data->hash_mac_addrs = rte_zmalloc(name,
+ RTE_ETHER_ADDR_LEN * port->attr.max_uc_mac_hash, 0);
+ if (dev->data->hash_mac_addrs == NULL) {
+ RNP_PMD_INIT_LOG(ERR, "Failed to allocate %d bytes "
+ "needed to store MAC addresses",
+ RTE_ETHER_ADDR_LEN * port->attr.max_uc_mac_hash);
+ return -ENOMEM;
+ }
+
+ rnp_set_default_mac(dev, port->mac_addr);
+ rte_ether_addr_copy((const struct rte_ether_addr *)port->mac_addr,
+ dev->data->mac_addrs);
+ /* MTU */
+ dev->data->mtu = RTE_ETHER_MAX_LEN -
+ RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN;
+ adapter->ports[p_id] = port;
+ rte_eth_dev_probing_finish(dev);
return 0;
}
@@ -211,9 +310,116 @@ static int32_t rnp_reset_hw_pf(struct rnp_hw *hw)
return 0;
}
+static void
+rnp_mac_res_take_in(struct rnp_eth_port *port,
+ uint8_t index)
+{
+ if (!port->mac_use_tb[index]) {
+ port->mac_use_tb[index] = true;
+ port->use_num_mac++;
+ }
+}
+
+static void
+rnp_mac_res_remove(struct rnp_eth_port *port,
+ uint8_t index)
+{
+ if (port->mac_use_tb[index]) {
+ port->mac_use_tb[index] = false;
+ port->use_num_mac--;
+ }
+}
+
+static int32_t rnp_set_mac_addr_pf(struct rnp_eth_port *port,
+ uint8_t *mac, uint8_t vm_pool,
+ uint8_t index)
+{
+ struct rnp_hw *hw = RNP_PORT_TO_HW(port);
+ struct rnp_port_attr *attr = &port->attr;
+ uint8_t hw_idx;
+ uint32_t value;
+
+ if (port->use_num_mac > port->attr.max_mac_addrs ||
+ index > port->attr.max_mac_addrs)
+ return -ENOMEM;
+
+ if (vm_pool != UINT8_MAX)
+ hw_idx = (attr->nr_port * attr->max_mac_addrs) + vm_pool + index;
+ else
+ hw_idx = (attr->nr_port * attr->max_mac_addrs) + index;
+
+ rnp_mac_res_take_in(port, hw_idx);
+
+ value = (mac[0] << 8) | mac[1];
+ value |= RNP_MAC_FILTER_EN;
+ RNP_MACADDR_UPDATE_HI(hw, hw_idx, value);
+
+ value = (mac[2] << 24) | (mac[3] << 16) | (mac[4] << 8) | mac[5];
+ RNP_MACADDR_UPDATE_LO(hw, hw_idx, value);
+
+ return 0;
+}
+
+static void
+rnp_remove_mac_from_hw(struct rnp_eth_port *port,
+ uint8_t vm_pool, uint8_t index)
+{
+ struct rnp_hw *hw = RNP_PORT_TO_HW(port);
+ struct rnp_port_attr *attr = &port->attr;
+ uint16_t hw_idx;
+
+ if (vm_pool != UINT8_MAX)
+ hw_idx = (attr->nr_port * attr->max_mac_addrs) + vm_pool + index;
+ else
+ hw_idx = (attr->nr_port * attr->max_mac_addrs) + index;
+
+ rnp_mac_res_remove(port, hw_idx);
+
+ rnp_eth_wr(hw, RNP_RAL_BASE_ADDR(hw_idx), 0);
+ rnp_eth_wr(hw, RNP_RAH_BASE_ADDR(hw_idx), 0);
+}
+
+static int32_t
+rnp_clear_mac_addr_pf(struct rnp_eth_port *port,
+ uint8_t vm_pool, uint8_t index)
+{
+ rnp_remove_mac_from_hw(port, vm_pool, index);
+
+ return 0;
+}
+
+static int32_t rnp_get_mac_addr_pf(struct rnp_eth_port *port,
+ uint8_t lane,
+ uint8_t *macaddr)
+{
+ struct rnp_hw *hw = RNP_DEV_TO_HW(port->eth_dev);
+
+ return rnp_fw_get_macaddr(port->eth_dev, hw->pf_vf_num, macaddr, lane);
+}
+
+static int32_t
+rnp_set_default_mac_pf(struct rnp_eth_port *port,
+ uint8_t *mac)
+{
+ struct rnp_eth_adapter *adap = RNP_PORT_TO_ADAPTER(port);
+ uint16_t max_vfs;
+
+ if (port->s_mode == RNP_SHARE_INDEPENDENT)
+ return rnp_set_rafb(port->eth_dev, (uint8_t *)mac,
+ UINT8_MAX, 0);
+
+ max_vfs = adap->max_vfs;
+
+ return rnp_set_rafb(port->eth_dev, mac, max_vfs, 0);
+}
+
const struct rnp_mac_api rnp_mac_ops = {
.reset_hw = rnp_reset_hw_pf,
- .init_hw = rnp_init_hw_pf
+ .init_hw = rnp_init_hw_pf,
+ .get_mac_addr = rnp_get_mac_addr_pf,
+ .set_default_mac = rnp_set_default_mac_pf,
+ .set_rafb = rnp_set_mac_addr_pf,
+ .clear_rafb = rnp_clear_mac_addr_pf
};
static void
@@ -228,7 +434,11 @@ rnp_common_ops_init(struct rnp_eth_adapter *adapter)
static int
rnp_special_ops_init(struct rte_eth_dev *eth_dev)
{
- RTE_SET_USED(eth_dev);
+ struct rnp_eth_adapter *adapter = RNP_DEV_TO_ADAPTER(eth_dev);
+ struct rnp_share_ops *share_priv;
+
+ share_priv = adapter->share_priv;
+ share_priv->mac_api = &rnp_mac_ops;
return 0;
}
@@ -237,9 +447,9 @@ static int
rnp_eth_dev_init(struct rte_eth_dev *dev)
{
struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+ struct rnp_eth_port *port = RNP_DEV_TO_PORT(dev);
struct rnp_eth_adapter *adapter = NULL;
char name[RTE_ETH_NAME_MAX_LEN] = " ";
- struct rnp_eth_port *port = NULL;
struct rte_eth_dev *eth_dev;
struct rnp_hw *hw = NULL;
int32_t p_id;
@@ -275,13 +485,13 @@ rnp_eth_dev_init(struct rte_eth_dev *dev)
return ret;
}
adapter->share_priv = dev->process_private;
+ port->adapt = adapter;
rnp_common_ops_init(adapter);
+ rnp_init_mbx_ops_pf(hw);
rnp_get_nic_attr(adapter);
/* We need Use Device Id To Change The Resource Mode */
rnp_special_ops_init(dev);
- port->adapt = adapter;
port->hw = hw;
- rnp_init_mbx_ops_pf(hw);
for (p_id = 0; p_id < adapter->num_ports; p_id++) {
/* port 0 resource has been allocated When Probe */
if (!p_id) {
diff --git a/drivers/net/rnp/rnp_mbx_fw.c b/drivers/net/rnp/rnp_mbx_fw.c
index 6fe008351b..856b3f956b 100644
--- a/drivers/net/rnp/rnp_mbx_fw.c
+++ b/drivers/net/rnp/rnp_mbx_fw.c
@@ -269,3 +269,115 @@ int rnp_mbx_fw_reset_phy(struct rte_eth_dev *dev)
return rnp_fw_send_cmd_wait(dev, &req, &reply);
}
+
+int
+rnp_fw_get_macaddr(struct rte_eth_dev *dev,
+ int pfvfnum,
+ u8 *mac_addr,
+ int nr_lane)
+{
+ struct rnp_hw *hw = RNP_DEV_TO_HW(dev);
+ struct mbx_req_cookie *cookie;
+ struct mbx_fw_cmd_reply reply;
+ struct mbx_fw_cmd_req req;
+ struct mac_addr *mac;
+ int err;
+
+ memset(&req, 0, sizeof(req));
+ memset(&reply, 0, sizeof(reply));
+
+ if (!mac_addr)
+ return -EINVAL;
+
+ if (hw->mbx.irq_enabled) {
+ cookie = rnp_memzone_reserve(hw->cookie_p_name, 0);
+ if (!cookie)
+ return -ENOMEM;
+ memset(cookie->priv, 0, cookie->priv_len);
+ mac = (struct mac_addr *)cookie->priv;
+ build_get_macaddress_req(&req, 1 << nr_lane, pfvfnum, cookie);
+ err = rnp_mbx_fw_post_req(dev, &req, cookie);
+ if (err)
+ goto quit;
+
+ if ((1 << nr_lane) & mac->lanes) {
+ memcpy(mac_addr, mac->addrs[nr_lane].mac, 6);
+ err = 0;
+ } else {
+ err = -EIO;
+ }
+quit:
+ return err;
+ }
+ build_get_macaddress_req(&req, 1 << nr_lane, pfvfnum, &req);
+ err = rnp_fw_send_cmd_wait(dev, &req, &reply);
+ if (err) {
+ RNP_PMD_LOG(ERR, "%s: failed. err:%d\n", __func__, err);
+ return err;
+ }
+
+ if ((1 << nr_lane) & reply.mac_addr.lanes) {
+ memcpy(mac_addr, reply.mac_addr.addrs[nr_lane].mac, 6);
+ return 0;
+ }
+
+ return -EIO;
+}
+
+int rnp_mbx_get_lane_stat(struct rte_eth_dev *dev)
+{
+ struct rnp_eth_port *port = RNP_DEV_TO_PORT(dev);
+ struct rnp_phy_meta *phy_meta = &port->attr.phy_meta;
+ struct rnp_hw *hw = RNP_DEV_TO_HW(dev);
+ struct lane_stat_data *lane_stat;
+ int nr_lane = port->attr.nr_lane;
+ struct mbx_req_cookie *cookie;
+ struct mbx_fw_cmd_reply reply;
+ struct mbx_fw_cmd_req req;
+ int err = 0;
+
+ memset(&req, 0, sizeof(req));
+
+ if (hw->mbx.irq_enabled) {
+ cookie = rnp_memzone_reserve(hw->cookie_p_name, 0);
+
+ if (!cookie)
+ return -ENOMEM;
+ memset(cookie->priv, 0, cookie->priv_len);
+ lane_stat = (struct lane_stat_data *)cookie->priv;
+ build_get_lane_status_req(&req, nr_lane, cookie);
+ err = rnp_mbx_fw_post_req(dev, &req, cookie);
+ if (err)
+ goto quit;
+ } else {
+ memset(&reply, 0, sizeof(reply));
+ build_get_lane_status_req(&req, nr_lane, &req);
+ err = rnp_fw_send_cmd_wait(dev, &req, &reply);
+ if (err)
+ goto quit;
+ lane_stat = (struct lane_stat_data *)reply.data;
+ }
+
+ phy_meta->supported_link = lane_stat->supported_link;
+ phy_meta->is_backplane = lane_stat->is_backplane;
+ phy_meta->phy_identifier = lane_stat->phy_addr;
+ phy_meta->link_autoneg = lane_stat->autoneg;
+ phy_meta->link_duplex = lane_stat->duplex;
+ phy_meta->phy_type = lane_stat->phy_type;
+ phy_meta->is_sgmii = lane_stat->is_sgmii;
+ phy_meta->fec = lane_stat->fec;
+
+ if (phy_meta->is_sgmii) {
+ phy_meta->media_type = RNP_MEDIA_TYPE_COPPER;
+ phy_meta->supported_link |=
+ RNP_SPEED_CAP_100M_HALF | RNP_SPEED_CAP_10M_HALF;
+ } else if (phy_meta->is_backplane) {
+ phy_meta->media_type = RNP_MEDIA_TYPE_BACKPLANE;
+ } else {
+ phy_meta->media_type = RNP_MEDIA_TYPE_FIBER;
+ }
+
+ return 0;
+quit:
+ return err;
+}
diff --git a/drivers/net/rnp/rnp_mbx_fw.h b/drivers/net/rnp/rnp_mbx_fw.h
index f842639c86..c4a04a3f09 100644
--- a/drivers/net/rnp/rnp_mbx_fw.h
+++ b/drivers/net/rnp/rnp_mbx_fw.h
@@ -19,7 +19,9 @@ struct mbx_req_cookie {
enum GENERIC_CMD {
/* link configuration admin commands */
GET_PHY_ABALITY = 0x0601,
+ GET_MAC_ADDRESS = 0x0602,
RESET_PHY = 0x0603,
+ GET_LANE_STATUS = 0x0610,
SET_EVENT_MASK = 0x0613,
};
@@ -82,6 +84,61 @@ struct phy_abilities {
};
} __rte_packed __rte_aligned(4);
+#define RNP_SPEED_CAP_UNKNOWN (0)
+#define RNP_SPEED_CAP_10M_FULL BIT(2)
+#define RNP_SPEED_CAP_100M_FULL BIT(3)
+#define RNP_SPEED_CAP_1GB_FULL BIT(4)
+#define RNP_SPEED_CAP_10GB_FULL BIT(5)
+#define RNP_SPEED_CAP_40GB_FULL BIT(6)
+#define RNP_SPEED_CAP_25GB_FULL BIT(7)
+#define RNP_SPEED_CAP_50GB_FULL BIT(8)
+#define RNP_SPEED_CAP_100GB_FULL BIT(9)
+#define RNP_SPEED_CAP_10M_HALF BIT(10)
+#define RNP_SPEED_CAP_100M_HALF BIT(11)
+#define RNP_SPEED_CAP_1GB_HALF BIT(12)
+
+struct lane_stat_data {
+ u8 nr_lane; /* 0-3 cur port correspond with hw lane */
+ u8 pci_gen : 4; /* nic cur pci speed genX: 1,2,3 */
+ u8 pci_lanes : 4; /* nic cur pci x1 x2 x4 x8 x16 */
+ u8 pma_type;
