Skip to contentSkip to conversation

bus/pci: optimize pci device probe

PATCH NAK [email protected] 5 participants 15 messages Apr 26, 2020 – May 6, 2020

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 bus/pci: optimize pci device probe

From: Jerin Jacob <[email protected]>

If the PCI device is not attached to any driver then there is no point in probing it. As an optimization, skip the PCI device probe if the PCI device driver of type RTE_KDRV_NONE.

---
Notes:
------
- virtio drivers does special treatment based on RTE_KDRV_UNKNOWN, That is the reason allowing RTE_KDRV_UNKNOWN in this patch.
- virio devices uses RTE_KDRV_UNKNOWN for some special meaning, IMO, if it would be better, if
a) Introduce the KDRV for virio
b) If the PCIe device of driver type NONE or UNKNOWN then not even add in pci list
in the scan, It will improve the boot time by avoiding operation on unwanted device like sorting the PCI devices, scanning it, probe it, managing it etc.

- Initial problem reported at http://patches.dpdk.org/patch/64999/ as boot time printf clutter on octeontx2 devices with a lot PCI devices which are of type RTE_KDRV_NONE.

 drivers/bus/pci/pci_common.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/bus/pci/pci_common.c +2 -0
index 3f5542076..2fa3d85ae 100644
@@ -271,6 +271,8 @@ pci_probe_all_drivers(struct rte_pci_device *dev)
 		return -EINVAL;
 
 	FOREACH_DRIVER_ON_PCIBUS(dr) {
+		if (dev->kdrv == RTE_KDRV_NONE)
+			continue;
 		rc = rte_pci_probe_one_driver(dr, dev);
 		if (rc < 0)
 			/* negative value is an error */

--
2.26.2

#1 lore raw

26/04/2020 19:38, [email protected]:

Quoted 7 lines

From: Jerin Jacob <[email protected]>

If the PCI device is not attached to any driver then there is no point in probing it. As an optimization, skip the PCI device probe if the PCI device driver of type RTE_KDRV_NONE.

Quoted 13 lines

---
Notes:
------
- virtio drivers does special treatment based on RTE_KDRV_UNKNOWN, That is the reason allowing RTE_KDRV_UNKNOWN in this patch.
- virio devices uses RTE_KDRV_UNKNOWN for some special meaning, IMO, if it would be better, if
a) Introduce the KDRV for virio
b) If the PCIe device of driver type NONE or UNKNOWN then not even add in pci list
in the scan, It will improve the boot time by avoiding operation on unwanted device like sorting the PCI devices, scanning it, probe it, managing it etc.

mlx4/mlx4 uses RTE_KDRV_UNKNOWN.

Quoted 3 lines

- Initial problem reported at http://patches.dpdk.org/patch/64999/ as boot time printf clutter on octeontx2 devices with a lot PCI devices which are of type RTE_KDRV_NONE.

Add a logtype for PCI driver and adjust log level accordingly to your preferences.

drivers/bus/pci/pci_common.c pci_probe_all_drivers(struct rte_pci_device *dev) · line 274

Quoted 5 lines
drivers/bus/pci/pci_common.c +2 -0
@@ -271,6 +271,8 @@ pci_probe_all_drivers(struct rte_pci_device *dev)
 	FOREACH_DRIVER_ON_PCIBUS(dr) {
+		if (dev->kdrv == RTE_KDRV_NONE)
+			continue;
 		rc = rte_pci_probe_one_driver(dr, dev);

Nack

#2 lore raw
Quoted 23 lines · Thomas Monjalon

26/04/2020 19:38, [email protected]:

Quoted 7 lines

From: Jerin Jacob <[email protected]>

If the PCI device is not attached to any driver then there is no point in probing it. As an optimization, skip the PCI device probe if the PCI device driver of type RTE_KDRV_NONE.

