net: macb: Wake-on-Lan magic packet fixes
Subsystems (3)
cover net: macb: Wake-on-Lan magic packet fixes
From: Nicolas Ferre <[email protected]>
Hi,
Here are some of my early patches in order to fix WoL magic-packet on the current macb driver.
Addition of this feature to GEM types of IPs is yet to come. I would like to have your feedback on these little patches first so that I can continue investigating the addition of GEM WoL magic-packet.
Harini, I know that you have patches for GEM in order to integrate WoL ARP mode [1]. I'll try to integrate some of your work but would need that this feature is better integrated in current code. For instance, the choice of "magic packet" or "ARP" should be done by ethtool options and DT properties. For matching with mainline users, MACB and GEM code must co-exist.
The use of dumb buffers for RX seems also fairly platform specific and we would need to think more about it.
[1]:
https://github.com/Xilinx/linux-xlnx/commit/e9648006e8d9132db2594e50e700af362b3c9226#diff-41909d180431659ccc1229aa30fd4e5a
https://github.com/Xilinx/linux-xlnx/commit/60a21c686f7e4e50489ae04b9bb1980b145e52ef
Nicolas Ferre (3):
net: macb: fix wakeup test in runtime suspend/resume routines
net: macb: mark device wake capable when "magic-packet" property
present
net: macb: fix macb_get/set_wol() when moving to phylink
drivers/net/ethernet/cadence/macb_main.c | 24 +++++++++++++----------- 1 file changed, 13 insertions(+), 11 deletions(-)
Signature · 1 line
2.20.1
RFC PATCH 1/3 net: macb: fix wakeup test in runtime suspend/resume routines
From: Nicolas Ferre <[email protected]>
Use the proper struct device pointer to check if the wakeup flag and wakeup source are positioned.
Use the one passed by function call which is equivalent to &bp->dev->dev.parent.
It's preventing the trigger of a spurious interrupt in case the Wake-on-Lan feature is used.
--- drivers/net/ethernet/cadence/macb_main.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/net/ethernet/cadence/macb_main.c +2 -2
@@ -4616,7 +4616,7 @@ static int __maybe_unused macb_runtime_suspend(struct device *dev)struct net_device *netdev =dev_get_drvdata (dev);struct macb *bp =netdev_priv (netdev); -if (!(device_may_wakeup (&bp->dev->dev))) { +if (!(device_may_wakeup (dev))) {clk_disable_unprepare (bp->tx_clk);clk_disable_unprepare (bp->hclk);clk_disable_unprepare (bp->pclk);
@@ -4632,7 +4632,7 @@ static int __maybe_unused macb_runtime_resume(struct device *dev)struct net_device *netdev =dev_get_drvdata (dev);struct macb *bp =netdev_priv (netdev); -if (!(device_may_wakeup (&bp->dev->dev))) { +if (!(device_may_wakeup (dev))) {clk_prepare_enable (bp->pclk);clk_prepare_enable (bp->hclk);clk_prepare_enable (bp->tx_clk);
Signature · 1 line
2.20.1
RFC PATCH 2/3 net: macb: mark device wake capable when "magic-packet" property present
In reply to [email protected]From: Nicolas Ferre <[email protected]>
Change the way the "magic-packet" DT property is handled in the macb_probe() function, matching DT binding documentation. Now we mark the device as "wakeup capable" instead of calling the device_init_wakeup() function that would enable the wakeup source.
For Ethernet WoL, enabling the wakeup_source is done by using ethtool and associated macb_set_wol() function that already calls device_set_wakeup_enable() for this purpose.
That would reduce power consumption by cutting more clocks if "magic-packet" property is set but WoL is not configured by ethtool.
--- drivers/net/ethernet/cadence/macb_main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/ethernet/cadence/macb_main.c +1 -1
@@ -4384,7 +4384,7 @@ static int macb_probe(struct platform_device *pdev) bp->wol =0 ;if (of_get_property (np,"magic-packet" ,NULL )) bp->wol |=MACB_WOL_HAS_MAGIC_PACKET ; -device_init_wakeup (&pdev->dev, bp->wol &MACB_WOL_HAS_MAGIC_PACKET ); +device_set_wakeup_capable (&pdev->dev, bp->wol &MACB_WOL_HAS_MAGIC_PACKET );spin_lock_init (&bp->lock);
Signature · 1 line
2.20.1
RFC PATCH 3/3 net: macb: fix macb_get/set_wol() when moving to phylink
In reply to [email protected]From: Nicolas Ferre <[email protected]>
Keep previous function goals and integrate phylink actions to them.
phylink_ethtool_get_wol() is not enough to figure out if Ethernet driver supports Wake-on-Lan.
Initialization of "supported" and "wolopts" members is done in phylink function, no need to keep them in calling function.
phylink_ethtool_set_wol() return value is not enough to determine if WoL is enabled for the calling Ethernet driver. Call if first but don't rely on its return value as most of simple PHY drivers don't implement a set_wol() function.