+ u8 phy_type; /* interface media type */
+
+ u16 linkup : 1; /* cur port link state */
+ u16 duplex : 1; /* duplex state only RJ45 valid */
+ u16 autoneg : 1; /* autoneg state */
+ u16 fec : 1; /* fec state */
+ u16 rev_an : 1;
+ u16 link_traing : 1; /* link-traing state */
+ u16 media_available : 1;
+ u16 is_sgmii : 1; /* 1: Twisted Pair 0: FIBRE */
+ u16 link_fault : 4;
+#define LINK_LINK_FAULT BIT(0)
+#define LINK_TX_FAULT BIT(1)
+#define LINK_RX_FAULT BIT(2)
+#define LINK_REMOTE_FAULT BIT(3)
+ u16 is_backplane : 1; /* 1: Backplane Mode */
+ union {
+ u8 phy_addr; /* Phy MDIO address */
+ struct {
+ u8 mod_abs : 1;
+ u8 fault : 1;
+ u8 tx_dis : 1;
+ u8 los : 1;
+ } sfp;
+ };
+ u8 sfp_connector;
+ u32 speed; /* Current Speed Value */
+
+ u32 si_main;
+ u32 si_pre;
+ u32 si_post;
+ u32 si_tx_boost;
+ u32 supported_link; /* Cur nic Support Link cap */
+ u32 phy_id;
+ u32 advertised_link; /* autoneg mode advertised cap */
+} __rte_packed __rte_aligned(4);
+
/* firmware -> driver */
struct mbx_fw_cmd_reply {
/* fw must set: DD, CMP, Error(if error), copy value */
@@ -99,6 +156,19 @@ struct mbx_fw_cmd_reply {
};
/* ===== data ==== [16-64] */
union {
+ char data[0];
+
+ struct mac_addr {
+ int lanes;
+ struct _addr {
+ /* for macaddr:01:02:03:04:05:06
+ * mac-hi=0x01020304 mac-lo=0x05060000
+ */
+ unsigned char mac[8];
+ } addrs[4];
+ } mac_addr;
+
+ struct lane_stat_data lanestat;
struct phy_abilities phy_abilities;
};
} __rte_packed __rte_aligned(4);
@@ -128,10 +198,19 @@ struct mbx_fw_cmd_req {
#define REQUEST_BY_PXE 0xa3
} get_phy_ablity;
+ struct {
+ int lane_mask;
+ int pfvf_num;
+ } get_mac_addr;
+
struct {
unsigned short enable_stat;
unsigned short event_mask; /* enum link_event_mask */
} stat_event_mask;
+
+ struct {
+ int nr_lane;
+ } get_lane_st;
};
} __rte_packed __rte_aligned(4);
@@ -146,6 +225,23 @@ build_phy_abalities_req(struct mbx_fw_cmd_req *req, void *cookie)
req->cookie = cookie;
}
+static inline void
+build_get_macaddress_req(struct mbx_fw_cmd_req *req,
+ int lane_mask,
+ int pfvfnum,
+ void *cookie)
+{
+ req->flags = 0;
+ req->opcode = GET_MAC_ADDRESS;
+ req->datalen = sizeof(req->get_mac_addr);
+ req->cookie = cookie;
+ req->reply_lo = 0;
+ req->reply_hi = 0;
+
+ req->get_mac_addr.lane_mask = lane_mask;
+ req->get_mac_addr.pfvf_num = pfvfnum;
+}
+
/* enum link_event_mask or */
static inline void
build_link_set_event_mask(struct mbx_fw_cmd_req *req,
@@ -175,9 +271,28 @@ build_reset_phy_req(struct mbx_fw_cmd_req *req,
req->cookie = cookie;
}
+static inline void
+build_get_lane_status_req(struct mbx_fw_cmd_req *req,
+ int nr_lane, void *cookie)
+{
+ req->flags = 0;
+ req->opcode = GET_LANE_STATUS;
+ req->datalen = sizeof(req->get_lane_st);
+ req->cookie = cookie;
+ req->reply_lo = 0;
+ req->reply_hi = 0;
+ req->get_lane_st.nr_lane = nr_lane;
+}
+
int rnp_mbx_get_capability(struct rte_eth_dev *dev,
int *lane_mask,
int *nic_mode);
int rnp_mbx_link_event_enable(struct rte_eth_dev *dev, int enable);
int rnp_mbx_fw_reset_phy(struct rte_eth_dev *dev);
+int
+rnp_fw_get_macaddr(struct rte_eth_dev *dev,
+ int pfvfnum,
+ u8 *mac_addr,
+ int nr_lane);
+int rnp_mbx_get_lane_stat(struct rte_eth_dev *dev);
#endif /* __RNP_MBX_FW_H__*/
diff --git a/drivers/net/rnp/rnp_rxtx.c b/drivers/net/rnp/rnp_rxtx.c
new file mode 100644
index 0000000000..cc27c063ce
--- /dev/null
+++ b/drivers/net/rnp/rnp_rxtx.c
@@ -0,0 +1,82 @@
+#include <stdbool.h>
+#include <stdint.h>
+#include <unistd.h>
+#include <stdlib.h>
+#include <assert.h>
+
+#include <rte_version.h>
+#include <rte_ether.h>
+#include <rte_cycles.h>
+#include <rte_ethdev.h>
+#include <rte_malloc.h>
+#include <rte_memzone.h>
+#include <rte_vxlan.h>
+#include <rte_gre.h>
+#ifdef RTE_ARCH_ARM64
+#include <rte_cpuflags_64.h>
+#elif defined(RTE_ARCH_ARM)
+#include <rte_cpuflags_32.h>
+#endif
+
+#include "base/rnp_hw.h"
+#include "rnp.h"
+#include "rnp_rxtx.h"
+#include "rnp_logs.h"
+
+int
+rnp_dev_rx_descriptor_status(void *rx_queue, uint16_t offset)
+{
+ RTE_SET_USED(rx_queue);
+ RTE_SET_USED(offset);
+
+ return 0;
+}
+
+int
+rnp_dev_tx_descriptor_status(void *tx_queue, uint16_t offset)
+{
+ RTE_SET_USED(tx_queue);
+ RTE_SET_USED(offset);
+
+ return 0;
+}
+
+uint32_t
+rnp_dev_rx_queue_count(void *rx_queue)
+{
+ RTE_SET_USED(rx_queue);
+
+ return 0;
+}
+
+__rte_always_inline uint16_t
+rnp_recv_pkts(void *_rxq,
+ struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
+{
+ RTE_SET_USED(_rxq);
+ RTE_SET_USED(rx_pkts);
+ RTE_SET_USED(nb_pkts);
+
+ return 0;
+}
+
+__rte_always_inline uint16_t
+rnp_xmit_pkts(void *_txq, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
+{
+ RTE_SET_USED(_txq);
+ RTE_SET_USED(tx_pkts);
+ RTE_SET_USED(nb_pkts);
+
+ return 0;
+}
+
+uint16_t rnp_prep_pkts(void *tx_queue,
+ struct rte_mbuf **tx_pkts,
+ uint16_t nb_pkts)
+{
+ RTE_SET_USED(tx_queue);
+ RTE_SET_USED(tx_pkts);
+ RTE_SET_USED(nb_pkts);
+
+ return 0;
+}
diff --git a/drivers/net/rnp/rnp_rxtx.h b/drivers/net/rnp/rnp_rxtx.h
new file mode 100644
index 0000000000..0352971fcb
--- /dev/null
+++ b/drivers/net/rnp/rnp_rxtx.h
@@ -0,0 +1,14 @@
+#ifndef __RNP_RXTX_H__
+#define __RNP_RXTX_H__
+
+uint32_t rnp_dev_rx_queue_count(void *rx_queue);
+int rnp_dev_rx_descriptor_status(void *rx_queue, uint16_t offset);
+int rnp_dev_tx_descriptor_status(void *tx_queue, uint16_t offset);
+uint16_t
+rnp_recv_pkts(void *_rxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts);
+uint16_t
+rnp_xmit_pkts(void *_txq, struct rte_mbuf **tx_pkts, uint16_t nb_pkts);
+uint16_t rnp_prep_pkts(void *tx_queue,
+ struct rte_mbuf **tx_pkts,
+ uint16_t nb_pkts);
+#endif /* __RNP_RXTX_H__ */
--
2.27.0
^ permalink raw reply related [flat|nested] 27+ messages in thread* Re: [PATCH v6 6/8] net/rnp add port info resource init
2023-09-01 2:30 ` [PATCH v6 6/8] net/rnp add port info resource init Wenbo Cao
@ 2023-09-05 16:56 ` Ferruh Yigit
2023-09-06 9:07 ` 11
0 siblings, 1 reply; 27+ messages in thread
From: Ferruh Yigit @ 2023-09-05 16:56 UTC (permalink / raw)
To: Wenbo Cao; +Cc: dev, thomas, andrew.rybchenko, yaojun
On 9/1/2023 3:30 AM, Wenbo Cao wrote:
> Add Api For FW Mac Info, Port Resoucre info init Code
> For Different Shape Of Nic
>
> Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
<...>
> @@ -47,11 +104,53 @@ rnp_init_port_resource(struct rnp_eth_adapter *adapter,
> uint8_t p_id)
> {
> struct rnp_eth_port *port = RNP_DEV_TO_PORT(dev);
> + struct rte_pci_device *pci_dev = adapter->pdev;
> + struct rnp_hw *hw = &adapter->hw;
>
> + port->adapt = adapter;
> + port->s_mode = adapter->s_mode;
> + port->port_stopped = 1;
> + port->hw = hw;
> port->eth_dev = dev;
> - adapter->ports[p_id] = port;
> +
> + dev->device = &pci_dev->device;
> + rte_eth_copy_pci_info(dev, pci_dev);
> dev->dev_ops = &rnp_eth_dev_ops;
> - RTE_SET_USED(name);
> + dev->rx_queue_count = rnp_dev_rx_queue_count;
> + dev->rx_descriptor_status = rnp_dev_rx_descriptor_status;
> + dev->tx_descriptor_status = rnp_dev_tx_descriptor_status;
> + dev->rx_pkt_burst = rnp_recv_pkts;
> + dev->tx_pkt_burst = rnp_xmit_pkts;
> + dev->tx_pkt_prepare = rnp_prep_pkts;
> +
> + rnp_setup_port_attr(port, dev, adapter->num_ports, p_id);
> + rnp_init_filter_setup(port, adapter->num_ports);
> + rnp_get_mac_addr(dev, port->mac_addr);
> + dev->data->mac_addrs = rte_zmalloc(name, sizeof(struct rte_ether_addr) *
> + port->attr.max_mac_addrs, 0);
> + if (!dev->data->mac_addrs) {
> + RNP_PMD_DRV_LOG(ERR, "Memory allocation "
> + "for MAC failed! Exiting.\n");
> + return -ENOMEM;
> + }
> + /* Allocate memory for storing hash filter MAC addresses */
> + dev->data->hash_mac_addrs = rte_zmalloc(name,
> + RTE_ETHER_ADDR_LEN * port->attr.max_uc_mac_hash, 0);
> + if (dev->data->hash_mac_addrs == NULL) {
> + RNP_PMD_INIT_LOG(ERR, "Failed to allocate %d bytes "
> + "needed to store MAC addresses",
> + RTE_ETHER_ADDR_LEN * port->attr.max_uc_mac_hash);
> + return -ENOMEM;
Should free 'dev->data->mac_addrs' here, or even better can be to
implement 'rnp_dev_close()' to free device resources.
> + }
> +
> + rnp_set_default_mac(dev, port->mac_addr);
I guess the 'port->mac_addr' got from device via 'rnp_get_mac_addr()'
above, but if 'rnp_get_mac_addr()' fails what will be the value. Should
there be check and set a random mac if required?
> + rte_ether_addr_copy((const struct rte_ether_addr *)port->mac_addr,
> + dev->data->mac_addrs);
> + /* MTU */
> + dev->data->mtu = RTE_ETHER_MAX_LEN -
> + RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN;
> + adapter->ports[p_id] = port;
> + rte_eth_dev_probing_finish(dev);
>
rte_eth_dev_probing_finish() is not required, as
'rte_eth_dev_pci_generic_probe()' calls it if dev_init() returns success.
<...>
> +static int32_t rnp_get_mac_addr_pf(struct rnp_eth_port *port,
> + uint8_t lane,
> + uint8_t *macaddr)
> +{
> + struct rnp_hw *hw = RNP_DEV_TO_HW(port->eth_dev);
> +
> + return rnp_fw_get_macaddr(port->eth_dev, hw->pf_vf_num, macaddr, lane);
> +}
> +
These are mac_ops functions, normally the reason to have mac_ops is to
support different HW that behaves slightly different and this difference
managed by different function pointers per device. Is this your usecase?