Quoted 13 lines

---
Notes:
------
- virtio drivers does special treatment based on RTE_KDRV_UNKNOWN, That is the reason allowing RTE_KDRV_UNKNOWN in this patch.
- virio devices uses RTE_KDRV_UNKNOWN for some special meaning, IMO, if it would be better, if
a) Introduce the KDRV for virio
b) If the PCIe device of driver type NONE or UNKNOWN then not even add in pci list
in the scan, It will improve the boot time by avoiding operation on unwanted device like sorting the PCI devices, scanning it, probe it, managing it etc.

Quoted 1 line

mlx4/mlx4 uses RTE_KDRV_UNKNOWN.

OK.

Quoted 8 lines
Quoted 4 lines

- Initial problem reported at http://patches.dpdk.org/patch/64999/ as boot time printf clutter on octeontx2 devices with a lot PCI devices which are of type RTE_KDRV_NONE.

Add a logtype for PCI driver and adjust log level accordingly to your preferences.

Quoted 7 lines

drivers/bus/pci/pci_common.c pci_probe_all_drivers(struct rte_pci_device *dev) · line 274

Quoted 5 lines
drivers/bus/pci/pci_common.c +2 -0
@@ -271,6 +271,8 @@ pci_probe_all_drivers(struct rte_pci_device *dev)
      FOREACH_DRIVER_ON_PCIBUS(dr) {
+             if (dev->kdrv == RTE_KDRV_NONE)
+                     continue;
              rc = rte_pci_probe_one_driver(dr, dev);

Nack

I understand mlx4/mlx5 is using RTE_KDRV_UNKNOWN, Here we are skipping the RTE_KDRV_NONE,
What is the use case for probing the devices with RTE_KDRV_NONE?

Quoted 2 lines
#3 lore raw

26/04/2020 20:41, Jerin Jacob:

Quoted 43 lines · Thomas Monjalon
Quoted 23 lines

26/04/2020 19:38, [email protected]:

Quoted 7 lines

From: Jerin Jacob <[email protected]>

If the PCI device is not attached to any driver then there is no point in probing it. As an optimization, skip the PCI device probe if the PCI device driver of type RTE_KDRV_NONE.

Quoted 13 lines

---
Notes:
------
- virtio drivers does special treatment based on RTE_KDRV_UNKNOWN, That is the reason allowing RTE_KDRV_UNKNOWN in this patch.
- virio devices uses RTE_KDRV_UNKNOWN for some special meaning, IMO, if it would be better, if
a) Introduce the KDRV for virio
b) If the PCIe device of driver type NONE or UNKNOWN then not even add in pci list
in the scan, It will improve the boot time by avoiding operation on unwanted device like sorting the PCI devices, scanning it, probe it, managing it etc.

Quoted 1 line

mlx4/mlx4 uses RTE_KDRV_UNKNOWN.

OK.

Quoted 7 lines
Quoted 3 lines

- Initial problem reported at http://patches.dpdk.org/patch/64999/ as boot time printf clutter on octeontx2 devices with a lot PCI devices which are of type RTE_KDRV_NONE.

Add a logtype for PCI driver and adjust log level accordingly to your preferences.

Quoted 7 lines

drivers/bus/pci/pci_common.c pci_probe_all_drivers(struct rte_pci_device *dev) · line 274

Quoted 5 lines
drivers/bus/pci/pci_common.c +2 -0
@@ -271,6 +271,8 @@ pci_probe_all_drivers(struct rte_pci_device *dev)
      FOREACH_DRIVER_ON_PCIBUS(dr) {
+             if (dev->kdrv == RTE_KDRV_NONE)
+                     continue;
              rc = rte_pci_probe_one_driver(dr, dev);

Nack

Quoted 3 lines

I understand mlx4/mlx5 is using RTE_KDRV_UNKNOWN, Here we are skipping the RTE_KDRV_NONE,
What is the use case for probing the devices with RTE_KDRV_NONE?

Maybe you are right. I don't remember the use case.
I think I remember these virtio and vmxnet3 PMD were not using UIO:
http://git.dpdk.org/old/virtio-net-pmd/tree/virtio_user.c
http://git.dpdk.org/old/vmxnet3-usermap/tree/pmd/vmxnet3.c

We need to know which case is using following code:

case RTE_KDRV_NONE:

#if defined(RTE_ARCH_X86)
        ret = pci_ioport_map(dev, bar, p);
#endif
        break;

David, please could you refresh our memory?

