Skip to contentSkip to conversation

fm10k fixes for suspend/resume and related

PATCH New Jacob Keller 2 participants 37 messages Jun 7, 2016 – Jul 11, 2016

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

v2 cover fm10k fixes for suspend/resume and related

This series implements several fixes for suspend/resume and related behaviors for both PF and VF driver. The major changes are to commonize the suspend/resume, io_error_detected, reset_notify and reinit flows to use one set of functions. This ensures that all flows set the same state, preventing multiple flows from interacting together. In addition it ensures that each flow does every step that it needs. This was previously overlooked because each flow was coded separately.

In addition, there are several bug fixes scattered around the series for issues that came up when testing the changes and the bugs that these changes fix.

Changes in V2:
* This series does not include the 2 patches Jeff already applied on the queue, which are the first two patches of the v1 series
* Added 2 new patches to fix a few issues found in testing

Jacob Keller (17):
fm10k: prevent multiple threads updating statistics
fm10k: don't stop reset due to FM10K_ERR_REQUESTS_PENDING
fm10k: perform data path reset even when switch is not ready
fm10k: use actual hardware registers when checking for pending Tx
fm10k: only warn when stop_hw fails with FM10K_ERR_REQUESTS_PENDING
fm10k: wait for queues to drain if stop_hw() fails once
fm10k: split fm10k_reinit into two functions
fm10k: implement prepare_suspend and handle_resume
fm10k: use common reset flow when handling io errors from PCI stack
fm10k: implement reset_notify handler for PCIe FLR events
fm10k: use common flow for suspend and resume
fm10k: enable bus master after every reset
fm10k: check if PCIe link is restored
fm10k: implement request_lport_map pointer
fm10k: force link to remain down for at least a second on resume
events
fm10k: return proper error code when pci_enable_msix_range fails
fm10k: bump version number

Ngai-Mint Kwan (1):
fm10k: Reset mailbox global interrupts

 drivers/net/ethernet/intel/fm10k/fm10k.h         |   2 +
 drivers/net/ethernet/intel/fm10k/fm10k_common.c  |   6 +-
 drivers/net/ethernet/intel/fm10k/fm10k_ethtool.c |   2 +
 drivers/net/ethernet/intel/fm10k/fm10k_main.c    |  14 +-
 drivers/net/ethernet/intel/fm10k/fm10k_mbx.h     |   2 +
 drivers/net/ethernet/intel/fm10k/fm10k_pci.c     | 309 +++++++++++++----------
 drivers/net/ethernet/intel/fm10k/fm10k_pf.c      |  38 ++-
 drivers/net/ethernet/intel/fm10k/fm10k_type.h    |   2 +
 drivers/net/ethernet/intel/fm10k/fm10k_vf.c      |  12 +-
 9 files changed, 223 insertions(+), 164 deletions(-)
Signature · 1 line
2.9.0.rc1.405.g81f467e
#1 lore raw

PATCH v2 1/18 fm10k: prevent multiple threads updating statistics

Also prevent updating stats while the interface is down. If we're already updating stats, just return doing nothing. When we take the device down, block stat updates until we come back up. This ensures that we avoid tearing down rings when we're updating statistics, and prevents updating statistics until we're up.

We can't re-use the __FM10K_DOWN for this because it wouldn't prevent multiple threads from accessing statistics. Neither does it prevent the case where we start updating stats and then start going down in another thread.

The fm10k_get_stats64 is except from this, because it has a completely different flow which does not suffer from the same issues as fm10k_update_stats might.