And since these are mac_ops, defined in the base folder header, it suits
better to have separate .c file for them which is under base file, what
do you think? Or am I getting these mac_ops wrong?
> +static int32_t
> +rnp_set_default_mac_pf(struct rnp_eth_port *port,
> + uint8_t *mac)
> +{
> + struct rnp_eth_adapter *adap = RNP_PORT_TO_ADAPTER(port);
> + uint16_t max_vfs;
> +
> + if (port->s_mode == RNP_SHARE_INDEPENDENT)
> + return rnp_set_rafb(port->eth_dev, (uint8_t *)mac,
> + UINT8_MAX, 0);
> +
> + max_vfs = adap->max_vfs;
> +
> + return rnp_set_rafb(port->eth_dev, mac, max_vfs, 0);
> +}
> +
> const struct rnp_mac_api rnp_mac_ops = {
> .reset_hw = rnp_reset_hw_pf,
> - .init_hw = rnp_init_hw_pf
> + .init_hw = rnp_init_hw_pf,
> + .get_mac_addr = rnp_get_mac_addr_pf,
> + .set_default_mac = rnp_set_default_mac_pf,
> + .set_rafb = rnp_set_mac_addr_pf,
> + .clear_rafb = rnp_clear_mac_addr_pf
> };
>
> static void
> @@ -228,7 +434,11 @@ rnp_common_ops_init(struct rnp_eth_adapter *adapter)
> static int
> rnp_special_ops_init(struct rte_eth_dev *eth_dev)
> {
> - RTE_SET_USED(eth_dev);
> + struct rnp_eth_adapter *adapter = RNP_DEV_TO_ADAPTER(eth_dev);
> + struct rnp_share_ops *share_priv;
> +
> + share_priv = adapter->share_priv;
> + share_priv->mac_api = &rnp_mac_ops;
>
Can you please describe why this 'rnp_special_ops_init()' is for (its
difference from rnp_common_ops_init()) ?
^ permalink raw reply [flat|nested] 27+ messages in thread* RE: [PATCH v6 6/8] net/rnp add port info resource init
2023-09-05 16:56 ` Ferruh Yigit
@ 2023-09-06 9:07 ` 11
0 siblings, 0 replies; 27+ messages in thread
From: 11 @ 2023-09-06 9:07 UTC (permalink / raw)
To: 'Ferruh Yigit'; +Cc: dev, thomas, andrew.rybchenko, yaojun
Hi Ferruh,
> -----Original Message-----
> From: Ferruh Yigit <ferruh.yigit@amd.com>
> Sent: 2023年9月6日 0:56
> To: Wenbo Cao <caowenbo@mucse.com>
> Cc: dev@dpdk.org; thomas@monjalon.net; andrew.rybchenko@oktetlabs.ru;
> yaojun@mucse.com
> Subject: Re: [PATCH v6 6/8] net/rnp add port info resource init
>
> On 9/1/2023 3:30 AM, Wenbo Cao wrote:
> > Add Api For FW Mac Info, Port Resoucre info init Code For Different
> > Shape Of Nic
> >
> > Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
>
> <...>
>
> > @@ -47,11 +104,53 @@ rnp_init_port_resource(struct rnp_eth_adapter
> *adapter,
> > uint8_t p_id)
> > {
> > struct rnp_eth_port *port = RNP_DEV_TO_PORT(dev);
> > + struct rte_pci_device *pci_dev = adapter->pdev;
> > + struct rnp_hw *hw = &adapter->hw;
> >
> > + port->adapt = adapter;
> > + port->s_mode = adapter->s_mode;
> > + port->port_stopped = 1;
> > + port->hw = hw;
> > port->eth_dev = dev;
> > - adapter->ports[p_id] = port;
> > +
> > + dev->device = &pci_dev->device;
> > + rte_eth_copy_pci_info(dev, pci_dev);
> > dev->dev_ops = &rnp_eth_dev_ops;
> > - RTE_SET_USED(name);
> > + dev->rx_queue_count = rnp_dev_rx_queue_count;
> > + dev->rx_descriptor_status = rnp_dev_rx_descriptor_status;
> > + dev->tx_descriptor_status = rnp_dev_tx_descriptor_status;
> > + dev->rx_pkt_burst = rnp_recv_pkts;
> > + dev->tx_pkt_burst = rnp_xmit_pkts;
> > + dev->tx_pkt_prepare = rnp_prep_pkts;
> > +
> > + rnp_setup_port_attr(port, dev, adapter->num_ports, p_id);
> > + rnp_init_filter_setup(port, adapter->num_ports);
> > + rnp_get_mac_addr(dev, port->mac_addr);
> > + dev->data->mac_addrs = rte_zmalloc(name, sizeof(struct
> rte_ether_addr) *
> > + port->attr.max_mac_addrs, 0);
> > + if (!dev->data->mac_addrs) {
> > + RNP_PMD_DRV_LOG(ERR, "Memory allocation "
> > + "for MAC failed! Exiting.\n");
> > + return -ENOMEM;
> > + }
> > + /* Allocate memory for storing hash filter MAC addresses */
> > + dev->data->hash_mac_addrs = rte_zmalloc(name,
> > + RTE_ETHER_ADDR_LEN * port-
> >attr.max_uc_mac_hash, 0);
> > + if (dev->data->hash_mac_addrs == NULL) {
> > + RNP_PMD_INIT_LOG(ERR, "Failed to allocate %d bytes "
> > + "needed to store MAC addresses",
> > + RTE_ETHER_ADDR_LEN * port-
> >attr.max_uc_mac_hash);
> > + return -ENOMEM;
>
> Should free 'dev->data->mac_addrs' here, or even better can be to implement
> 'rnp_dev_close()' to free device resources.
>
>
> > + dev->data->mac_addrs = rte_zmalloc(name, sizeof(struct
Yes, for the failed branch, need to free the last alloc success argument mac_addrs
> > + }
> > +
> > + rnp_set_default_mac(dev, port->mac_addr);
>
> I guess the 'port->mac_addr' got from device via 'rnp_get_mac_addr()'
> above, but if 'rnp_get_mac_addr()' fails what will be the value. Should there be
> check and set a random mac if required?
>
Yes, you are right :) , it the rnp_get_mac_addr failed, this may have fireware issue thing about
Communicate.This need to be check and warning.
> > + rte_ether_addr_copy((const struct rte_ether_addr *)port->mac_addr,
> > + dev->data->mac_addrs);
> > + /* MTU */
> > + dev->data->mtu = RTE_ETHER_MAX_LEN -
> > + RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN;
> > + adapter->ports[p_id] = port;
> > + rte_eth_dev_probing_finish(dev);
> >
>
> rte_eth_dev_probing_finish() is not required, as
> 'rte_eth_dev_pci_generic_probe()' calls it if dev_init() returns success.
>
> <...>
>
For N10 chip is just have two pcie-bdf, the one pcie-bdf may have multiple port(max 4)
'rte_eth_dev_pci_generic_probe' probe the pcie-pdf port(master port port 0)
the other port alloc by 'rnp_alloc_eth_port', not set 'rte_eth_dev_probing_finish()'
so the 1,2,3 port will set 'rte_eth_dev_probing_finish()'.
> > +static int32_t rnp_get_mac_addr_pf(struct rnp_eth_port *port,
> > + uint8_t lane,
> > + uint8_t *macaddr)
> > +{
> > + struct rnp_hw *hw = RNP_DEV_TO_HW(port->eth_dev);
> > +
> > + return rnp_fw_get_macaddr(port->eth_dev, hw->pf_vf_num, macaddr,
> > +lane); }
> > +
>
> These are mac_ops functions, normally the reason to have mac_ops is to support
> different HW that behaves slightly different and this difference managed by
> different function pointers per device. Is this your usecase?
>
This is because for one pcie-bdf one port and one pcie-bdf multiple port usecase
For hareware design limit, we max can support eight port but just have two pcie-bdf.
We must use two set of api to resolve one pcie-bdf one port and
one pcie-bdf multiple port for the port Control independence feature.
Such as vlan-filter, mac unicast/Multicast filter, promise, Multicast promise, Broadcast promise.
So that some API is common used for one-port-pcie-bdf and multiple-port-one-pcie-bdf :(.
And some mac-api must change when known the nic is multiple-port mode.
> And since these are mac_ops, defined in the base folder header, it suits better to
> have separate .c file for them which is under base file, what do you think? Or am
> I getting these mac_ops wrong?
>
You are right, 😊 mac_ops api need to depart from rnp.c to a separate.c .
> > +static int32_t
> > +rnp_set_default_mac_pf(struct rnp_eth_port *port,
> > + uint8_t *mac)
> > +{
> > + struct rnp_eth_adapter *adap = RNP_PORT_TO_ADAPTER(port);
> > + uint16_t max_vfs;
> > +
> > + if (port->s_mode == RNP_SHARE_INDEPENDENT)
> > + return rnp_set_rafb(port->eth_dev, (uint8_t *)mac,
> > + UINT8_MAX, 0);
> > +
> > + max_vfs = adap->max_vfs;
> > +
> > + return rnp_set_rafb(port->eth_dev, mac, max_vfs, 0); }
> > +
> > const struct rnp_mac_api rnp_mac_ops = {
> > .reset_hw = rnp_reset_hw_pf,
> > - .init_hw = rnp_init_hw_pf
> > + .init_hw = rnp_init_hw_pf,
> > + .get_mac_addr = rnp_get_mac_addr_pf,
> > + .set_default_mac = rnp_set_default_mac_pf,
> > + .set_rafb = rnp_set_mac_addr_pf,
> > + .clear_rafb = rnp_clear_mac_addr_pf
> > };
> >
> > static void
> > @@ -228,7 +434,11 @@ rnp_common_ops_init(struct rnp_eth_adapter
> > *adapter) static int rnp_special_ops_init(struct rte_eth_dev
> > *eth_dev) {
> > - RTE_SET_USED(eth_dev);
> > + struct rnp_eth_adapter *adapter = RNP_DEV_TO_ADAPTER(eth_dev);
> > + struct rnp_share_ops *share_priv;
> > +
> > + share_priv = adapter->share_priv;
> > + share_priv->mac_api = &rnp_mac_ops;
> >
>
> Can you please describe why this 'rnp_special_ops_init()' is for (its difference
> from rnp_common_ops_init()) ?
>
>
In 'rnp_common_ops_init' both one pcie-bdf one port and one pcie-bdf multiple port,
will use mailbox to communicase with fireware.
'rnp_special_ops_init() is design to change the mac api when the nic is multiple-port
mode.
For now, the code is just for one-pcie-one-port mode.
^ permalink raw reply [flat|nested] 27+ messages in thread
* [PATCH v6 7/8] net/rnp add devargs runtime parsing functions
2023-09-01 2:30 [PATCH v6 0/8] [v6]drivers/net Add Support mucse N10 Pmd Driver Wenbo Cao
` (5 preceding siblings ...)
2023-09-01 2:30 ` [PATCH v6 6/8] net/rnp add port info resource init Wenbo Cao
@ 2023-09-01 2:30 ` Wenbo Cao
2023-09-05 15:46 ` Ferruh Yigit
2023-09-01 2:30 ` [PATCH v6 8/8] net/rnp handle device interrupts Wenbo Cao
2023-09-05 15:34 ` [PATCH v6 0/8] [v6]drivers/net Add Support mucse N10 Pmd Driver Ferruh Yigit
8 siblings, 1 reply; 27+ messages in thread
From: Wenbo Cao @ 2023-09-01 2:30 UTC (permalink / raw)
To: Wenbo Cao; +Cc: dev, ferruh.yigit, thomas, andrew.rybchenko, yaojun
add various runtime devargs command line options
supported by this driver.
Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
---
drivers/net/rnp/rnp.h | 22 +++++
drivers/net/rnp/rnp_ethdev.c | 166 +++++++++++++++++++++++++++++++++++
drivers/net/rnp/rnp_mbx_fw.c | 164 ++++++++++++++++++++++++++++++++++
drivers/net/rnp/rnp_mbx_fw.h | 69 +++++++++++++++
4 files changed, 421 insertions(+)
diff --git a/drivers/net/rnp/rnp.h b/drivers/net/rnp/rnp.h
index 9ae801414b..7e797cf177 100644
--- a/drivers/net/rnp/rnp.h
+++ b/drivers/net/rnp/rnp.h
@@ -107,6 +107,8 @@ struct rnp_eth_port {
struct rnp_eth_adapter *adapt;
uint8_t mac_addr[RTE_ETHER_ADDR_LEN];
struct rnp_hw *hw;
+ uint8_t rx_func_sec; /* force set io rx_func */
+ uint8_t tx_func_sec; /* force set io tx func */
struct rte_eth_dev *eth_dev;
struct rnp_port_attr attr;
/* Recvice Mac Address Record Table */
@@ -122,6 +124,13 @@ struct rnp_share_ops {
const struct rnp_mac_api *mac_api;
} __rte_cache_aligned;
+enum {
+ RNP_IO_FUNC_USE_NONE = 0,
+ RNP_IO_FUNC_USE_VEC,
+ RNP_IO_FUNC_USE_SIMPLE,
+ RNP_IO_FUNC_USE_COMMON,
+};
+
struct rnp_eth_adapter {
enum rnp_work_mode mode;
enum rnp_resource_share_m s_mode; /* Port Resource Share Policy */
@@ -135,6 +144,19 @@ struct rnp_eth_adapter {
int max_link_speed;
uint8_t num_ports; /* Cur Pf Has physical Port Num */
uint8_t lane_mask;
+
+ uint8_t rx_func_sec; /* force set io rx_func */
+ uint8_t tx_func_sec; /* force set io tx func*/
+ /*fw-update*/
+ bool do_fw_update;
+ char *fw_path;
+
+ bool loopback_en;
+ bool fw_sfp_10g_1g_auto_det;
+ int fw_force_speed_1g;
+#define FOCE_SPEED_1G_NOT_SET (-1)
+#define FOCE_SPEED_1G_DISABLED (0)
+#define FOCE_SPEED_1G_ENABLED (1)
} __rte_cache_aligned;
#define RNP_DEV_TO_PORT(eth_dev) \
diff --git a/drivers/net/rnp/rnp_ethdev.c b/drivers/net/rnp/rnp_ethdev.c
index f0ef2ef329..131da42259 100644
--- a/drivers/net/rnp/rnp_ethdev.c
+++ b/drivers/net/rnp/rnp_ethdev.c
@@ -6,6 +6,7 @@
#include <rte_io.h>
#include <rte_malloc.h>
#include <ethdev_driver.h>
+#include <rte_kvargs.h>
#include "rnp.h"
#include "rnp_api.h"
@@ -14,6 +15,13 @@
#include "rnp_rxtx.h"
#include "rnp_logs.h"
+#define RNP_HW_MAC_LOOPBACK_ARG "hw_loopback"
+#define RNP_FW_UPDATE "fw_update"
+#define RNP_RX_FUNC_SELECT "rx_func_sec"
+#define RNP_TX_FUNC_SELECT "tx_func_sec"
+#define RNP_FW_4X10G_10G_1G_DET "fw_4x10g_10g_1g_auto_det"
+#define RNP_FW_FORCE_SPEED_1G "fw_force_1g_speed"
+
static int
rnp_mac_rx_disable(struct rte_eth_dev *dev)
{
@@ -108,6 +116,8 @@ rnp_init_port_resource(struct rnp_eth_adapter *adapter,
struct rnp_hw *hw = &adapter->hw;
port->adapt = adapter;
+ port->rx_func_sec = adapter->rx_func_sec;
+ port->tx_func_sec = adapter->tx_func_sec;
port->s_mode = adapter->s_mode;
port->port_stopped = 1;
port->hw = hw;
@@ -443,6 +453,154 @@ rnp_special_ops_init(struct rte_eth_dev *eth_dev)
return 0;
}
+static const char *const rnp_valid_arguments[] = {
+ RNP_HW_MAC_LOOPBACK_ARG,
+ RNP_FW_UPDATE,
+ RNP_RX_FUNC_SELECT,
+ RNP_TX_FUNC_SELECT,
+ RNP_FW_4X10G_10G_1G_DET,
+ RNP_FW_FORCE_SPEED_1G,
+ NULL
+};
+
+static int
+rnp_parse_handle_devarg(const char *key, const char *value,
+ void *extra_args)
+{
+ struct rnp_eth_adapter *adapter = NULL;
+
+ if (value == NULL || extra_args == NULL)
+ return -EINVAL;
+
+ if (strcmp(key, RNP_HW_MAC_LOOPBACK_ARG) == 0) {
+ uint64_t *n = extra_args;
+ *n = (uint16_t)strtoul(value, NULL, 10);
+ if (*n > UINT16_MAX && errno == ERANGE) {
+ RNP_PMD_DRV_LOG(ERR, "invalid extra param value\n");
+ return -1;
+ }
+ } else if (strcmp(key, RNP_FW_UPDATE) == 0) {
+ adapter = (struct rnp_eth_adapter *)extra_args;
+ adapter->do_fw_update = true;
+ adapter->fw_path = strdup(value);
+ } else if (strcmp(key, RNP_FW_4X10G_10G_1G_DET) == 0) {
+ adapter = (struct rnp_eth_adapter *)extra_args;
+ if (adapter->num_ports == 2 && adapter->hw.speed == 10 * 1000) {
+ adapter->fw_sfp_10g_1g_auto_det =
+ (strcmp(value, "on") == 0) ? true : false;
+ } else {
+ adapter->fw_sfp_10g_1g_auto_det = false;
+ }
+ } else if (strcmp(key, RNP_FW_FORCE_SPEED_1G) == 0) {
+ adapter = (struct rnp_eth_adapter *)extra_args;
+ if (adapter->num_ports == 2) {
+ if (strcmp(value, "on") == 0)
+ adapter->fw_force_speed_1g = FOCE_SPEED_1G_ENABLED;
+ else if (strcmp(value, "off") == 0)
+ adapter->fw_force_speed_1g = FOCE_SPEED_1G_DISABLED;
+ }
+ } else {
+ return -1;
+ }
+
+ return 0;
+}
+
+static int
+rnp_parse_io_select_func(const char *key, const char *value, void *extra_args)
+{
+ uint8_t select = RNP_IO_FUNC_USE_NONE;
+
+ RTE_SET_USED(key);
+
+ if (strcmp(value, "vec") == 0)
+ select = RNP_IO_FUNC_USE_VEC;
+ else if (strcmp(value, "simple") == 0)
+ select = RNP_IO_FUNC_USE_SIMPLE;
+ else if (strcmp(value, "common") == 0)
+ select = RNP_IO_FUNC_USE_COMMON;
+
+ *(uint8_t *)extra_args = select;
+
+ return 0;
+}
+
+static int
+rnp_parse_devargs(struct rnp_eth_adapter *adapter,
+ struct rte_devargs *devargs)
+{
+ uint8_t rx_io_func = RNP_IO_FUNC_USE_NONE;
+ uint8_t tx_io_func = RNP_IO_FUNC_USE_NONE;
+ struct rte_kvargs *kvlist;
+ bool loopback_en = false;
+ int ret = 0;
+
+ adapter->do_fw_update = false;
+ adapter->fw_sfp_10g_1g_auto_det = false;
+ adapter->fw_force_speed_1g = FOCE_SPEED_1G_NOT_SET;
+
+ if (!devargs)
+ goto def;
+
+ kvlist = rte_kvargs_parse(devargs->args, rnp_valid_arguments);
+ if (kvlist == NULL)
+ goto def;
+
+ if (rte_kvargs_count(kvlist, RNP_HW_MAC_LOOPBACK_ARG) == 1)
+ ret = rte_kvargs_process(kvlist, RNP_HW_MAC_LOOPBACK_ARG,
+ &rnp_parse_handle_devarg, &loopback_en);
+
+ if (rte_kvargs_count(kvlist, RNP_FW_4X10G_10G_1G_DET) == 1)
+ ret = rte_kvargs_process(kvlist,
+ RNP_FW_4X10G_10G_1G_DET,
+ &rnp_parse_handle_devarg,
+ adapter);
+
+ if (rte_kvargs_count(kvlist, RNP_FW_FORCE_SPEED_1G) == 1)
+ ret = rte_kvargs_process(kvlist,
+ RNP_FW_FORCE_SPEED_1G,
+ &rnp_parse_handle_devarg,
+ adapter);
+
+ if (rte_kvargs_count(kvlist, RNP_FW_UPDATE) == 1)
+ ret = rte_kvargs_process(kvlist, RNP_FW_UPDATE,
+ &rnp_parse_handle_devarg, adapter);
+ if (rte_kvargs_count(kvlist, RNP_RX_FUNC_SELECT) == 1)
+ ret = rte_kvargs_process(kvlist, RNP_RX_FUNC_SELECT,
+ &rnp_parse_io_select_func, &rx_io_func);
+ if (rte_kvargs_count(kvlist, RNP_TX_FUNC_SELECT) == 1)
+ ret = rte_kvargs_process(kvlist, RNP_TX_FUNC_SELECT,
+ &rnp_parse_io_select_func, &tx_io_func);
+ rte_kvargs_free(kvlist);
+def:
+ adapter->loopback_en = loopback_en;
+ adapter->rx_func_sec = rx_io_func;
+ adapter->tx_func_sec = tx_io_func;
+
+ return ret;
+}
+
+static int rnp_post_handle(struct rnp_eth_adapter *adapter)
+{
+ bool on = false;
+
+ if (!adapter->eth_dev)
+ return -ENOMEM;
+ if (adapter->do_fw_update && adapter->fw_path) {
+ rnp_fw_update(adapter);
+ adapter->do_fw_update = 0;
+ }
+
+ if (adapter->fw_sfp_10g_1g_auto_det)
+ return rnp_hw_set_fw_10g_1g_auto_detch(adapter->eth_dev, 1);
+
+ on = (adapter->fw_force_speed_1g == FOCE_SPEED_1G_ENABLED) ? 1 : 0;
+ if (adapter->fw_force_speed_1g != FOCE_SPEED_1G_NOT_SET)
+ return rnp_hw_set_fw_force_speed_1g(adapter->eth_dev, on);
+
+ return 0;
+}
+
static int
rnp_eth_dev_init(struct rte_eth_dev *dev)
{
@@ -492,6 +650,11 @@ rnp_eth_dev_init(struct rte_eth_dev *dev)
/* We need Use Device Id To Change The Resource Mode */
rnp_special_ops_init(dev);
port->hw = hw;
+ ret = rnp_parse_devargs(adapter, pci_dev->device.devargs);
+ if (ret) {
+ PMD_DRV_LOG(ERR, "parse_devargs failed");
+ return ret;
+ }
for (p_id = 0; p_id < adapter->num_ports; p_id++) {
/* port 0 resource has been allocated When Probe */
if (!p_id) {
@@ -517,6 +680,9 @@ rnp_eth_dev_init(struct rte_eth_dev *dev)
rnp_mac_rx_disable(eth_dev);
rnp_mac_tx_disable(eth_dev);
}
+ ret = rnp_post_handle(adapter);
+ if (ret)
+ goto eth_alloc_error;
return 0;
eth_alloc_error:
diff --git a/drivers/net/rnp/rnp_mbx_fw.c b/drivers/net/rnp/rnp_mbx_fw.c
index 856b3f956b..0c3f499cf2 100644
--- a/drivers/net/rnp/rnp_mbx_fw.c
+++ b/drivers/net/rnp/rnp_mbx_fw.c
@@ -105,6 +105,27 @@ static int rnp_mbx_fw_post_req(struct rte_eth_dev *dev,
return err;
}
+static int
+rnp_mbx_write_posted_locked(struct rte_eth_dev *dev, struct mbx_fw_cmd_req *req)
+{
+ const struct rnp_mbx_api *ops = RNP_DEV_TO_MBX_OPS(dev);
+ struct rnp_hw *hw = RNP_DEV_TO_HW(dev);
+ int err = 0;
+
+ rte_spinlock_lock(&hw->fw_lock);
+
+ err = ops->write_posted(dev, (u32 *)req,
+ (req->datalen + MBX_REQ_HDR_LEN) / 4, MBX_FW);
+ if (err) {
+ RNP_PMD_LOG(ERR, "%s failed!\n", __func__);
+ goto quit;
+ }
+
+quit:
+ rte_spinlock_unlock(&hw->fw_lock);
+ return err;
+}
+
static int rnp_fw_get_capablity(struct rte_eth_dev *dev,
struct phy_abilities *abil)
{
@@ -381,3 +402,146 @@ int rnp_mbx_get_lane_stat(struct rte_eth_dev *dev)
quit:
return err;
}
+
+static int rnp_maintain_req(struct rte_eth_dev *dev,
+ int cmd,
+ int arg0,
+ int req_data_bytes,
+ int reply_bytes,
+ phys_addr_t dma_phy_addr)
+{
+ struct rnp_hw *hw = RNP_DEV_TO_HW(dev);
+ struct mbx_req_cookie *cookie = NULL;
+ struct mbx_fw_cmd_req req;
+ int err;
+
+ if (!hw->mbx.irq_enabled)
+ return -EIO;
+ cookie = rnp_memzone_reserve(hw->cookie_p_name, 0);
+ if (!cookie)
+ return -ENOMEM;
+ memset(&req, 0, sizeof(req));
+ cookie->timeout_ms = 60 * 1000; /* 60s */
+
+ build_maintain_req(&req,
+ cookie,
+ cmd,
+ arg0,
+ req_data_bytes,
+ reply_bytes,
+ dma_phy_addr & 0xffffffff,
+ (dma_phy_addr >> 32) & 0xffffffff);
+
+ err = rnp_mbx_fw_post_req(dev, &req, cookie);
+
+ return (err) ? -EIO : 0;
+}
+
+int rnp_fw_update(struct rnp_eth_adapter *adapter)
+{
+ const struct rte_memzone *rz = NULL;
+ struct maintain_req *mt;
+ FILE *file;
+ int fsz;
+#define MAX_FW_BIN_SZ (552 * 1024)
+#define FW_256KB (256 * 1024)
+
+ RNP_PMD_LOG(INFO, "%s: %s\n", __func__, adapter->fw_path);
+
+ file = fopen(adapter->fw_path, "rb");