#4 lore raw
Quoted 45 lines · Thomas Monjalon

26/04/2020 20:41, Jerin Jacob:

Quoted 43 lines · Thomas Monjalon
Quoted 23 lines

26/04/2020 19:38, [email protected]:

Quoted 7 lines

From: Jerin Jacob <[email protected]>

If the PCI device is not attached to any driver then there is no point in probing it. As an optimization, skip the PCI device probe if the PCI device driver of type RTE_KDRV_NONE.

Quoted 13 lines

---
Notes:
------
- virtio drivers does special treatment based on RTE_KDRV_UNKNOWN, That is the reason allowing RTE_KDRV_UNKNOWN in this patch.
- virio devices uses RTE_KDRV_UNKNOWN for some special meaning, IMO, if it would be better, if
a) Introduce the KDRV for virio
b) If the PCIe device of driver type NONE or UNKNOWN then not even add in pci list
in the scan, It will improve the boot time by avoiding operation on unwanted device like sorting the PCI devices, scanning it, probe it, managing it etc.

Quoted 1 line

mlx4/mlx4 uses RTE_KDRV_UNKNOWN.

OK.

Quoted 14 lines
Quoted 3 lines

- Initial problem reported at http://patches.dpdk.org/patch/64999/ as boot time printf clutter on octeontx2 devices with a lot PCI devices which are of type RTE_KDRV_NONE.

Add a logtype for PCI driver and adjust log level accordingly to your preferences.

drivers/bus/pci/pci_common.c pci_probe_all_drivers(struct rte_pci_device *dev) · line 274

Quoted 5 lines
drivers/bus/pci/pci_common.c +2 -0
@@ -271,6 +271,8 @@ pci_probe_all_drivers(struct rte_pci_device *dev)
      FOREACH_DRIVER_ON_PCIBUS(dr) {
+             if (dev->kdrv == RTE_KDRV_NONE)
+                     continue;
              rc = rte_pci_probe_one_driver(dr, dev);

Nack

Quoted 8 lines
Quoted 3 lines

I understand mlx4/mlx5 is using RTE_KDRV_UNKNOWN, Here we are skipping the RTE_KDRV_NONE,
What is the use case for probing the devices with RTE_KDRV_NONE?

Maybe you are right. I don't remember the use case.
I think I remember these virtio and vmxnet3 PMD were not using UIO:
http://git.dpdk.org/old/virtio-net-pmd/tree/virtio_user.c
http://git.dpdk.org/old/vmxnet3-usermap/tree/pmd/vmxnet3.c

Looks it is OLD stale drivers with DPDK 1.3 or something.

Currently, virtio is using, following drivers.

RTE_PMD_REGISTER_PCI_TABLE(net_virtio, pci_id_virtio_map); RTE_PMD_REGISTER_KMOD_DEP(net_virtio, "* igb_uio | uio_pci_generic | vfio-pci");

Quoted 8 lines

We need to know which case is using following code:

case RTE_KDRV_NONE:

#if defined(RTE_ARCH_X86)
        ret = pci_ioport_map(dev, bar, p);
#endif
        break;

AFAIK, it does not affect the scanning. virtio maintainers?

Quoted 3 lines

David, please could you refresh our memory?

#5 lore raw
In reply to Thomas Monjalon
Quoted 56 lines · Thomas Monjalon

26/04/2020 20:41, Jerin Jacob:

Quoted 43 lines · Thomas Monjalon
Quoted 23 lines

26/04/2020 19:38, [email protected]:

Quoted 7 lines

From: Jerin Jacob <[email protected]>

If the PCI device is not attached to any driver then there is no point in probing it. As an optimization, skip the PCI device probe if the PCI device driver of type RTE_KDRV_NONE.