---
 drivers/net/ethernet/intel/fm10k/fm10k.h     |  1 +
 drivers/net/ethernet/intel/fm10k/fm10k_pci.c | 14 ++++++++++++++
 2 files changed, 15 insertions(+)
drivers/net/ethernet/intel/fm10k/fm10k.h +1 -0
index e98b86bf0ca1..c8d0817766bf 100644
@@ -362,6 +362,7 @@ enum fm10k_state_t {
 	__FM10K_SERVICE_DISABLE,
 	__FM10K_MBX_LOCK,
 	__FM10K_LINK_DOWN,
+	__FM10K_UPDATING_STATS,
 };
 
 static inline void fm10k_mbx_lock(struct fm10k_intfc *interface)
drivers/net/ethernet/intel/fm10k/fm10k_pci.c +14 -0
index be0b7dea6e72..44393763fe77 100644
@@ -372,6 +372,10 @@ void fm10k_update_stats(struct fm10k_intfc *interface)
 	u64 bytes, pkts;
 	int i;
 
+	/* ensure only one thread updates stats at a time */
+	if (test_and_set_bit(__FM10K_UPDATING_STATS, &interface->state))
+		return;
+
 	/* do not allow stats update via service task for next second */
 	interface->next_stats_update = jiffies + HZ;
 
@@ -449,6 +453,8 @@ void fm10k_update_stats(struct fm10k_intfc *interface)
 	/* Fill out the OS statistics structure */
 	net_stats->rx_errors = rx_errors;
 	net_stats->rx_dropped = interface->stats.nodesc_drop.count;
+
+	clear_bit(__FM10K_UPDATING_STATS, &interface->state);
 }
 
 /**
@@ -1572,6 +1578,9 @@ void fm10k_up(struct fm10k_intfc *interface)
 	/* configure interrupts */
 	hw->mac.ops.update_int_moderator(hw);
 
+	/* enable statistics capture again */
+	clear_bit(__FM10K_UPDATING_STATS, &interface->state);
+
 	/* clear down bit to indicate we are ready to go */
 	clear_bit(__FM10K_DOWN, &interface->state);
 
@@ -1629,6 +1638,10 @@ void fm10k_down(struct fm10k_intfc *interface)
 	/* capture stats one last time before stopping interface */
 	fm10k_update_stats(interface);
 
+	/* prevent updating statistics while we're down */
+	while (test_and_set_bit(__FM10K_UPDATING_STATS, &interface->state))
+		usleep_range(1000, 2000);
+
 	/* Disable DMA engine for Tx/Rx */
 	err = hw->mac.ops.stop_hw(hw);
 	if (err)
@@ -1757,6 +1770,7 @@ static int fm10k_sw_init(struct fm10k_intfc *interface,
 
 	/* Start off interface as being down */
 	set_bit(__FM10K_DOWN, &interface->state);
+	set_bit(__FM10K_UPDATING_STATS, &interface->state);
 
 	return 0;
 }
Signature · 1 line
2.9.0.rc1.405.g81f467e
#2 lore raw

PATCH v2 1/18 fm10k: prevent multiple threads updating statistics

-----Original Message-----
From: Intel-wired-lan [mailto:intel-wired-lan-bounces at lists.osuosl.org] On Behalf Of Jacob Keller
Sent: Tuesday, June 7, 2016 4:09 PM
To: Intel Wired LAN <[email protected]>
Subject: [Intel-wired-lan] [PATCH v2 01/18] fm10k: prevent multiple threads updating statistics

Also prevent updating stats while the interface is down. If we're already updating stats, just return doing nothing. When we take the device down, block stat updates until we come back up. This ensures that we avoid tearing down rings when we're updating statistics, and prevents updating statistics until we're up.

We can't re-use the __FM10K_DOWN for this because it wouldn't prevent multiple threads from accessing statistics. Neither does it prevent the case where we start updating stats and then start going down in another thread.

The fm10k_get_stats64 is except from this, because it has a completely different flow which does not suffer from the same issues as fm10k_update_stats might.

---

#3 lore raw

PATCH v2 2/18 fm10k: Reset mailbox global interrupts

In reply to Jacob Keller

From: Ngai-Mint Kwan <[email protected]>

When a data path reset is initiated, write control to the PCIE_GMBX is yanked from the switch manager. The switch manager writes to this register to clear mailbox global interrupt bits as part of its mailbox interrupt handling routine. When the device recovers from the data path reset and these bits are not cleared, it will prevent future mailbox global interrupts from being triggered. Upon confirming that the device has exited from a data path reset, clear these bits to ensure the proper functioning of the mailbox global interrupt.

