[v1] net/axgbe: use PCI root complex device to distinguish AMD hardware

Message ID 20211126102403.13223-1-chandu@amd.com (mailing list archive)
State Superseded, archived
Delegated to: Ferruh Yigit
Headers
Series [v1] net/axgbe: use PCI root complex device to distinguish AMD hardware |

Checks

Context Check Description
ci/checkpatch success coding style OK
ci/Intel-compilation success Compilation OK
ci/intel-Testing success Testing PASS
ci/github-robot: build success github build: passed
ci/iol-mellanox-Performance success Performance Testing PASS
ci/iol-broadcom-Performance success Performance Testing PASS
ci/iol-broadcom-Functional success Functional Testing PASS
ci/iol-intel-Performance success Performance Testing PASS
ci/iol-intel-Functional success Functional Testing PASS
ci/iol-x86_64-unit-testing success Testing PASS
ci/iol-aarch64-unit-testing success Testing PASS
ci/iol-aarch64-compile-testing success Testing PASS
ci/iol-x86_64-compile-testing success Testing PASS

Commit Message

Namburu, Chandu-babu Nov. 26, 2021, 10:24 a.m. UTC
  "bus/pci: optimize bus scan" broke axgbe on V1000/R1000.
RV root complex pci device does not have any kernel driver assigned
so it is removed from pci scan list which is used in
"net/axgbe: add a HW quirk for register definitions"

Get root complex device id directly from pci sysfs instead of
pci scan list

Signed-off-by: Chandubabu Namburu <chandu@amd.com>
---
 drivers/net/axgbe/axgbe_ethdev.c | 39 ++++++++++++++++----------------
 1 file changed, 20 insertions(+), 19 deletions(-)
  

Comments

Awan, Arsalan Dec. 1, 2021, 8:45 a.m. UTC | #1
Hi Chandu,

I have tested this in the following configurations and it works very well in all cases:

* V1000 with Yocto v3.3.2 & Linux v5.10.47
* E3000 with Yocto v3.3.2 & Linux v5.10.47
* V1000 with Ubuntu Server v20.04.3 LTS
* E3000 with Ubuntu Server v20.04.3 LTS

I think this is a much better approach than scanning the whole PCI bus and comparing
each and every device against the PCI Root Complex Device ID of the platform.

This solution looks good to me!

Acked by: Arsalan H. Awan <Arsalan_Awan@mentor.com>
  
David Marchand Dec. 1, 2021, 9 a.m. UTC | #2
On Fri, Nov 26, 2021 at 11:24 AM Chandubabu Namburu <chandu@amd.com> wrote:
>
> "bus/pci: optimize bus scan" broke axgbe on V1000/R1000.
> RV root complex pci device does not have any kernel driver assigned
> so it is removed from pci scan list which is used in
> "net/axgbe: add a HW quirk for register definitions"
>
> Get root complex device id directly from pci sysfs instead of
> pci scan list

Please end the sentences with a '.'.

As you described, this change is a fix.
Please add a Fixes: tag and Cc: stable@dpdk.org in the commitlog if
you think it should be backported to LTS releases.
https://doc.dpdk.org/guides/contributing/patches.html#patch-for-stable-releases


>
> Signed-off-by: Chandubabu Namburu <chandu@amd.com>
> ---
>  drivers/net/axgbe/axgbe_ethdev.c | 39 ++++++++++++++++----------------
>  1 file changed, 20 insertions(+), 19 deletions(-)
>
> diff --git a/drivers/net/axgbe/axgbe_ethdev.c b/drivers/net/axgbe/axgbe_ethdev.c
> index 9cd056d04a..26babde354 100644
> --- a/drivers/net/axgbe/axgbe_ethdev.c
> +++ b/drivers/net/axgbe/axgbe_ethdev.c
> @@ -10,6 +10,8 @@
>  #include "axgbe_regs.h"
>  #include "rte_time.h"
>
> +#include "eal_filesystem.h"
> +
>  static int eth_axgbe_dev_init(struct rte_eth_dev *eth_dev);
>  static int  axgbe_dev_configure(struct rte_eth_dev *dev);
>  static int  axgbe_dev_start(struct rte_eth_dev *dev);
> @@ -1923,28 +1925,27 @@ static void axgbe_default_config(struct axgbe_port *pdata)
>         pdata->power_down = 0;
>  }
>
> -static int
> -pci_device_cmp(const struct rte_device *dev, const void *_pci_id)
> +/*
> + * Return PCI root complex device id on success else 0
> + */
> +static uint16_t
> +get_pci_rc_devid(void)
>  {
> -       const struct rte_pci_device *pdev = RTE_DEV_TO_PCI_CONST(dev);
> -       const struct rte_pci_id *pcid = _pci_id;
> +       char pci_sysfs[PATH_MAX];
> +       const struct rte_pci_addr pci_rc_addr = {0, 0, 0, 0};
> +       unsigned long device_id;
>
> -       if (pdev->id.vendor_id == AMD_PCI_VENDOR_ID &&
> -                       pdev->id.device_id == pcid->device_id)
> -               return 0;
> -       return 1;
> -}
> +       snprintf(pci_sysfs, sizeof(pci_sysfs), "%s/" PCI_PRI_FMT "/device",
> +                rte_pci_get_sysfs_path(), pci_rc_addr.domain,
> +                pci_rc_addr.bus, pci_rc_addr.devid, pci_rc_addr.function);