Quoted 13 lines

---
Notes:
------
- virtio drivers does special treatment based on RTE_KDRV_UNKNOWN, That is the reason allowing RTE_KDRV_UNKNOWN in this patch.
- virio devices uses RTE_KDRV_UNKNOWN for some special meaning, IMO, if it would be better, if
a) Introduce the KDRV for virio
b) If the PCIe device of driver type NONE or UNKNOWN then not even add in pci list
in the scan, It will improve the boot time by avoiding operation on unwanted device like sorting the PCI devices, scanning it, probe it, managing it etc.

Quoted 1 line

mlx4/mlx4 uses RTE_KDRV_UNKNOWN.

OK.

Quoted 7 lines
Quoted 3 lines

- Initial problem reported at http://patches.dpdk.org/patch/64999/ as boot time printf clutter on octeontx2 devices with a lot PCI devices which are of type RTE_KDRV_NONE.

Add a logtype for PCI driver and adjust log level accordingly to your preferences.

Quoted 7 lines

drivers/bus/pci/pci_common.c pci_probe_all_drivers(struct rte_pci_device *dev) · line 274

Quoted 5 lines
drivers/bus/pci/pci_common.c +2 -0
@@ -271,6 +271,8 @@ pci_probe_all_drivers(struct rte_pci_device *dev)
      FOREACH_DRIVER_ON_PCIBUS(dr) {
+             if (dev->kdrv == RTE_KDRV_NONE)
+                     continue;
              rc = rte_pci_probe_one_driver(dr, dev);

Nack

Quoted 3 lines

I understand mlx4/mlx5 is using RTE_KDRV_UNKNOWN, Here we are skipping the RTE_KDRV_NONE,
What is the use case for probing the devices with RTE_KDRV_NONE?

Maybe you are right. I don't remember the use case.
I think I remember these virtio and vmxnet3 PMD were not using UIO:
http://git.dpdk.org/old/virtio-net-pmd/tree/virtio_user.c
http://git.dpdk.org/old/vmxnet3-usermap/tree/pmd/vmxnet3.c

We need to know which case is using following code:

Quoted 7 lines

case RTE_KDRV_NONE:

#if defined(RTE_ARCH_X86)
        ret = pci_ioport_map(dev, bar, p);
#endif
        break;

David, please could you refresh our memory?

The in-tree virtio-net driver directly calls rte_pci_map_device / rte_pci_ioport_map depending on virtio legacy/modern modes. This is why the virtio pci driver does not ask for RTE_PCI_DRV_NEED_MAPPING to the pci bus.

In ioport mode, there were two options for historical reasons because of the virtio driver you mention.
This driver did not rely on uio, and this behavior was later merged to the current in-tree driver.
It ended up (not clean) in the pci bus driver because virtio was the only user of this code.

Removing this special case could break x86 applications running with legacy virtio.

On the plus side, we have been announcing for some time in virtio:
RTE_PMD_REGISTER_KMOD_DEP(net_virtio, "* igb_uio | uio_pci_generic | vfio-pci");

Signature · 1 line
David Marchand
#6 lore raw
Quoted 80 lines · David Marchand
Quoted 56 lines · Thomas Monjalon

26/04/2020 20:41, Jerin Jacob:

Quoted 43 lines · Thomas Monjalon
Quoted 23 lines

26/04/2020 19:38, [email protected]:

Quoted 7 lines

From: Jerin Jacob <[email protected]>

If the PCI device is not attached to any driver then there is no point in probing it. As an optimization, skip the PCI device probe if the PCI device driver of type RTE_KDRV_NONE.

Quoted 13 lines

---
Notes:
------
- virtio drivers does special treatment based on RTE_KDRV_UNKNOWN, That is the reason allowing RTE_KDRV_UNKNOWN in this patch.
- virio devices uses RTE_KDRV_UNKNOWN for some special meaning, IMO, if it would be better, if
a) Introduce the KDRV for virio
b) If the PCIe device of driver type NONE or UNKNOWN then not even add in pci list
in the scan, It will improve the boot time by avoiding operation on unwanted device like sorting the PCI devices, scanning it, probe it, managing it etc.

Quoted 1 line

mlx4/mlx4 uses RTE_KDRV_UNKNOWN.