---
 drivers/net/ethernet/intel/fm10k/fm10k_mbx.h | 2 ++
 drivers/net/ethernet/intel/fm10k/fm10k_pf.c  | 4 ++++
 2 files changed, 6 insertions(+)
drivers/net/ethernet/intel/fm10k/fm10k_mbx.h +2 -0
index b7dbc8a84c05..35c1dbad1330 100644
@@ -41,6 +41,8 @@ struct fm10k_mbx_info;
 #define FM10K_MBX_ACK_INTERRUPT			0x00000010
 #define FM10K_MBX_INTERRUPT_ENABLE		0x00000020
 #define FM10K_MBX_INTERRUPT_DISABLE		0x00000040
+#define FM10K_MBX_GLOBAL_REQ_INTERRUPT		0x00000200
+#define FM10K_MBX_GLOBAL_ACK_INTERRUPT		0x00000400
 #define FM10K_MBICR(_n)		((_n) + 0x18840)
 #define FM10K_GMBX		0x18842
 
drivers/net/ethernet/intel/fm10k/fm10k_pf.c +4 -0
index dc75507c9926..69e2c822db00 100644
@@ -77,6 +77,10 @@ static s32 fm10k_reset_hw_pf(struct fm10k_hw *hw)
 	if (!(reg & FM10K_IP_NOTINRESET))
 		err = FM10K_ERR_RESET_FAILED;
 
+	/* Reset mailbox global interrupts */
+	reg = FM10K_MBX_GLOBAL_REQ_INTERRUPT | FM10K_MBX_GLOBAL_ACK_INTERRUPT;
+	fm10k_write_reg(hw, FM10K_GMBX, reg);
+
 out:
 	return err;
 }
Signature · 1 line
2.9.0.rc1.405.g81f467e
#4 lore raw

PATCH v2 2/18 fm10k: Reset mailbox global interrupts

-----Original Message-----
From: Intel-wired-lan [mailto:intel-wired-lan-bounces at lists.osuosl.org] On Behalf Of Jacob Keller
Sent: Tuesday, June 7, 2016 4:09 PM
To: Intel Wired LAN <[email protected]>
Cc: Kwan, Ngai-mint <[email protected]>
Subject: [Intel-wired-lan] [PATCH v2 02/18] fm10k: Reset mailbox global interrupts

From: Ngai-Mint Kwan <[email protected]>

When a data path reset is initiated, write control to the PCIE_GMBX is yanked from the switch manager. The switch manager writes to this register to clear mailbox global interrupt bits as part of its mailbox interrupt handling routine. When the device recovers from the data path reset and these bits are not cleared, it will prevent future mailbox global interrupts from being triggered. Upon confirming that the device has exited from a data path reset, clear these bits to ensure the proper functioning of the mailbox global interrupt.

---

#5 lore raw

PATCH v2 3/18 fm10k: don't stop reset due to FM10K_ERR_REQUESTS_PENDING

In reply to Jacob Keller

Don't report FM10K_ERR_REQUESTS_PENDING when we fail to disable queues within the timeout. This can occur due to a hardware Tx hang, or when the switch ethernet fabric is resetting while we are transmitting traffic. It can sometimes take up to 500ms before the Tx DMA engine gives up. Instead, just skip the DMA engine check and perform a data-path reset anyways. Add a statistic counter to keep track of the number of resets occurring while we have pending DMA on the rings.

In order to prevent having to re-assign err to 0, re-order the last few items of the reset_hw_pf function so that we don't perform "return err" at the end.