The use of rte_pci_get_sysfs_path() and in general is ugly because it
only works for Linux.
We could come up with a better API in the pci bus, but I don't think
it is worth it atm.

We can live with your implementation since this driver is only
compiled on Linux.
  
Namburu, Chandu-babu Dec. 2, 2021, 5:53 a.m. UTC | #3
[Public]

Hi David Marchand,

Thank you for your comments. Will change patch and submit v2 for review

Regards,
Chandu

-----Original Message-----
From: David Marchand <david.marchand@redhat.com> 
Sent: Wednesday, December 1, 2021 2:31 PM
To: Namburu, Chandu-babu <chandu@amd.com>
Cc: dev <dev@dpdk.org>; Sebastian, Selwin <Selwin.Sebastian@amd.com>; Arsalan_Awan@mentor.com; Yigit, Ferruh <ferruh.yigit@intel.com>
Subject: Re: [PATCH v1] net/axgbe: use PCI root complex device to distinguish AMD hardware

On Fri, Nov 26, 2021 at 11:24 AM Chandubabu Namburu <chandu@amd.com> wrote:
>
> "bus/pci: optimize bus scan" broke axgbe on V1000/R1000.
> RV root complex pci device does not have any kernel driver assigned so 
> it is removed from pci scan list which is used in
> "net/axgbe: add a HW quirk for register definitions"
>
> Get root complex device id directly from pci sysfs instead of pci scan 
> list

Please end the sentences with a '.'.

As you described, this change is a fix.
Please add a Fixes: tag and Cc: stable@dpdk.org in the commitlog if you think it should be backported to LTS releases.
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdoc.dpdk.org%2Fguides%2Fcontributing%2Fpatches.html%23patch-for-stable-releases&amp;data=04%7C01%7Cchandu%40amd.com%7C6618c303ecb146be0b5908d9b4a91c01%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637739460691977618%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&amp;sdata=piC1U%2BPcPP2A0xwM3%2FA5qDUFXNLAj2SrLBywTFc3DFM%3D&amp;reserved=0


>
> Signed-off-by: Chandubabu Namburu <chandu@amd.com>
> ---
>  drivers/net/axgbe/axgbe_ethdev.c | 39 
> ++++++++++++++++----------------
>  1 file changed, 20 insertions(+), 19 deletions(-)
>
> diff --git a/drivers/net/axgbe/axgbe_ethdev.c 
> b/drivers/net/axgbe/axgbe_ethdev.c
> index 9cd056d04a..26babde354 100644
> --- a/drivers/net/axgbe/axgbe_ethdev.c
> +++ b/drivers/net/axgbe/axgbe_ethdev.c
> @@ -10,6 +10,8 @@
>  #include "axgbe_regs.h"
>  #include "rte_time.h"
>
> +#include "eal_filesystem.h"
> +
>  static int eth_axgbe_dev_init(struct rte_eth_dev *eth_dev);  static 
> int  axgbe_dev_configure(struct rte_eth_dev *dev);  static int  
> axgbe_dev_start(struct rte_eth_dev *dev); @@ -1923,28 +1925,27 @@ 
> static void axgbe_default_config(struct axgbe_port *pdata)
>         pdata->power_down = 0;
>  }
>
> -static int
> -pci_device_cmp(const struct rte_device *dev, const void *_pci_id)
> +/*
> + * Return PCI root complex device id on success else 0  */ static 
> +uint16_t
> +get_pci_rc_devid(void)
>  {
> -       const struct rte_pci_device *pdev = RTE_DEV_TO_PCI_CONST(dev);
> -       const struct rte_pci_id *pcid = _pci_id;
> +       char pci_sysfs[PATH_MAX];
> +       const struct rte_pci_addr pci_rc_addr = {0, 0, 0, 0};
> +       unsigned long device_id;
>
> -       if (pdev->id.vendor_id == AMD_PCI_VENDOR_ID &&
> -                       pdev->id.device_id == pcid->device_id)
> -               return 0;
> -       return 1;
> -}
> +       snprintf(pci_sysfs, sizeof(pci_sysfs), "%s/" PCI_PRI_FMT "/device",
> +                rte_pci_get_sysfs_path(), pci_rc_addr.domain,
> +                pci_rc_addr.bus, pci_rc_addr.devid, 
> + pci_rc_addr.function);