OK.

Quoted 7 lines
Quoted 3 lines

- Initial problem reported at http://patches.dpdk.org/patch/64999/ as boot time printf clutter on octeontx2 devices with a lot PCI devices which are of type RTE_KDRV_NONE.

Add a logtype for PCI driver and adjust log level accordingly to your preferences.

Quoted 7 lines

drivers/bus/pci/pci_common.c pci_probe_all_drivers(struct rte_pci_device *dev) · line 274

Quoted 5 lines
drivers/bus/pci/pci_common.c +2 -0
@@ -271,6 +271,8 @@ pci_probe_all_drivers(struct rte_pci_device *dev)
      FOREACH_DRIVER_ON_PCIBUS(dr) {
+             if (dev->kdrv == RTE_KDRV_NONE)
+                     continue;
              rc = rte_pci_probe_one_driver(dr, dev);

Nack

Quoted 3 lines

I understand mlx4/mlx5 is using RTE_KDRV_UNKNOWN, Here we are skipping the RTE_KDRV_NONE,
What is the use case for probing the devices with RTE_KDRV_NONE?

Maybe you are right. I don't remember the use case.
I think I remember these virtio and vmxnet3 PMD were not using UIO:
http://git.dpdk.org/old/virtio-net-pmd/tree/virtio_user.c
http://git.dpdk.org/old/vmxnet3-usermap/tree/pmd/vmxnet3.c

We need to know which case is using following code:

Quoted 7 lines

case RTE_KDRV_NONE:

#if defined(RTE_ARCH_X86)
        ret = pci_ioport_map(dev, bar, p);
#endif
        break;

David, please could you refresh our memory?

The in-tree virtio-net driver directly calls rte_pci_map_device / rte_pci_ioport_map depending on virtio legacy/modern modes. This is why the virtio pci driver does not ask for RTE_PCI_DRV_NEED_MAPPING to the pci bus.

In ioport mode, there were two options for historical reasons because of the virtio driver you mention.
This driver did not rely on uio, and this behavior was later merged to the current in-tree driver.
It ended up (not clean) in the pci bus driver because virtio was the only user of this code.

Quoted 6 lines

Removing this special case could break x86 applications running with legacy virtio.

On the plus side, we have been announcing for some time in virtio:
RTE_PMD_REGISTER_KMOD_DEP(net_virtio, "* igb_uio | uio_pci_generic | vfio-pci");

What is to conclude?
# The In-tree virtio driver uses ""* igb_uio | uio_pci_generic | vfio-pci"" driver as backend and it does not need RTE_KDRV_NONE? OR
# The in-tree, legacy virtio(const struct virtio_pci_ops legacy_op) can work without any kernel driver in the backend. So RTE_KDRV_NONE need?

Quoted 5 lines

--
David Marchand

#7 lore raw
Quoted 90 lines · Jerin Jacob
Quoted 80 lines · David Marchand
Quoted 56 lines · Thomas Monjalon

26/04/2020 20:41, Jerin Jacob:

Quoted 43 lines · Thomas Monjalon
Quoted 23 lines

26/04/2020 19:38, [email protected]:

Quoted 7 lines

From: Jerin Jacob <[email protected]>

If the PCI device is not attached to any driver then there is no point in probing it. As an optimization, skip the PCI device probe if the PCI device driver of type RTE_KDRV_NONE.

Quoted 13 lines

---
Notes:
------
- virtio drivers does special treatment based on RTE_KDRV_UNKNOWN, That is the reason allowing RTE_KDRV_UNKNOWN in this patch.
- virio devices uses RTE_KDRV_UNKNOWN for some special meaning, IMO, if it would be better, if
a) Introduce the KDRV for virio
b) If the PCIe device of driver type NONE or UNKNOWN then not even add in pci list
in the scan, It will improve the boot time by avoiding operation on unwanted device like sorting the PCI devices, scanning it, probe it, managing it etc.

Quoted 1 line

mlx4/mlx4 uses RTE_KDRV_UNKNOWN.