---
 drivers/net/ethernet/intel/fm10k/fm10k_ethtool.c |  2 ++
 drivers/net/ethernet/intel/fm10k/fm10k_pf.c      | 24 ++++++++++++++----------
 drivers/net/ethernet/intel/fm10k/fm10k_type.h    |  1 +
 drivers/net/ethernet/intel/fm10k/fm10k_vf.c      | 12 +++++++-----
 4 files changed, 24 insertions(+), 15 deletions(-)
drivers/net/ethernet/intel/fm10k/fm10k_ethtool.c +2 -0
index 9b5195435c87..c04cbe9c9f7c 100644
@@ -76,6 +76,8 @@ static const struct fm10k_stats fm10k_gstrings_global_stats[] = {
 	FM10K_STAT("mac_rules_used", hw.swapi.mac.used),
 	FM10K_STAT("mac_rules_avail", hw.swapi.mac.avail),
 
+	FM10K_STAT("reset_while_pending", hw.mac.reset_while_pending),
+
 	FM10K_STAT("tx_hang_count", tx_timeout_count),
 };
 
drivers/net/ethernet/intel/fm10k/fm10k_pf.c +14 -10
index 69e2c822db00..7fbd94ba6745 100644
@@ -51,8 +51,12 @@ static s32 fm10k_reset_hw_pf(struct fm10k_hw *hw)
 
 	/* shut down all rings */
 	err = fm10k_disable_queues_generic(hw, FM10K_MAX_QUEUES);
-	if (err)
+	if (err == FM10K_ERR_REQUESTS_PENDING) {
+		hw->mac.reset_while_pending++;
+		goto force_reset;
+	} else if (err) {
 		return err;
+	}
 
 	/* Verify that DMA is no longer active */
 	reg = fm10k_read_reg(hw, FM10K_DMA_CTRL);
@@ -62,27 +66,27 @@ static s32 fm10k_reset_hw_pf(struct fm10k_hw *hw)
 	/* verify the switch is ready for reset */
 	reg = fm10k_read_reg(hw, FM10K_DMA_CTRL2);
 	if (!(reg & FM10K_DMA_CTRL2_SWITCH_READY))
-		goto out;
+		return FM10K_ERR_DMA_PENDING;
 
+force_reset:
 	/* Inititate data path reset */
-	reg |= FM10K_DMA_CTRL_DATAPATH_RESET;
+	reg = FM10K_DMA_CTRL_DATAPATH_RESET;
 	fm10k_write_reg(hw, FM10K_DMA_CTRL, reg);
 
 	/* Flush write and allow 100us for reset to complete */
 	fm10k_write_flush(hw);
 	udelay(FM10K_RESET_TIMEOUT);
 
-	/* Verify we made it out of reset */
-	reg = fm10k_read_reg(hw, FM10K_IP);
-	if (!(reg & FM10K_IP_NOTINRESET))
-		err = FM10K_ERR_RESET_FAILED;
-
 	/* Reset mailbox global interrupts */
 	reg = FM10K_MBX_GLOBAL_REQ_INTERRUPT | FM10K_MBX_GLOBAL_ACK_INTERRUPT;
 	fm10k_write_reg(hw, FM10K_GMBX, reg);
 
-out:
-	return err;
+	/* Verify we made it out of reset */
+	reg = fm10k_read_reg(hw, FM10K_IP);
+	if (!(reg & FM10K_IP_NOTINRESET))
+		return FM10K_ERR_RESET_FAILED;
+
+	return 0;
 }
 
 /**
drivers/net/ethernet/intel/fm10k/fm10k_type.h +1 -0
index b8bc06183720..1d65ad85d72e 100644
@@ -562,6 +562,7 @@ struct fm10k_mac_info {
 	bool tx_ready;
 	u32 dglort_map;
 	u8 itr_scale;
+	u64 reset_while_pending;
 };
 
 struct fm10k_swapi_table_info {
drivers/net/ethernet/intel/fm10k/fm10k_vf.c +7 -5
index 3b06685ea63b..337ba65a9411 100644
@@ -34,7 +34,7 @@ static s32 fm10k_stop_hw_vf(struct fm10k_hw *hw)
 