+ if (!file) {
+ RNP_PMD_LOG(ERR,
+ "RNP: [%s] %s can't open for read\n",
+ __func__,
+ adapter->fw_path);
+ return -ENOENT;
+ }
+ /* get dma */
+ rz = rte_memzone_reserve("fw_update", MAX_FW_BIN_SZ, SOCKET_ID_ANY, 4);
+ if (rz == NULL) {
+ RNP_PMD_LOG(ERR, "RNP: [%s] not memory:%d\n", __func__,
+ MAX_FW_BIN_SZ);
+ return -EFBIG;
+ }
+ memset(rz->addr, 0xff, rz->len);
+ mt = (struct maintain_req *)rz->addr;
+
+ /* read data */
+ fsz = fread(mt->data, 1, rz->len, file);
+ if (fsz <= 0) {
+ RNP_PMD_LOG(INFO, "RNP: [%s] read failed! err:%d\n",
+ __func__, fsz);
+ return -EIO;
+ }
+ fclose(file);
+
+ if (fsz > ((256 + 4) * 1024)) {
+ printf("fw length:%d is two big. not supported!\n", fsz);
+ return -EINVAL;
+ }
+ RNP_PMD_LOG(NOTICE, "RNP: fw update ...\n");
+ fflush(stdout);
+
+ /* ==== update fw */
+ mt->magic = MAINTAIN_MAGIC;
+ mt->cmd = MT_WRITE_FLASH;
+ mt->arg0 = 1;
+ mt->req_data_bytes = (fsz > FW_256KB) ? FW_256KB : fsz;
+ mt->reply_bytes = 0;
+
+ if (rnp_maintain_req(adapter->eth_dev, mt->cmd, mt->arg0,
+ mt->req_data_bytes, mt->reply_bytes, rz->iova))
+ RNP_PMD_LOG(ERR, "maintain request failed!\n");
+ else
+ RNP_PMD_LOG(INFO, "maintail request done!\n");
+
+ /* ==== update cfg */
+ if (fsz > FW_256KB) {
+ mt->magic = MAINTAIN_MAGIC;
+ mt->cmd = MT_WRITE_FLASH;
+ mt->arg0 = 2;
+ mt->req_data_bytes = 4096;
+ mt->reply_bytes = 0;
+ memcpy(mt->data, mt->data + FW_256KB, mt->req_data_bytes);
+
+ if (rnp_maintain_req(adapter->eth_dev,
+ mt->cmd, mt->arg0, mt->req_data_bytes,
+ mt->reply_bytes, rz->iova))
+ RNP_PMD_LOG(ERR, "maintain request failed!\n");
+ else
+ RNP_PMD_LOG(INFO, "maintail request done!\n");
+ }
+
+ RNP_PMD_LOG(NOTICE, "done\n");
+ fflush(stdout);
+
+ rte_memzone_free(rz);
+
+ exit(0);
+
+ return 0;
+}
+
+static int rnp_mbx_set_dump(struct rte_eth_dev *dev, int flag)
+{
+ struct rnp_eth_port *port = RNP_DEV_TO_PORT(dev);
+ struct mbx_fw_cmd_req req;
+ int err;
+
+ memset(&req, 0, sizeof(req));
+ build_set_dump(&req, port->attr.nr_lane, flag);
+
+ err = rnp_mbx_write_posted_locked(dev, &req);
+
+ return err;
+}
+
+int rnp_hw_set_fw_10g_1g_auto_detch(struct rte_eth_dev *dev, int enable)
+{
+ return rnp_mbx_set_dump(dev, 0x01140000 | (enable & 1));
+}
+
+int rnp_hw_set_fw_force_speed_1g(struct rte_eth_dev *dev, int enable)
+{
+ return rnp_mbx_set_dump(dev, 0x01150000 | (enable & 1));
+}
diff --git a/drivers/net/rnp/rnp_mbx_fw.h b/drivers/net/rnp/rnp_mbx_fw.h
index c4a04a3f09..39fb0f6ef3 100644
--- a/drivers/net/rnp/rnp_mbx_fw.h
+++ b/drivers/net/rnp/rnp_mbx_fw.h
@@ -16,6 +16,17 @@ struct mbx_req_cookie {
int priv_len;
char priv[RNP_MAX_SHARE_MEM];
};
+struct maintain_req {
+ int magic;
+#define MAINTAIN_MAGIC 0xa6a7a8a9
+
+ int cmd;
+ int arg0;
+ int req_data_bytes;
+ int reply_bytes;
+ char data[0];
+} __rte_packed;
+
enum GENERIC_CMD {
/* link configuration admin commands */
GET_PHY_ABALITY = 0x0601,
@@ -23,6 +34,9 @@ enum GENERIC_CMD {
RESET_PHY = 0x0603,
GET_LANE_STATUS = 0x0610,
SET_EVENT_MASK = 0x0613,
+ /* fw update */
+ FW_MAINTAIN = 0x0701,
+ SET_DUMP = 0x0a10,
};
enum link_event_mask {
@@ -211,6 +225,21 @@ struct mbx_fw_cmd_req {
struct {
int nr_lane;
} get_lane_st;
+
+ struct {
+ int cmd;
+#define MT_WRITE_FLASH 1
+ int arg0;
+ int req_bytes;
+ int reply_bytes;
+ int ddr_lo;
+ int ddr_hi;
+ } maintain;
+
+ struct {
+ int flag;
+ int nr_lane;
+ } set_dump;
};
} __rte_packed __rte_aligned(4);
@@ -284,6 +313,43 @@ build_get_lane_status_req(struct mbx_fw_cmd_req *req,
req->get_lane_st.nr_lane = nr_lane;
}
+static inline void
+build_maintain_req(struct mbx_fw_cmd_req *req,
+ void *cookie,
+ int cmd,
+ int arg0,
+ int req_bytes,
+ int reply_bytes,
+ u32 dma_phy_lo,
+ u32 dma_phy_hi)
+{
+ req->flags = 0;
+ req->opcode = FW_MAINTAIN;
+ req->datalen = sizeof(req->maintain);
+ req->cookie = cookie;
+ req->reply_lo = 0;
+ req->reply_hi = 0;
+ req->maintain.cmd = cmd;
+ req->maintain.arg0 = arg0;
+ req->maintain.req_bytes = req_bytes;
+ req->maintain.reply_bytes = reply_bytes;
+ req->maintain.ddr_lo = dma_phy_lo;
+ req->maintain.ddr_hi = dma_phy_hi;
+}
+
+static inline void
+build_set_dump(struct mbx_fw_cmd_req *req, int nr_lane, int flag)
+{
+ req->flags = 0;
+ req->opcode = SET_DUMP;
+ req->datalen = sizeof(req->set_dump);
+ req->cookie = NULL;
+ req->reply_lo = 0;
+ req->reply_hi = 0;
+ req->set_dump.flag = flag;
+ req->set_dump.nr_lane = nr_lane;
+}
+
int rnp_mbx_get_capability(struct rte_eth_dev *dev,
int *lane_mask,
int *nic_mode);
@@ -295,4 +361,7 @@ rnp_fw_get_macaddr(struct rte_eth_dev *dev,
u8 *mac_addr,
int nr_lane);
int rnp_mbx_get_lane_stat(struct rte_eth_dev *dev);
+int rnp_fw_update(struct rnp_eth_adapter *adapter);
+int rnp_hw_set_fw_10g_1g_auto_detch(struct rte_eth_dev *dev, int enable);
+int rnp_hw_set_fw_force_speed_1g(struct rte_eth_dev *dev, int enable);
#endif /* __RNP_MBX_FW_H__*/
--
2.27.0
^ permalink raw reply related [flat|nested] 27+ messages in thread* Re: [PATCH v6 7/8] net/rnp add devargs runtime parsing functions
2023-09-01 2:30 ` [PATCH v6 7/8] net/rnp add devargs runtime parsing functions Wenbo Cao
@ 2023-09-05 15:46 ` Ferruh Yigit
2023-09-06 9:13 ` 11
0 siblings, 1 reply; 27+ messages in thread
From: Ferruh Yigit @ 2023-09-05 15:46 UTC (permalink / raw)
To: Wenbo Cao; +Cc: dev, thomas, andrew.rybchenko, yaojun
On 9/1/2023 3:30 AM, Wenbo Cao wrote:
> add various runtime devargs command line options
> supported by this driver.
>
> Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
<...>
>
> +#define RNP_HW_MAC_LOOPBACK_ARG "hw_loopback"
> +#define RNP_FW_UPDATE "fw_update"
> +#define RNP_RX_FUNC_SELECT "rx_func_sec"
> +#define RNP_TX_FUNC_SELECT "tx_func_sec"
> +#define RNP_FW_4X10G_10G_1G_DET "fw_4x10g_10g_1g_auto_det"
> +#define RNP_FW_FORCE_SPEED_1G "fw_force_1g_speed"
> +
Please document these runtime arguments in the device document.
Also please add 'RTE_PMD_REGISTER_PARAM_STRING()' macros to document
argument for pmdinfogen, please see samples in existing code.
<...>
> +
> +static int
> +rnp_parse_io_select_func(const char *key, const char *value, void *extra_args)
> +{
> + uint8_t select = RNP_IO_FUNC_USE_NONE;
> +
> + RTE_SET_USED(key);
> +
> + if (strcmp(value, "vec") == 0)
> + select = RNP_IO_FUNC_USE_VEC;
> + else if (strcmp(value, "simple") == 0)
> + select = RNP_IO_FUNC_USE_SIMPLE;
> + else if (strcmp(value, "common") == 0)
> + select = RNP_IO_FUNC_USE_COMMON;
> +
There is already an generic eal argument that lets you select between
vector and scalar datapath implementation:
--force-max-simd-bitwidth=<val>
<...>
> +static int rnp_post_handle(struct rnp_eth_adapter *adapter)
> +{
> + bool on = false;
> +
> + if (!adapter->eth_dev)
> + return -ENOMEM;
> + if (adapter->do_fw_update && adapter->fw_path) {
> + rnp_fw_update(adapter);
> + adapter->do_fw_update = 0;
This patch also enables FW upgrade, can you please detail this in the
commit log? Or even you can consider to split this part into separate patch.
^ permalink raw reply [flat|nested] 27+ messages in thread* RE: [PATCH v6 7/8] net/rnp add devargs runtime parsing functions
2023-09-05 15:46 ` Ferruh Yigit
@ 2023-09-06 9:13 ` 11
0 siblings, 0 replies; 27+ messages in thread
From: 11 @ 2023-09-06 9:13 UTC (permalink / raw)
To: 'Ferruh Yigit'; +Cc: dev, thomas, andrew.rybchenko, yaojun
> -----Original Message-----
> From: Ferruh Yigit <ferruh.yigit@amd.com>
> Sent: 2023年9月5日 23:47
> To: Wenbo Cao <caowenbo@mucse.com>
> Cc: dev@dpdk.org; thomas@monjalon.net; andrew.rybchenko@oktetlabs.ru;
> yaojun@mucse.com
> Subject: Re: [PATCH v6 7/8] net/rnp add devargs runtime parsing functions
>
> On 9/1/2023 3:30 AM, Wenbo Cao wrote:
> > add various runtime devargs command line options supported by this
> > driver.
> >
> > Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
>
> <...>
>
> >
> > +#define RNP_HW_MAC_LOOPBACK_ARG "hw_loopback"
> > +#define RNP_FW_UPDATE "fw_update"
> > +#define RNP_RX_FUNC_SELECT "rx_func_sec"
> > +#define RNP_TX_FUNC_SELECT "tx_func_sec"
> > +#define RNP_FW_4X10G_10G_1G_DET "fw_4x10g_10g_1g_auto_det"
> > +#define RNP_FW_FORCE_SPEED_1G "fw_force_1g_speed"
> > +
>
> Please document these runtime arguments in the device document.
>
> Also please add 'RTE_PMD_REGISTER_PARAM_STRING()' macros to document
> argument for pmdinfogen, please see samples in existing code.
>
> <...>
>
OK, I will check the samples.
> > +
> > +static int
> > +rnp_parse_io_select_func(const char *key, const char *value, void
> > +*extra_args) {
> > + uint8_t select = RNP_IO_FUNC_USE_NONE;
> > +
> > + RTE_SET_USED(key);
> > +
> > + if (strcmp(value, "vec") == 0)
> > + select = RNP_IO_FUNC_USE_VEC;
> > + else if (strcmp(value, "simple") == 0)
> > + select = RNP_IO_FUNC_USE_SIMPLE;
> > + else if (strcmp(value, "common") == 0)
> > + select = RNP_IO_FUNC_USE_COMMON;
> > +
>
> There is already an generic eal argument that lets you select between vector and
> scalar datapath implementation:
> --force-max-simd-bitwidth=<val>
>
> <...>
>
> --force-max-simd-bitwidth=<val>
This option I have noticed before. Thanks a lot.
I will check the generic eal argume code, and reedit the interrelated code
> > +static int rnp_post_handle(struct rnp_eth_adapter *adapter) {
> > + bool on = false;
> > +
> > + if (!adapter->eth_dev)
> > + return -ENOMEM;
> > + if (adapter->do_fw_update && adapter->fw_path) {
> > + rnp_fw_update(adapter);
> > + adapter->do_fw_update = 0;
>
> This patch also enables FW upgrade, can you please detail this in the commit log?
> Or even you can consider to split this part into separate patch.
>
>
>
For this, I will split it to another commit.