OK.

Quoted 7 lines
Quoted 3 lines

- Initial problem reported at http://patches.dpdk.org/patch/64999/ as boot time printf clutter on octeontx2 devices with a lot PCI devices which are of type RTE_KDRV_NONE.

Add a logtype for PCI driver and adjust log level accordingly to your preferences.

Quoted 7 lines

drivers/bus/pci/pci_common.c pci_probe_all_drivers(struct rte_pci_device *dev) · line 274

Quoted 5 lines
drivers/bus/pci/pci_common.c +2 -0
@@ -271,6 +271,8 @@ pci_probe_all_drivers(struct rte_pci_device *dev)
      FOREACH_DRIVER_ON_PCIBUS(dr) {
+             if (dev->kdrv == RTE_KDRV_NONE)
+                     continue;
              rc = rte_pci_probe_one_driver(dr, dev);

Nack

Quoted 3 lines

I understand mlx4/mlx5 is using RTE_KDRV_UNKNOWN, Here we are skipping the RTE_KDRV_NONE,
What is the use case for probing the devices with RTE_KDRV_NONE?

Maybe you are right. I don't remember the use case.
I think I remember these virtio and vmxnet3 PMD were not using UIO:
http://git.dpdk.org/old/virtio-net-pmd/tree/virtio_user.c
http://git.dpdk.org/old/vmxnet3-usermap/tree/pmd/vmxnet3.c

We need to know which case is using following code:

Quoted 7 lines

case RTE_KDRV_NONE:

#if defined(RTE_ARCH_X86)
        ret = pci_ioport_map(dev, bar, p);
#endif
        break;

David, please could you refresh our memory?

The in-tree virtio-net driver directly calls rte_pci_map_device / rte_pci_ioport_map depending on virtio legacy/modern modes. This is why the virtio pci driver does not ask for RTE_PCI_DRV_NEED_MAPPING to the pci bus.

In ioport mode, there were two options for historical reasons because of the virtio driver you mention.
This driver did not rely on uio, and this behavior was later merged to the current in-tree driver.
It ended up (not clean) in the pci bus driver because virtio was the only user of this code.

Quoted 6 lines

Removing this special case could break x86 applications running with legacy virtio.

On the plus side, we have been announcing for some time in virtio:
RTE_PMD_REGISTER_KMOD_DEP(net_virtio, "* igb_uio | uio_pci_generic | vfio-pci");

Quoted 7 lines

What is to conclude?
# The In-tree virtio driver uses ""* igb_uio | uio_pci_generic | vfio-pci"" driver as backend and it does not need RTE_KDRV_NONE? OR
# The in-tree, legacy virtio(const struct virtio_pci_ops legacy_op) can work without any kernel driver in the backend. So RTE_KDRV_NONE need?

Ping. What is the conclusion? If it is former then this patch is valid.

Quoted 9 lines
Quoted 9 lines

--
David Marchand

#8 lore raw
Quoted 16 lines · Jerin Jacob
Quoted 8 lines
Quoted 7 lines

Removing this special case could break x86 applications running with legacy virtio.

On the plus side, we have been announcing for some time in virtio:
RTE_PMD_REGISTER_KMOD_DEP(net_virtio, "* igb_uio | uio_pci_generic | vfio-pci");

Quoted 7 lines

What is to conclude?
# The In-tree virtio driver uses ""* igb_uio | uio_pci_generic | vfio-pci"" driver as backend and it does not need RTE_KDRV_NONE? OR
# The in-tree, legacy virtio(const struct virtio_pci_ops legacy_op) can work without any kernel driver in the backend. So RTE_KDRV_NONE need?

Quoted 1 line

Ping. What is the conclusion? If it is former then this patch is valid.

I am fine with dropping the legacy part, but I wanted to hear from Maxime at least.

Signature · 1 line
David Marchand
#9 lore raw

Hi,

Quoted 17 lines · David Marchand
Quoted 16 lines · Jerin Jacob
Quoted 15 lines
Quoted 7 lines

Removing this special case could break x86 applications running with legacy virtio.