 	/* we need to disable the queues before taking further steps */
 	err = fm10k_stop_hw_generic(hw);
-	if (err)
+	if (err && err != FM10K_ERR_REQUESTS_PENDING)
 		return err;
 
 	/* If permanent address is set then we need to restore it */
@@ -67,7 +67,7 @@ static s32 fm10k_stop_hw_vf(struct fm10k_hw *hw)
 		fm10k_write_reg(hw, FM10K_TDLEN(i), tdlen);
 	}
 
-	return 0;
+	return err;
 }
 
 /**
@@ -83,7 +83,9 @@ static s32 fm10k_reset_hw_vf(struct fm10k_hw *hw)
 
 	/* shut down queues we own and reset DMA configuration */
 	err = fm10k_stop_hw_vf(hw);
-	if (err)
+	if (err == FM10K_ERR_REQUESTS_PENDING)
+		hw->mac.reset_while_pending++;
+	else if (err)
 		return err;
 
 	/* Inititate VF reset */
@@ -96,9 +98,9 @@ static s32 fm10k_reset_hw_vf(struct fm10k_hw *hw)
 	/* Clear reset bit and verify it was cleared */
 	fm10k_write_reg(hw, FM10K_VFCTRL, 0);
 	if (fm10k_read_reg(hw, FM10K_VFCTRL) & FM10K_VFCTRL_RST)
-		err = FM10K_ERR_RESET_FAILED;
+		return FM10K_ERR_RESET_FAILED;
 
-	return err;
+	return 0;
 }
 
 /**
Signature · 1 line
2.9.0.rc1.405.g81f467e
#6 lore raw

PATCH v2 3/18 fm10k: don't stop reset due to FM10K_ERR_REQUESTS_PENDING

-----Original Message-----
From: Intel-wired-lan [mailto:intel-wired-lan-bounces at lists.osuosl.org] On Behalf Of Jacob Keller
Sent: Tuesday, June 7, 2016 4:09 PM
To: Intel Wired LAN <[email protected]>
Subject: [Intel-wired-lan] [PATCH v2 03/18] fm10k: don't stop reset due to FM10K_ERR_REQUESTS_PENDING

Don't report FM10K_ERR_REQUESTS_PENDING when we fail to disable queues within the timeout. This can occur due to a hardware Tx hang, or when the switch ethernet fabric is resetting while we are transmitting traffic. It can sometimes take up to 500ms before the Tx DMA engine gives up. Instead, just skip the DMA engine check and perform a data-path reset anyways. Add a statistic counter to keep track of the number of resets occurring while we have pending DMA on the rings.

In order to prevent having to re-assign err to 0, re-order the last few items of the reset_hw_pf function so that we don't perform "return err" at the end.

---

#7 lore raw

PATCH v2 4/18 fm10k: perform data path reset even when switch is not ready

In reply to Jacob Keller

A while ago, an additional check for the switch being ready was added to reset_hw. A recent refactor accidentally made this check return an error code on failure which caused fm10k_probe to fail when the switch wasn't brought up first. The original reasoning for the check was to prevent additional data path reset when the fabric wasn't ready yet. However, there isn't a compelling reason to keep the check, as the data path reset will restore hardware to a known good state. Remove the check and perform the data path reset regardless of the switch manager state.

An alternative fix is to return FM10K_SUCCESS instead, and bypass the actual data path reset. This should be fine as we will perform a reset_hw once the switch is active. However, since data path reset will reset many parts of the hardware it seems better to just perform the reset regardless of switch state.