^ permalink raw reply [flat|nested] 27+ messages in thread
* [PATCH v6 8/8] net/rnp handle device interrupts
2023-09-01 2:30 [PATCH v6 0/8] [v6]drivers/net Add Support mucse N10 Pmd Driver Wenbo Cao
` (6 preceding siblings ...)
2023-09-01 2:30 ` [PATCH v6 7/8] net/rnp add devargs runtime parsing functions Wenbo Cao
@ 2023-09-01 2:30 ` Wenbo Cao
2023-09-05 15:34 ` [PATCH v6 0/8] [v6]drivers/net Add Support mucse N10 Pmd Driver Ferruh Yigit
8 siblings, 0 replies; 27+ messages in thread
From: Wenbo Cao @ 2023-09-01 2:30 UTC (permalink / raw)
To: Wenbo Cao; +Cc: dev, ferruh.yigit, thomas, andrew.rybchenko, yaojun
Handle device lsc interrupt event
Signed-off-by: Wenbo Cao <caowenbo@mucse.com>
---
drivers/net/rnp/base/rnp_hw.h | 5 +
drivers/net/rnp/base/rnp_mac_regs.h | 279 ++++++++++++++++++++++++++++
drivers/net/rnp/rnp.h | 8 +
drivers/net/rnp/rnp_ethdev.c | 17 ++
drivers/net/rnp/rnp_mbx.h | 3 +-
drivers/net/rnp/rnp_mbx_fw.c | 233 +++++++++++++++++++++++
drivers/net/rnp/rnp_mbx_fw.h | 38 +++-
7 files changed, 580 insertions(+), 3 deletions(-)
create mode 100644 drivers/net/rnp/base/rnp_mac_regs.h
diff --git a/drivers/net/rnp/base/rnp_hw.h b/drivers/net/rnp/base/rnp_hw.h
index 395b9d5c71..5c50484c6c 100644
--- a/drivers/net/rnp/base/rnp_hw.h
+++ b/drivers/net/rnp/base/rnp_hw.h
@@ -10,6 +10,7 @@
#include "rnp_osdep.h"
#include "rnp_dma_regs.h"
#include "rnp_eth_regs.h"
+#include "rnp_mac_regs.h"
#include "rnp_cfg.h"
static inline unsigned int rnp_rd_reg(volatile void *addr)
@@ -48,6 +49,10 @@ static inline void rnp_wr_reg(volatile void *reg, int val)
rnp_eth_wr(hw, RNP_RAL_BASE_ADDR(hw_idx), val)
#define RNP_MACADDR_UPDATE_HI(hw, hw_idx, val) \
rnp_eth_wr(hw, RNP_RAH_BASE_ADDR(hw_idx), val)
+#define rnp_mac_rd(hw, id, off) \
+ rnp_rd_reg((char *)(hw)->mac_base[id] + (off))
+#define rnp_mac_wr(hw, id, off, val) \
+ rnp_wr_reg((char *)(hw)->mac_base[id] + (off), val)
struct rnp_hw;
/* Mbx Operate info */
enum MBX_ID {
diff --git a/drivers/net/rnp/base/rnp_mac_regs.h b/drivers/net/rnp/base/rnp_mac_regs.h
new file mode 100644
index 0000000000..f9466b3841
--- /dev/null
+++ b/drivers/net/rnp/base/rnp_mac_regs.h
@@ -0,0 +1,279 @@
+#ifndef __RNP_MAC_REGS_H__
+#define __RNP_MAC_REGS_H__
+
+#include "rnp_osdep.h"
+#define RNP_MAC_TX_CFG (0x0)
+
+/* Transmitter Enable */
+#define RNP_MAC_TE BIT(0)
+/* Jabber Disable */
+#define RNP_MAC_JD BIT(16)
+#define RNP_SPEED_SEL_1G (BIT(30) | BIT(29) | BIT(28))
+#define RNP_SPEED_SEL_10G BIT(30)
+#define RNP_SPEED_SEL_40G (0)
+#define RNP_MAC_RX_CFG (0x4)
+/* Receiver Enable */
+#define RNP_MAC_RE BIT(0)
+/* Automatic Pad or CRC Stripping */
+#define RNP_MAC_ACS BIT(1)
+/* CRC stripping for Type packets */
+#define RNP_MAC_CST BIT(2)
+/* Disable CRC Check */
+#define RNP_MAC_DCRCC BIT(3)
+/* Enable Max Frame Size Limit */
+#define RNP_MAC_GPSLCE BIT(6)
+/* Watchdog Disable */
+#define RNP_MAC_WD BIT(7)
+/* Jumbo Packet Support En */
+#define RNP_MAC_JE BIT(8)
+/* Loopback Mode */
+#define RNP_MAC_LM BIT(10)
+/* Giant Packet Size Limit */
+#define RNP_MAC_GPSL_MASK GENMASK(29, 16)
+#define RNP_MAC_MAX_GPSL (1518)
+#define RNP_MAC_CPSL_SHIFT (16)
+
+#define RNP_MAC_PKT_FLT_CTRL (0x8)
+
+/* Receive All */
+#define RNP_MAC_RA BIT(31)
+/* Pass Control Packets */
+#define RNP_MAC_PCF GENMASK(7, 6)
+#define RNP_MAC_PCF_OFFSET (6)
+/* Mac Filter ALL Ctrl Frame */
+#define RNP_MAC_PCF_FAC (0)
+/* Mac Forward ALL Ctrl Frame Except Pause */
+#define RNP_MAC_PCF_NO_PAUSE (1)
+/* Mac Forward All Ctrl Pkt */
+#define RNP_MAC_PCF_PA (2)
+/* Mac Forward Ctrl Frame Match Unicast */
+#define RNP_MAC_PCF_PUN (3)
+/* Promiscuous Mode */
+#define RNP_MAC_PROMISC_EN BIT(0)
+/* Hash Unicast */
+#define RNP_MAC_HUC BIT(1)
+/* Hash Multicast */
+#define RNP_MAC_HMC BIT(2)
+/* Pass All Multicast */
+#define RNP_MAC_PM BIT(4)
+/* Disable Broadcast Packets */
+#define RNP_MAC_DBF BIT(5)
+/* Hash or Perfect Filter */
+#define RNP_MAC_HPF BIT(10)
+#define RNP_MAC_VTFE BIT(16)
+/* Interrupt Status */
+#define RNP_MAC_INT_STATUS _MAC_(0xb0)
+#define RNP_MAC_LS_MASK GENMASK(25, 24)
+#define RNP_MAC_LS_UP (0)
+#define RNP_MAC_LS_LOCAL_FAULT BIT(25)
+#define RNP_MAC_LS_REMOTE_FAULT (BIT(25) | BIT(24))
+/* Unicast Mac Hash Table */
+#define RNP_MAC_UC_HASH_TB(n) _MAC_(0x10 + ((n) * 0x4))
+
+
+#define RNP_MAC_LPI_CTRL (0xd0)
+
+/* PHY Link Status Disable */
+#define RNP_MAC_PLSDIS BIT(18)
+/* PHY Link Status */
+#define RNP_MAC_PLS BIT(17)
+
+/* MAC VLAN CTRL Strip REG */
+#define RNP_MAC_VLAN_TAG (0x50)
+
+/* En Inner VLAN Strip Action */
+#define RNP_MAC_EIVLS GENMASK(29, 28)
+/* Inner VLAN Strip Action Shift */
+#define RNP_MAC_IV_EIVLS_SHIFT (28)
+/* Inner Vlan Don't Strip*/
+#define RNP_MAC_IV_STRIP_NONE (0x0)
+/* Inner Vlan Strip When Filter Match Success */
+#define RNP_MAC_IV_STRIP_PASS (0x1)
+/* Inner Vlan STRIP When Filter Match FAIL */
+#define RNP_MAC_IV_STRIP_FAIL (0x2)
+/* Inner Vlan STRIP Always */
+#define RNP_MAC_IV_STRIP_ALL (0X3)
+/* VLAN Strip Mode Ctrl Shift */
+#define RNP_VLAN_TAG_CTRL_EVLS_SHIFT (21)
+/* En Double Vlan Processing */
+#define RNP_MAC_VLAN_EDVLP BIT(26)
+/* VLAN Tag Hash Table Match Enable */
+#define RNP_MAC_VLAN_VTHM BIT(25)
+/* Enable VLAN Tag in Rx status */
+#define RNP_MAC_VLAN_EVLRXS BIT(24)
+/* Disable VLAN Type Check */
+#define RNP_MAC_VLAN_DOVLTC BIT(20)
+/* Enable S-VLAN */
+#define RNP_MAC_VLAN_ESVL BIT(18)
+/* Enable 12-Bit VLAN Tag Comparison Filter */
+#define RNP_MAC_VLAN_ETV BIT(16)
+#define RNP_MAC_VLAN_HASH_EN GENMASK(15, 0)
+#define RNP_MAC_VLAN_VID GENMASK(15, 0)
+/* VLAN Don't Strip */
+#define RNP_MAC_VLAN_STRIP_NONE (0x0 << RNP_VLAN_TAG_CTRL_EVLS_SHIFT)
+/* VLAN Filter Success Then STRIP */
+#define RNP_MAC_VLAN_STRIP_PASS (0x1 << RNP_VLAN_TAG_CTRL_EVLS_SHIFT)
+/* VLAN Filter Failed Then STRIP */
+#define RNP_MAC_VLAN_STRIP_FAIL (0x2 << RNP_VLAN_TAG_CTRL_EVLS_SHIFT)
+/* All Vlan Will Strip */
+#define RNP_MAC_VLAN_STRIP_ALL (0x3 << RNP_VLAN_TAG_CTRL_EVLS_SHIFT)
+
+#define RNP_MAC_VLAN_HASH_TB (0x58)
+#define RNP_MAC_VLAN_HASH_MASK GENMASK(15, 0)
+
+/* MAC VLAN CTRL INSERT REG */
+#define RNP_MAC_VLAN_INCL (0x60)
+#define RNP_MAC_INVLAN_INCL (0x64)
+
+/* VLAN Tag Input */
+/* VLAN_Tag Insert From Description */
+#define RNP_MAC_VLAN_VLTI BIT(20)
+/* C-VLAN or S-VLAN */
+#define RNP_MAC_VLAN_CSVL BIT(19)
+#define RNP_MAC_VLAN_INSERT_CVLAN (0 << 19)
+#define RNP_MAC_VLAN_INSERT_SVLAN (1 << 19)
+/* VLAN Tag Control in Transmit Packets */
+#define RNP_MAC_VLAN_VLC GENMASK(17, 16)
+/* VLAN Tag Control Offset Bit */
+#define RNP_MAC_VLAN_VLC_SHIFT (16)
+/* Don't Anything ON TX VLAN*/
+#define RNP_MAC_VLAN_VLC_NONE (0x0 << RNP_MAC_VLAN_VLC_SHIFT)
+/* MAC Delete VLAN */
+#define RNP_MAC_VLAN_VLC_DEL (0x1 << RNP_MAC_VLAN_VLC_SHIFT)
+/* MAC Add VLAN */
+#define RNP_MAC_VLAN_VLC_ADD (0x2 << RNP_MAC_VLAN_VLC_SHIFT)
+/* MAC Replace VLAN */
+#define RNP_MAC_VLAN_VLC_REPLACE (0x3 << RNP_MAC_VLAN_VLC_SHIFT)
+/* VLAN Tag for Transmit Packets For Insert/Remove */
+#define RNP_MAC_VLAN_VLT GENMASK(15, 0)
+/* TX Peer TC Flow Ctrl */
+
+#define RNP_MAC_Q0_TX_FC(n) (0x70 + ((n) * 0x4))
+
+/* Edit Pause Time */
+#define RNP_MAC_FC_PT GENMASK(31, 16)
+#define RNP_MAC_FC_PT_OFFSET (16)
+/* Disable Zero-Quanta Pause */
+#define RNP_MAC_FC_DZPQ BIT(7)
+/* Pause Low Threshold */
+#define RNP_MAC_FC_PLT GENMASK(6, 4)
+#define RNP_MAC_FC_PLT_OFFSET (4)
+#define RNP_MAC_FC_PLT_4_SLOT (0)
+#define RNP_MAC_FC_PLT_28_SLOT (1)
+#define RNP_MAC_FC_PLT_36_SLOT (2)
+#define RNP_MAC_FC_PLT_144_SLOT (3)
+#define RNP_MAC_FC_PLT_256_SLOT (4)
+/* Transmit Flow Control Enable */
+#define RNP_MAC_FC_TEE BIT(1)
+/* Transmit Flow Control Busy Immediately */
+#define RNP_MAC_FC_FCB BIT(0)
+/* Mac RX Flow Ctrl*/
+
+#define RNP_MAC_RX_FC (0x90)
+
+/* Rx Priority Based Flow Control Enable */
+#define RNP_MAC_RX_FC_PFCE BIT(8)
+/* Unicast Pause Packet Detect */
+#define RNP_MAC_RX_FC_UP BIT(1)
+/* Receive Flow Control Enable */
+#define RNP_MAC_RX_FC_RFE BIT(0)
+
+/* Rx Mac Address Base */
+#define RNP_MAC_ADDR_DEF_HI _MAC_(0x0300)
+
+#define RNP_MAC_AE BIT(31)
+#define RNP_MAC_ADDR_LO(n) _MAC_((0x0304) + ((n) * 0x8))
+#define RNP_MAC_ADDR_HI(n) _MAC_((0x0300) + ((n) * 0x8))
+
+/* Mac Manage Counts */
+#define RNP_MMC_CTRL _MAC_(0x0800)
+#define RNP_MMC_RSTONRD BIT(2)
+/* Tx Good And Bad Bytes Base */
+#define RNP_MMC_TX_GBOCTGB _MAC_(0x0814)
+/* Tx Good And Bad Frame Num Base */
+#define RNP_MMC_TX_GBFRMB _MAC_(0x081c)
+/* Tx Good Broadcast Frame Num Base */
+#define RNP_MMC_TX_BCASTB _MAC_(0x0824)
+/* Tx Good Multicast Frame Num Base */
+#define RNP_MMC_TX_MCASTB _MAC_(0x082c)
+/* Tx 64Bytes Frame Num */
+#define RNP_MMC_TX_64_BYTESB _MAC_(0x0834)
+#define RNP_MMC_TX_65TO127_BYTESB _MAC_(0x083c)
+#define RNP_MMC_TX_128TO255_BYTEB _MAC_(0x0844)