On the plus side, we have been announcing for some time in virtio:
RTE_PMD_REGISTER_KMOD_DEP(net_virtio, "* igb_uio | uio_pci_generic | vfio-pci");

What is to conclude?
# The In-tree virtio driver uses ""* igb_uio | uio_pci_generic | vfio-pci"" driver as backend and it does not need RTE_KDRV_NONE? OR
# The in-tree, legacy virtio(const struct virtio_pci_ops legacy_op) can work without any kernel driver in the backend. So RTE_KDRV_NONE need?

Quoted 6 lines
Quoted 1 line

Ping. What is the conclusion? If it is former then this patch is valid.

I am fine with dropping the legacy part, but I wanted to hear from Maxime at least.

IIUC, it means that with Jerin patch, Virtio Legacy devices support will be dropped as they won't be probed anymore?

Maxime

#10 lore raw
Quoted 23 lines · Maxime Coquelin

Hi,

Quoted 19 lines · David Marchand
Quoted 17 lines · Jerin Jacob
Quoted 8 lines
Quoted 7 lines

Removing this special case could break x86 applications running with legacy virtio.

On the plus side, we have been announcing for some time in virtio:
RTE_PMD_REGISTER_KMOD_DEP(net_virtio, "* igb_uio | uio_pci_generic | vfio-pci");

Quoted 7 lines

What is to conclude?
# The In-tree virtio driver uses ""* igb_uio | uio_pci_generic | vfio-pci"" driver as backend and it does not need RTE_KDRV_NONE? OR
# The in-tree, legacy virtio(const struct virtio_pci_ops legacy_op) can work without any kernel driver in the backend. So RTE_KDRV_NONE need?

Ping. What is the conclusion? If it is former then this patch is valid.

Quoted 7 lines
Quoted 4 lines

I am fine with dropping the legacy part, but I wanted to hear from Maxime at least.

IIUC, it means that with Jerin patch, Virtio Legacy devices support will be dropped as they won't be probed anymore?

The device drivers with RTE_KDRV_NONE as the backend will not be probed.
1) Are Virtio Legacy devices are type of RTE_KDRV_NONE?
2) if yes, Would you like to support for virtio legacy device?
3) if yes, Please fix RTE_PMD_REGISTER_KMOD_DEP(net_virtio, "* igb_uio | uio_pci_generic | vfio-pci");

Quoted 3 lines

Maxime

#11 lore raw
Quoted 30 lines · Jerin Jacob
Quoted 28 lines · Maxime Coquelin

Hi,

Quoted 23 lines · David Marchand
Quoted 16 lines · Jerin Jacob
Quoted 8 lines
Quoted 7 lines

Removing this special case could break x86 applications running with legacy virtio.

On the plus side, we have been announcing for some time in virtio:
RTE_PMD_REGISTER_KMOD_DEP(net_virtio, "* igb_uio | uio_pci_generic | vfio-pci");

Quoted 7 lines

What is to conclude?
# The In-tree virtio driver uses ""* igb_uio | uio_pci_generic | vfio-pci"" driver as backend and it does not need RTE_KDRV_NONE? OR
# The in-tree, legacy virtio(const struct virtio_pci_ops legacy_op) can work without any kernel driver in the backend. So RTE_KDRV_NONE need?

Quoted 1 line

Ping. What is the conclusion? If it is former then this patch is valid.

I am fine with dropping the legacy part, but I wanted to hear from Maxime at least.

Quoted 5 lines
Quoted 2 lines

IIUC, it means that with Jerin patch, Virtio Legacy devices support will be dropped as they won't be probed anymore?

The device drivers with RTE_KDRV_NONE as the backend will not be probed.
1) Are Virtio Legacy devices are type of RTE_KDRV_NONE?

Virtio Legacy devices can be probed with no kernel driver.

Quoted 1 line

2) if yes, Would you like to support for virtio legacy device?

I am OK to remove legacy + RTE_KDRV_NONE case, but I think it needs an announcement and being done in a later release to let end-users using that configuration time to do the change.