--- drivers/net/ethernet/cadence/macb_main.c | 18 ++++++++++-------- 1 file changed, 10 insertions(+), 8 deletions(-)
drivers/net/ethernet/cadence/macb_main.c +10 -8
@@ -2813,21 +2813,23 @@ static void macb_get_wol(struct net_device *netdev, struct ethtool_wolinfo *wol) {struct macb *bp =netdev_priv (netdev); - wol->supported =0 ; - wol->wolopts =0 ; - -if (bp->wol &MACB_WOL_HAS_MAGIC_PACKET ) +if (bp->wol &MACB_WOL_HAS_MAGIC_PACKET ) {phylink_ethtool_get_wol (bp->phylink, wol); + wol->supported |=WAKE_MAGIC ; + +if (bp->wol &MACB_WOL_ENABLED ) + wol->wolopts |=WAKE_MAGIC ; + } }static int macb_set_wol (struct net_device *netdev,struct ethtool_wolinfo *wol) {struct macb *bp =netdev_priv (netdev); -int ret; - ret =phylink_ethtool_set_wol (bp->phylink, wol); -if (!ret) -return 0 ; +/* Pass the order to phylink layer. +* Don't test return value as set_wol() is often not supported. +*/ +phylink_ethtool_set_wol (bp->phylink, wol);if (!(bp->wol &MACB_WOL_HAS_MAGIC_PACKET ) || (wol->wolopts & ~WAKE_MAGIC ))
Signature · 1 line
2.20.1
Hi Nicolas,
Quoted 10 lines · On Fri, Apr 3, 2020 at 6:45 PM <[email protected]> wrote:
From: Nicolas Ferre <[email protected]>
Hi,
Here are some of my early patches in order to fix WoL magic-packet on the current macb driver.
Addition of this feature to GEM types of IPs is yet to come. I would like to have your feedback on these little patches first so that I can continue investigating the addition of GEM WoL magic-packet.
Quoted 5 lines
Harini, I know that you have patches for GEM in order to integrate WoL ARP mode [1]. I'll try to integrate some of your work but would need that this feature is better integrated in current code. For instance, the choice of "magic packet" or "ARP" should be done by ethtool options and DT properties. For matching with mainline users, MACB and GEM code must co-exist.
Agree. I'll try to test this series and get back to you next week.
Quoted 2 lines
The use of dumb buffers for RX seems also fairly platform specific and we would need to think more about it.
I know that the IP versions from r1p10 have a mechanism to disable DMA queues (bit 0 of the queue pointer register) which is cleaner. But for earlier IP versions,
I remember discussing with Cadence and there is no way to keep RX enabled for WOL
with RX DMA disabled. I'm afraid that means there should be a bare minimum memory
region with a dummy descriptor if you do not want to process the packets. That memory
should also be accessible while the rest of the system is powered down. Please let me
know if you think of any other solution.
Regards,
Harini
Hi Harini,
Quoted 19 lines · Harini Katakam
Hi Nicolas,
Quoted 10 lines · On Fri, Apr 3, 2020 at 6:45 PM <[email protected]> wrote:
From: Nicolas Ferre <[email protected]>
Hi,
Here are some of my early patches in order to fix WoL magic-packet on the current macb driver.
Addition of this feature to GEM types of IPs is yet to come. I would like to have your feedback on these little patches first so that I can continue investigating the addition of GEM WoL magic-packet.Quoted 5 lines
Harini, I know that you have patches for GEM in order to integrate WoL ARP mode [1]. I'll try to integrate some of your work but would need that this feature is better integrated in current code. For instance, the choice of "magic packet" or "ARP" should be done by ethtool options and DT properties. For matching with mainline users, MACB and GEM code must co-exist.
Quoted 8 lines
Agree. I'll try to test this series and get back to you next week.
Quoted 2 lines
The use of dumb buffers for RX seems also fairly platform specific and we would need to think more about it.
I know that the IP versions from r1p10 have a mechanism to disable DMA queues (bit 0 of the queue pointer register) which is cleaner. But for earlier IP versions,
Which IP name are you referring to? GEM, GEM-GXL? What is the value of register 0xFC then?
Quoted 8 lines
I remember discussing with Cadence and there is no way to keep RX enabled for WOL
with RX DMA disabled. I'm afraid that means there should be a bare minimum memory
region with a dummy descriptor if you do not want to process the packets. That memory
should also be accessible while the rest of the system is powered down. Please let me
Very interesting information Harini, thanks a lot for having shared it.
My GEM IP has 0xFC at value: 0x00020203. But I don't see a way to keep DMA queues disabled by using the famous bit that you mention above.
Quoted 1 line
know if you think of any other solution.
I'm trying all this right now. I keep you posted.
Thanks and best regards,
Nicolas
Signature · 1 line
Nicolas Ferre
Hi Nicolas,
Quoted 9 lines · Nicolas Ferre
Hi Harini,
Quoted 5 lines · Harini Katakam
Hi Nicolas,
Quoted 2 lines · On Fri, Apr 3, 2020 at 6:45 PM <[email protected]> wrote:
From: Nicolas Ferre <[email protected]>
<snip>
Quoted 7 lines
Quoted 4 lines
I know that the IP versions from r1p10 have a mechanism to disable DMA queues (bit 0 of the queue pointer register) which is cleaner. But for earlier IP versions,
Which IP name are you referring to? GEM, GEM-GXL? What is the value of register
0xFCthen?
GEM_GXL
Quoted 10 lines
Quoted 9 lines
I remember discussing with Cadence and there is no way to keep RX enabled for WOL
with RX DMA disabled. I'm afraid that means there should be a bare minimum memory
region with a dummy descriptor if you do not want to process the packets. That memory
should also be accessible while the rest of the system is powered down. Please let me
Quoted 4 lines
Very interesting information Harini, thanks a lot for having shared it.
My GEM IP has
0xFCat value:0x00020203. But I don't see a way to keep DMA queues disabled by using the famous bit that you mention above.
Yeah, it is not possible in this revision. This is part of the GEM_GXL r1p10 or higher I think. I can't be sure of all the possible variations of the revision reg
because the scheme changed at some point but it looks like this:0x00070100
bits 27:16 (module_ID), bits16:0 (module_revision); they could increase.
Regards,
Harini