The use of rte_pci_get_sysfs_path() and in general is ugly because it only works for Linux.
We could come up with a better API in the pci bus, but I don't think it is worth it atm.

We can live with your implementation since this driver is only compiled on Linux.


--
David Marchand
  

Patch

diff --git a/drivers/net/axgbe/axgbe_ethdev.c b/drivers/net/axgbe/axgbe_ethdev.c
index 9cd056d04a..26babde354 100644
--- a/drivers/net/axgbe/axgbe_ethdev.c
+++ b/drivers/net/axgbe/axgbe_ethdev.c
@@ -10,6 +10,8 @@ 
 #include "axgbe_regs.h"
 #include "rte_time.h"
 
+#include "eal_filesystem.h"
+
 static int eth_axgbe_dev_init(struct rte_eth_dev *eth_dev);
 static int  axgbe_dev_configure(struct rte_eth_dev *dev);
 static int  axgbe_dev_start(struct rte_eth_dev *dev);
@@ -1923,28 +1925,27 @@  static void axgbe_default_config(struct axgbe_port *pdata)
 	pdata->power_down = 0;
 }
 
-static int
-pci_device_cmp(const struct rte_device *dev, const void *_pci_id)
+/*
+ * Return PCI root complex device id on success else 0
+ */
+static uint16_t
+get_pci_rc_devid(void)
 {
-	const struct rte_pci_device *pdev = RTE_DEV_TO_PCI_CONST(dev);
-	const struct rte_pci_id *pcid = _pci_id;
+	char pci_sysfs[PATH_MAX];
+	const struct rte_pci_addr pci_rc_addr = {0, 0, 0, 0};
+	unsigned long device_id;
 
-	if (pdev->id.vendor_id == AMD_PCI_VENDOR_ID &&
-			pdev->id.device_id == pcid->device_id)
-		return 0;
-	return 1;
-}
+	snprintf(pci_sysfs, sizeof(pci_sysfs), "%s/" PCI_PRI_FMT "/device",
+		 rte_pci_get_sysfs_path(), pci_rc_addr.domain,
+		 pci_rc_addr.bus, pci_rc_addr.devid, pci_rc_addr.function);
 
-static bool
-pci_search_device(int device_id)
-{
-	struct rte_bus *pci_bus;
-	struct rte_pci_id dev_id;
+	/* get device id */
+	if (eal_parse_sysfs_value(pci_sysfs, &device_id) < 0) {
+		PMD_INIT_LOG(ERR, "Error in reading PCI sysfs\n");
+		return 0;
+	}
 
-	dev_id.device_id = device_id;
-	pci_bus = rte_bus_find_by_name("pci");
-	return (pci_bus != NULL) &&
-		(pci_bus->find_device(NULL, pci_device_cmp, &dev_id) != NULL);
+	return (uint16_t)device_id;
 }
 
 /*
@@ -1986,7 +1987,7 @@  eth_axgbe_dev_init(struct rte_eth_dev *eth_dev)
 	/*
 	 * Use root complex device ID to differentiate RV AXGBE vs SNOWY AXGBE
 	 */
-	if (pci_search_device(AMD_PCI_RV_ROOT_COMPLEX_ID)) {
+	if ((get_pci_rc_devid()) == AMD_PCI_RV_ROOT_COMPLEX_ID) {
 		pdata->xpcs_window_def_reg = PCS_V2_RV_WINDOW_DEF;
 		pdata->xpcs_window_sel_reg = PCS_V2_RV_WINDOW_SELECT;
 	} else {