Quoted 2 lines

3) if yes, Please fix RTE_PMD_REGISTER_KMOD_DEP(net_virtio, "* igb_uio | uio_pci_generic | vfio-pci");

While support gets removed, what about:

RTE_PMD_REGISTER_KMOD_DEP(net_virtio, "* igb_uio | uio_pci_generic | vfio-pci | none"); ?

Maxime

#12 lore raw
Quoted 40 lines · Maxime Coquelin
Quoted 30 lines · Jerin Jacob
Quoted 28 lines · Maxime Coquelin

Hi,

Quoted 23 lines · David Marchand
Quoted 16 lines · Jerin Jacob
Quoted 8 lines
Quoted 7 lines

Removing this special case could break x86 applications running with legacy virtio.

On the plus side, we have been announcing for some time in virtio:
RTE_PMD_REGISTER_KMOD_DEP(net_virtio, "* igb_uio | uio_pci_generic | vfio-pci");

Quoted 7 lines

What is to conclude?
# The In-tree virtio driver uses ""* igb_uio | uio_pci_generic | vfio-pci"" driver as backend and it does not need RTE_KDRV_NONE? OR
# The in-tree, legacy virtio(const struct virtio_pci_ops legacy_op) can work without any kernel driver in the backend. So RTE_KDRV_NONE need?

Quoted 1 line

Ping. What is the conclusion? If it is former then this patch is valid.

I am fine with dropping the legacy part, but I wanted to hear from Maxime at least.

Quoted 5 lines
Quoted 2 lines

IIUC, it means that with Jerin patch, Virtio Legacy devices support will be dropped as they won't be probed anymore?

The device drivers with RTE_KDRV_NONE as the backend will not be probed.
1) Are Virtio Legacy devices are type of RTE_KDRV_NONE?

Quoted 7 lines

Virtio Legacy devices can be probed with no kernel driver.

Quoted 1 line

2) if yes, Would you like to support for virtio legacy device?

I am OK to remove legacy + RTE_KDRV_NONE case, but I think it needs an announcement and being done in a later release to let end-users using that configuration time to do the change.

OK

Quoted 8 lines
Quoted 3 lines

3) if yes, Please fix RTE_PMD_REGISTER_KMOD_DEP(net_virtio, "* igb_uio | uio_pci_generic | vfio-pci");

While support gets removed, what about:

RTE_PMD_REGISTER_KMOD_DEP(net_virtio, "* igb_uio | uio_pci_generic | vfio-pci | none"); ?

+1

Quoted 3 lines

Maxime

#13 lore raw
Quoted 9 lines · Jerin Jacob
Quoted 7 lines
Quoted 2 lines

3) if yes, Please fix RTE_PMD_REGISTER_KMOD_DEP(net_virtio, "* igb_uio | uio_pci_generic | vfio-pci");

While support gets removed, what about:

RTE_PMD_REGISTER_KMOD_DEP(net_virtio, "* igb_uio | uio_pci_generic | vfio-pci | none"); ?

+1

I know some scripts rely on such info to modprobe and bind the devices based on such info.
So "none" does not make sense, there is no "none" kmod.

Since we want to deprecate this capability, let's not advertise this.

Signature · 1 line
David Marchand
#14 lore raw
Quoted 9 lines · David Marchand
Quoted 8 lines · Jerin Jacob
Quoted 7 lines
Quoted 2 lines

3) if yes, Please fix RTE_PMD_REGISTER_KMOD_DEP(net_virtio, "* igb_uio | uio_pci_generic | vfio-pci");

While support gets removed, what about:

RTE_PMD_REGISTER_KMOD_DEP(net_virtio, "* igb_uio | uio_pci_generic | vfio-pci | none"); ?

Quoted 7 lines
Quoted 1 line

+1

I know some scripts rely on such info to modprobe and bind the devices based on such info.
So "none" does not make sense, there is no "none" kmod.

Since we want to deprecate this capability, let's not advertise this.

Makes sense.

Jerin, will add this announcement in the release note? I think we can remove it in v20.08.

Thanks,
Maxime