+#define RNP_MMC_TX_256TO511_BYTEB _MAC_(0x084c)
+#define RNP_MMC_TX_512TO1023_BYTEB _MAC_(0x0854)
+#define RNP_MMC_TX_1024TOMAX_BYTEB _MAC_(0x085c)
+/* Tx Good And Bad Unicast Frame Num Base */
+#define RNP_MMC_TX_GBUCASTB _MAC_(0x0864)
+/* Tx Good And Bad Multicast Frame Num Base */
+#define RNP_MMC_TX_GBMCASTB _MAC_(0x086c)
+/* Tx Good And Bad Broadcast Frame NUM Base */
+#define RNP_MMC_TX_GBBCASTB _MAC_(0x0874)
+/* Tx Frame Underflow Error */
+#define RNP_MMC_TX_UNDRFLWB _MAC_(0x087c)
+/* Tx Good Frame Bytes Base */
+#define RNP_MMC_TX_GBYTESB _MAC_(0x0884)
+/* Tx Good Frame Num Base*/
+#define RNP_MMC_TX_GBRMB _MAC_(0x088c)
+/* Tx Good Pause Frame Num Base */
+#define RNP_MMC_TX_PAUSEB _MAC_(0x0894)
+/* Tx Good Vlan Frame Num Base */
+#define RNP_MMC_TX_VLANB _MAC_(0x089c)
+
+/* Rx Good And Bad Frames Num Base */
+#define RNP_MMC_RX_GBFRMB _MAC_(0x0900)
+/* Rx Good And Bad Frames Bytes Base */
+#define RNP_MMC_RX_GBOCTGB _MAC_(0x0908)
+/* Rx Good Framse Bytes Base */
+#define RNP_MMC_RX_GOCTGB _MAC_(0x0910)
+/* Rx Good Broadcast Frames Num Base */
+#define RNP_MMC_RX_BCASTGB _MAC_(0x0918)
+/* Rx Good Multicast Frames Num Base */
+#define RNP_MMC_RX_MCASTGB _MAC_(0x0920)
+/* Rx Crc Error Frames Num Base */
+#define RNP_MMC_RX_CRCERB _MAC_(0x0928)
+/* Rx Less Than 64Byes with Crc Err Base*/
+#define RNP_MMC_RX_RUNTERB _MAC_(0x0930)
+/* Receive Jumbo Frame Error */
+#define RNP_MMC_RX_JABBER_ERR _MAC_(0x0934)
+/* Shorter Than 64Bytes without Any Errora Base */
+#define RNP_MMC_RX_USIZEGB _MAC_(0x0938)
+/* Len Oversize Than Support */
+#define RNP_MMC_RX_OSIZEGB _MAC_(0x093c)
+/* Rx 64Byes Frame Num Base */
+#define RNP_MMC_RX_64_BYTESB _MAC_(0x0940)
+/* Rx 65Bytes To 127Bytes Frame Num Base */
+#define RNP_MMC_RX_65TO127_BYTESB _MAC_(0x0948)
+/* Rx 128Bytes To 255Bytes Frame Num Base */
+#define RNP_MMC_RX_128TO255_BYTESB _MAC_(0x0950)
+/* Rx 256Bytes To 511Bytes Frame Num Base */
+#define RNP_MMC_RX_256TO511_BYTESB _MAC_(0x0958)
+/* Rx 512Bytes To 1023Bytes Frame Num Base */
+#define RNP_MMC_RX_512TO1203_BYTESB _MAC_(0x0960)
+/* Rx Len Bigger Than 1024Bytes Base */
+#define RNP_MMC_RX_1024TOMAX_BYTESB _MAC_(0x0968)
+/* Rx Unicast Frame Good Num Base */
+#define RNP_MMC_RX_UCASTGB _MAC_(0x0970)
+/* Rx Length Error Of Frame Part */
+#define RNP_MMC_RX_LENERRB _MAC_(0x0978)
+/* Rx received with a Length field not equal to the valid frame size */
+#define RNP_MMC_RX_OUTOF_RANGE _MAC_(0x0980)
+/* Rx Pause Frame Good Num Base */
+#define RNP_MMC_RX_PAUSEB _MAC_(0x0988)
+/* Rx Vlan Frame Good Num Base */
+#define RNP_MMC_RX_VLANGB _MAC_(0x0998)
+/* Rx With A Watchdog Timeout Err Frame Base */
+#define RNP_MMC_RX_WDOGERRB _MAC_(0x09a0)
+
+/* 1588 */
+#define RNP_MAC_TS_CTRL _MAC_(0X0d00)
+#define RNP_MAC_SUB_SECOND_INCREMENT _MAC_(0x0d04)
+#define RNP_MAC_SYS_TIME_SEC_CFG _MAC_(0x0d08)
+#define RNP_MAC_SYS_TIME_NANOSEC_CFG _MAC_(0x0d0c)
+#define RNP_MAC_SYS_TIME_SEC_UPDATE _MAC_(0x0d10)
+#define RNP_MAC_SYS_TIME_NANOSEC_UPDATE _MAC_(0x0d14)
+#define RNP_MAC_TS_ADDEND _MAC_(0x0d18)
+#define RNP_MAC_TS_STATS _MAC_(0x0d20)
+#define RNP_MAC_INTERRUPT_ENABLE _MAC_(0x00b4)
+
+#endif /* __RNP_MAC_REGS_H__ */
diff --git a/drivers/net/rnp/rnp.h b/drivers/net/rnp/rnp.h
index 7e797cf177..413c8b0a73 100644
--- a/drivers/net/rnp/rnp.h
+++ b/drivers/net/rnp/rnp.h
@@ -111,6 +111,8 @@ struct rnp_eth_port {
uint8_t tx_func_sec; /* force set io tx func */
struct rte_eth_dev *eth_dev;
struct rnp_port_attr attr;
+ uint64_t state;
+ rte_spinlock_t rx_mac_lock; /* Lock For Mac_cfg resource write */
/* Recvice Mac Address Record Table */
uint8_t mac_use_tb[RNP_MAX_MAC_ADDRS];
uint8_t use_num_mac;
@@ -131,6 +133,12 @@ enum {
RNP_IO_FUNC_USE_COMMON,
};
+enum rnp_port_state {
+ RNP_PORT_STATE_PAUSE = 0,
+ RNP_PORT_STATE_FINISH,
+ RNP_PORT_STATE_SETTING,
+};
+
struct rnp_eth_adapter {
enum rnp_work_mode mode;
enum rnp_resource_share_m s_mode; /* Port Resource Share Policy */
diff --git a/drivers/net/rnp/rnp_ethdev.c b/drivers/net/rnp/rnp_ethdev.c
index 131da42259..22adfbeabc 100644
--- a/drivers/net/rnp/rnp_ethdev.c
+++ b/drivers/net/rnp/rnp_ethdev.c
@@ -601,10 +601,23 @@ static int rnp_post_handle(struct rnp_eth_adapter *adapter)
return 0;
}
+static void rnp_dev_interrupt_handler(void *param)
+{
+ struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
+ struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+ struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
+ struct rnp_eth_adapter *adapter = RNP_DEV_TO_ADAPTER(dev);
+
+ rte_intr_disable(intr_handle);
+ rnp_fw_msg_handler(adapter);
+ rte_intr_enable(intr_handle);
+}
+
static int
rnp_eth_dev_init(struct rte_eth_dev *dev)
{
struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+ struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
struct rnp_eth_port *port = RNP_DEV_TO_PORT(dev);
struct rnp_eth_adapter *adapter = NULL;
char name[RTE_ETH_NAME_MAX_LEN] = " ";
@@ -680,6 +693,10 @@ rnp_eth_dev_init(struct rte_eth_dev *dev)
rnp_mac_rx_disable(eth_dev);
rnp_mac_tx_disable(eth_dev);
}
+ rte_intr_disable(intr_handle);
+ /* Enable Link Update Event Interrupt */
+ rte_intr_callback_register(intr_handle,
+ rnp_dev_interrupt_handler, dev);
ret = rnp_post_handle(adapter);
if (ret)
goto eth_alloc_error;
diff --git a/drivers/net/rnp/rnp_mbx.h b/drivers/net/rnp/rnp_mbx.h
index 87949c1726..d6b78e32a7 100644
--- a/drivers/net/rnp/rnp_mbx.h
+++ b/drivers/net/rnp/rnp_mbx.h
@@ -13,7 +13,8 @@
/* Mbx Ctrl state */
#define RNP_VFMAILBOX_SIZE (14) /* 16 32 bit words - 64 bytes */
-#define TSRN10_VFMBX_SIZE (RNP_VFMAILBOX_SIZE)
+#define RNP_FW_MAILBOX_SIZE RNP_VFMAILBOX_SIZE
+#define RNP_VFMBX_SIZE (RNP_VFMAILBOX_SIZE)
#define RNP_VT_MSGTYPE_ACK (0x80000000)
#define RNP_VT_MSGTYPE_NACK (0x40000000)
diff --git a/drivers/net/rnp/rnp_mbx_fw.c b/drivers/net/rnp/rnp_mbx_fw.c
index 0c3f499cf2..a0a163e98c 100644
--- a/drivers/net/rnp/rnp_mbx_fw.c
+++ b/drivers/net/rnp/rnp_mbx_fw.c
@@ -545,3 +545,236 @@ int rnp_hw_set_fw_force_speed_1g(struct rte_eth_dev *dev, int enable)
{
return rnp_mbx_set_dump(dev, 0x01150000 | (enable & 1));
}
+
+static inline int
+rnp_mbx_fw_reply_handler(struct rnp_eth_adapter *adapter __rte_unused,
+ struct mbx_fw_cmd_reply *reply)
+{
+ struct mbx_req_cookie *cookie;
+ /* dbg_here; */
+ cookie = reply->cookie;
+ if (!cookie || cookie->magic != COOKIE_MAGIC) {
+ RNP_PMD_LOG(ERR,
+ "[%s] invalid cookie:%p opcode: "
+ "0x%x v0:0x%x\n",
+ __func__,
+ cookie,
+ reply->opcode,
+ *((int *)reply));
+ return -EIO;
+ }
+
+ if (cookie->priv_len > 0)
+ memcpy(cookie->priv, reply->data, cookie->priv_len);
+
+ cookie->done = 1;
+
+ if (reply->flags & FLAGS_ERR)
+ cookie->errcode = reply->error_code;
+ else
+ cookie->errcode = 0;
+
+ return 0;
+}
+
+void rnp_link_stat_mark(struct rnp_hw *hw, int nr_lane, int up)
+{
+ u32 v;
+
+ rte_spinlock_lock(&hw->fw_lock);
+ v = rnp_rd_reg(hw->link_sync);
+ v &= ~(0xffff0000);
+ v |= 0xa5a40000;
+ if (up)
+ v |= BIT(nr_lane);
+ else
+ v &= ~BIT(nr_lane);
+ rnp_wr_reg(hw->link_sync, v);
+
+ rte_spinlock_unlock(&hw->fw_lock);
+}
+
+void rnp_link_report(struct rte_eth_dev *dev, bool link_en)
+{
+ struct rnp_eth_port *port = RNP_DEV_TO_PORT(dev);
+ struct rnp_hw *hw = RNP_DEV_TO_HW(dev);
+ struct rte_eth_link link;
+
+ link.link_duplex = link_en ? port->attr.phy_meta.link_duplex :
+ RTE_ETH_LINK_FULL_DUPLEX;
+ link.link_status = link_en ? RTE_ETH_LINK_UP : RTE_ETH_LINK_DOWN;
+ link.link_speed = link_en ? port->attr.speed :
+ RTE_ETH_SPEED_NUM_UNKNOWN;
+ RNP_PMD_LOG(INFO,
+ "\nPF[%d]link changed: changed_lane:0x%x, "
+ "status:0x%x\n",
+ hw->pf_vf_num & RNP_PF_NB_MASK ? 1 : 0,
+ port->attr.nr_port,
+ link_en);
+ link.link_autoneg = port->attr.phy_meta.link_autoneg
+ ? RTE_ETH_LINK_SPEED_AUTONEG
+ : RTE_ETH_LINK_SPEED_FIXED;
+ /* Report Link Info To Upper Firmwork */
+ rte_eth_linkstatus_set(dev, &link);
+ /* Notice Event Process Link Status Change */
+ rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
+ /* Notce Firmware LSC Event SW Received */
+ rnp_link_stat_mark(hw, port->attr.nr_port, link_en);
+}
+
+static void rnp_dev_alarm_link_handler(void *param)
+{
+ struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
+ struct rnp_eth_port *port = RNP_DEV_TO_PORT(dev);
+ uint32_t status;
+
+ status = port->attr.link_ready;
+ rnp_link_report(dev, status);
+}
+
+static void rnp_link_event(struct rnp_eth_adapter *adapter,
+ struct mbx_fw_cmd_req *req)
+{
+ struct rnp_hw *hw = &adapter->hw;
+ struct rnp_eth_port *port;
+ bool link_change = false;
+ uint32_t lane_bit;
+ uint32_t sync_bit;
+ uint32_t link_en;
+ uint32_t ctrl;
+ int i;
+
+ for (i = 0; i < adapter->num_ports; i++) {
+ port = adapter->ports[i];
+ if (port == NULL)
+ continue;
+ link_change = false;
+ lane_bit = port->attr.nr_port;
+ if (__atomic_load_n(&port->state, __ATOMIC_RELAXED)
+ != RNP_PORT_STATE_FINISH)
+ continue;
+ if (!(BIT(lane_bit) & req->link_stat.changed_lanes))
+ continue;
+ link_en = BIT(lane_bit) & req->link_stat.lane_status;
+ sync_bit = BIT(lane_bit) & rnp_rd_reg(hw->link_sync);
+
+ if (link_en) {
+ /* Port Link Change To Up */
+ if (!port->attr.link_ready) {
+ link_change = true;
+ port->attr.link_ready = true;
+ }
+ if (req->link_stat.port_st_magic == SPEED_VALID_MAGIC) {
+ port->attr.speed = req->link_stat.st[lane_bit].speed;
+ port->attr.phy_meta.link_duplex =
+ req->link_stat.st[lane_bit].duplex;
+ port->attr.phy_meta.link_autoneg =
+ req->link_stat.st[lane_bit].autoneg;
+ RNP_PMD_INIT_LOG(INFO,
+ "phy_id %d speed %d duplex "
+ "%d issgmii %d PortID %d\n",
+ req->link_stat.st[lane_bit].phy_addr,
+ req->link_stat.st[lane_bit].speed,
+ req->link_stat.st[lane_bit].duplex,
+ req->link_stat.st[lane_bit].is_sgmii,
+ port->attr.rte_pid);
+ }
+ } else {
+ /* Port Link to Down */
+ if (port->attr.link_ready) {
+ link_change = true;
+ port->attr.link_ready = false;
+ }
+ }
+ if (link_change || sync_bit != link_en) {
+ /* WorkAround For Hardware When Link Down
+ * Eth Module Tx-side Can't Drop In some condition
+ * So back The Packet To Rx Side To Drop Packet
+ */
+ /* To Protect Conflict Hw Resource */
+ rte_spinlock_lock(&port->rx_mac_lock);
+ ctrl = rnp_mac_rd(hw, lane_bit, RNP_MAC_RX_CFG);
+ if (port->attr.link_ready) {
+ ctrl &= ~RNP_MAC_LM;
+ rnp_eth_wr(hw,
+ RNP_RX_FIFO_FULL_THRETH(lane_bit),
+ RNP_RX_DEFAULT_VAL);
+ } else {
+ rnp_eth_wr(hw,
+ RNP_RX_FIFO_FULL_THRETH(lane_bit),
+ RNP_RX_WORKAROUND_VAL);
+ ctrl |= RNP_MAC_LM;
+ }
+ rnp_mac_wr(hw, lane_bit, RNP_MAC_RX_CFG, ctrl);
+ rte_spinlock_unlock(&port->rx_mac_lock);
+ rte_eal_alarm_set(RNP_ALARM_INTERVAL,
+ rnp_dev_alarm_link_handler,
+ (void *)port->eth_dev);
+ }
+ }
+}
+
+static inline int
+rnp_mbx_fw_req_handler(struct rnp_eth_adapter *adapter,
+ struct mbx_fw_cmd_req *req)
+{
+ switch (req->opcode) {
+ case LINK_STATUS_EVENT:
+ rnp_link_event(adapter, req);
+ break;
+ default:
+ break;
+ }
+
+ return 0;
+}
+
+static inline int rnp_rcv_msg_from_fw(struct rnp_eth_adapter *adapter)
+{
+ const struct rnp_mbx_api *ops = RNP_DEV_TO_MBX_OPS(adapter->eth_dev);
+ struct rnp_hw *hw = &adapter->hw;
+ u32 msgbuf[RNP_FW_MAILBOX_SIZE];
+ uint16_t check_state;
+ int retval;
+
+ retval = ops->read(hw, msgbuf, RNP_FW_MAILBOX_SIZE, MBX_FW);
+ if (retval) {
+ PMD_DRV_LOG(ERR, "Error receiving message from FW\n");
+ return retval;
+ }
+#define RNP_MBX_SYNC_MASK GENMASK(15, 0)
+
+ check_state = msgbuf[0] & RNP_MBX_SYNC_MASK;
+ /* this is a message we already processed, do nothing */
+ if (check_state & FLAGS_DD)
+ return rnp_mbx_fw_reply_handler(adapter,
+ (struct mbx_fw_cmd_reply *)msgbuf);
+ else
+ return rnp_mbx_fw_req_handler(adapter,
+ (struct mbx_fw_cmd_req *)msgbuf);
+
+ return 0;
+}
+
+static void rnp_rcv_ack_from_fw(struct rnp_eth_adapter *adapter)
+{
+ struct rnp_hw *hw __rte_unused = &adapter->hw;
+ u32 msg __rte_unused = RNP_VT_MSGTYPE_NACK;
+ /* do-nothing */
+}
+
+int rnp_fw_msg_handler(struct rnp_eth_adapter *adapter)
+{
+ const struct rnp_mbx_api *ops = RNP_DEV_TO_MBX_OPS(adapter->eth_dev);
+ struct rnp_hw *hw = &adapter->hw;
+
+ /* == check cpureq */
+ if (!ops->check_for_msg(hw, MBX_FW))
+ rnp_rcv_msg_from_fw(adapter);
+
+ /* process any acks */
+ if (!ops->check_for_ack(hw, MBX_FW))
+ rnp_rcv_ack_from_fw(adapter);
+
+ return 0;
+}
diff --git a/drivers/net/rnp/rnp_mbx_fw.h b/drivers/net/rnp/rnp_mbx_fw.h
index 39fb0f6ef3..e0174c1d22 100644
--- a/drivers/net/rnp/rnp_mbx_fw.h
+++ b/drivers/net/rnp/rnp_mbx_fw.h
@@ -32,6 +32,8 @@ enum GENERIC_CMD {
GET_PHY_ABALITY = 0x0601,
GET_MAC_ADDRESS = 0x0602,
RESET_PHY = 0x0603,
+ GET_LINK_STATUS = 0x0607,
+ LINK_STATUS_EVENT = 0x0608,
GET_LANE_STATUS = 0x0610,
SET_EVENT_MASK = 0x0613,
/* fw update */
@@ -98,6 +100,21 @@ struct phy_abilities {
};
} __rte_packed __rte_aligned(4);
+struct port_stat {
+ u8 phy_addr; /* Phy MDIO address */
+
+ u8 duplex : 1; /* FIBRE is always 1,Twisted Pair 1 or 0 */
+ u8 autoneg : 1; /* autoned state */
+ u8 fec : 1;
+ u8 an_rev : 1;
+ u8 link_traing : 1;
+ u8 is_sgmii : 1; /* avild fw >= 0.5.0.17 */
+ u16 speed; /* cur port linked speed */
+
+ u16 pause : 4;
+ u16 rev : 12;
+} __rte_packed;
+
#define RNP_SPEED_CAP_UNKNOWN (0)
#define RNP_SPEED_CAP_10M_FULL BIT(2)
#define RNP_SPEED_CAP_100M_FULL BIT(3)
@@ -186,8 +203,14 @@ struct mbx_fw_cmd_reply {
struct phy_abilities phy_abilities;
};
} __rte_packed __rte_aligned(4);
-
-#define MBX_REQ_HDR_LEN 24
+/* == flags == */
+#define FLAGS_DD BIT(0) /* driver clear 0, FW must set 1 */
+#define FLAGS_CMP BIT(1) /* driver clear 0, FW mucst set */
+/* driver clear 0, FW must set only if it reporting an error */
+#define FLAGS_ERR BIT(2)
+
+#define MBX_REQ_HDR_LEN (24)
+#define RNP_ALARM_INTERVAL (50000) /* unit us */
/* driver -> firmware */
struct mbx_fw_cmd_req {
unsigned short flags; /* 0-1 */
@@ -240,6 +263,14 @@ struct mbx_fw_cmd_req {
int flag;
int nr_lane;
} set_dump;
+
+ struct {
+ unsigned short changed_lanes;
+ unsigned short lane_status;
+ unsigned int port_st_magic;
+#define SPEED_VALID_MAGIC 0xa4a6a8a9
+ struct port_stat st[4];
+ } link_stat; /* FW->RC */
};
} __rte_packed __rte_aligned(4);
@@ -364,4 +395,7 @@ int rnp_mbx_get_lane_stat(struct rte_eth_dev *dev);
int rnp_fw_update(struct rnp_eth_adapter *adapter);
int rnp_hw_set_fw_10g_1g_auto_detch(struct rte_eth_dev *dev, int enable);
int rnp_hw_set_fw_force_speed_1g(struct rte_eth_dev *dev, int enable);
+void rnp_link_stat_mark(struct rnp_hw *hw, int nr_lane, int up);
+void rnp_link_report(struct rte_eth_dev *dev, bool link_en);
+int rnp_fw_msg_handler(struct rnp_eth_adapter *adapter);
#endif /* __RNP_MBX_FW_H__*/
--
2.27.0
^ permalink raw reply related [flat|nested] 27+ messages in thread* Re: [PATCH v6 0/8] [v6]drivers/net Add Support mucse N10 Pmd Driver
2023-09-01 2:30 [PATCH v6 0/8] [v6]drivers/net Add Support mucse N10 Pmd Driver Wenbo Cao
` (7 preceding siblings ...)
2023-09-01 2:30 ` [PATCH v6 8/8] net/rnp handle device interrupts Wenbo Cao
@ 2023-09-05 15:34 ` Ferruh Yigit
8 siblings, 0 replies; 27+ messages in thread
From: Ferruh Yigit @ 2023-09-05 15:34 UTC (permalink / raw)
To: Wenbo Cao; +Cc: dev, thomas, andrew.rybchenko, yaojun
On 9/1/2023 3:30 AM, Wenbo Cao wrote:
> For This patchset just to support the basic chip init work
> and user can just found the eth_dev, but can't control more.
>
Hi Wenbo,
After this patch, driver doesn't implement any of the dev_ops (struct
eth_dev_ops), so I assume it is not functional at all.
But meanwhile patchset has devargs parsing, fw_update, LSC interrupts,
implemented etc.
I think it makes testing/reviewing harder to add more code before basic
device functionality is missing.
What about implementing following main dev_ops and implement a simple
datapath, to have minimum working driver, later add features on top of it:
dev_configure
dev_start
dev_stop
dev_close
dev_infos_get
rx_queue_setup
tx_queue_setup
Also I have comment on putting 'mac_ops' on process_private, I suspect
there is a design problem there, please check details in relevant patch.
Thanks,
Ferruh
> For Now just support 2*10g nic,the chip can support
> 2*10g,4*10g,4*1g,8*1g,8*10g.
> The Feature rx side can support rx-cksum-offload,rss,vlan-filter
> flow_clow,uncast_filter,mcast_filter,1588,Jumbo-frame
> The Feature tx side can supprt tx-cksum-offload,tso,vxlan-tso
> flow director base on ntuple pattern of tcp/udp/ip/ eth_hdr->type
> for sriov is also support.
>
> Because of the chip desgin defect, for multiple-port mode
> one pci-bdf will have multiple-port (max can have four ports)
> so this code must be care of one bdf init multiple-port.
>
> v6:
> * fixed the doc(rst) format problem advise by Thomas Monjalon
>
> v5:
> * fixed the symbol name require by the style documentation
> rx_queue_setup
> v4:
> * one patch has been forgot to upload :(
>
> v3:
> * fixed http://dpdk.org/patch/129830 FreeBSD 13 compile Issue
> * change iobar type to void suggest by Stephen Hemminger
> * add KMOD_DEP support for vfio-pci
> * change run-cmd argument parse check for invalid extra_args
>
> v2:
> * fixed MAINTAIN maillist fullname format
> * fixed driver/net/meson the order issue of new driver to driver list
> * improve virtual point function usage suggest by Stephen Hemminger
>
>
> Wenbo Cao (8):
> net/rnp: add skeleton
> net/rnp: add ethdev probe and remove
> net/rnp: add device init and uninit
> net/rnp: add mbx basic api feature
> net/rnp add reset code for Chip Init process
> net/rnp add port info resource init
> net/rnp add devargs runtime parsing functions
> net/rnp handle device interrupts
>
>
^ permalink raw reply [flat|nested] 27+ messages in thread