---
 drivers/net/ethernet/intel/fm10k/fm10k_pf.c | 5 -----
 1 file changed, 5 deletions(-)
drivers/net/ethernet/intel/fm10k/fm10k_pf.c +0 -5
index 7fbd94ba6745..23f3566b17fc 100644
@@ -63,11 +63,6 @@ static s32 fm10k_reset_hw_pf(struct fm10k_hw *hw)
 	if (reg & (FM10K_DMA_CTRL_TX_ACTIVE | FM10K_DMA_CTRL_RX_ACTIVE))
 		return FM10K_ERR_DMA_PENDING;
 
-	/* verify the switch is ready for reset */
-	reg = fm10k_read_reg(hw, FM10K_DMA_CTRL2);
-	if (!(reg & FM10K_DMA_CTRL2_SWITCH_READY))
-		return FM10K_ERR_DMA_PENDING;
-
 force_reset:
 	/* Inititate data path reset */
 	reg = FM10K_DMA_CTRL_DATAPATH_RESET;
Signature · 1 line
2.9.0.rc1.405.g81f467e
#8 lore raw

PATCH v2 4/18 fm10k: perform data path reset even when switch is not ready

-----Original Message-----
From: Intel-wired-lan [mailto:intel-wired-lan-bounces at lists.osuosl.org] On Behalf Of Jacob Keller
Sent: Tuesday, June 7, 2016 4:09 PM
To: Intel Wired LAN <[email protected]>
Subject: [Intel-wired-lan] [PATCH v2 04/18] fm10k: perform data path reset even when switch is not ready

A while ago, an additional check for the switch being ready was added to reset_hw. A recent refactor accidentally made this check return an error code on failure which caused fm10k_probe to fail when the switch wasn't brought up first. The original reasoning for the check was to prevent additional data path reset when the fabric wasn't ready yet. However, there isn't a compelling reason to keep the check, as the data path reset will restore hardware to a known good state. Remove the check and perform the data path reset regardless of the switch manager state.

An alternative fix is to return FM10K_SUCCESS instead, and bypass the actual data path reset. This should be fine as we will perform a reset_hw once the switch is active. However, since data path reset will reset many parts of the hardware it seems better to just perform the reset regardless of switch state.

---

#9 lore raw

PATCH v2 5/18 fm10k: use actual hardware registers when checking for pending Tx

In reply to Jacob Keller
---
 drivers/net/ethernet/intel/fm10k/fm10k_main.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)
drivers/net/ethernet/intel/fm10k/fm10k_main.c +5 -3
index a9ccc1eb3ea4..c6a464551577 100644
@@ -1130,9 +1130,11 @@ static u64 fm10k_get_tx_completed(struct fm10k_ring *ring)
 
 static u64 fm10k_get_tx_pending(struct fm10k_ring *ring)
 {
-	/* use SW head and tail until we have real hardware */
-	u32 head = ring->next_to_clean;
-	u32 tail = ring->next_to_use;
+	struct fm10k_intfc *interface = ring->q_vector->interface;
+	struct fm10k_hw *hw = &interface->hw;
+
+	u32 head = fm10k_read_reg(hw, FM10K_TDH(ring->reg_idx));
+	u32 tail = fm10k_read_reg(hw, FM10K_TDT(ring->reg_idx));
 
 	return ((head <= tail) ? tail : tail + ring->count) - head;
 }
Signature · 1 line
2.9.0.rc1.405.g81f467e
#10 lore raw

PATCH v2 5/18 fm10k: use actual hardware registers when checking for pending Tx

-----Original Message-----
From: Intel-wired-lan [mailto:intel-wired-lan-bounces at lists.osuosl.org] On Behalf Of Jacob Keller
Sent: Tuesday, June 7, 2016 4:09 PM
To: Intel Wired LAN <[email protected]>
Subject: [Intel-wired-lan] [PATCH v2 05/18] fm10k: use actual hardware registers when checking for pending Tx

---

#11 lore raw

PATCH v2 6/18 fm10k: only warn when stop_hw fails with FM10K_ERR_REQUESTS_PENDING

In reply to Jacob Keller

When stop_hw() routine fails with FM10K_ERR_REQUESTS_PENDING, this indicates that the Tx or Rx queues did not shutdown within the time limit. Print a more suitable message at the dev_info level instead of dev_err.

---
 drivers/net/ethernet/intel/fm10k/fm10k_pci.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/net/ethernet/intel/fm10k/fm10k_pci.c +4 -1
index 44393763fe77..4dfd1284a8de 100644
@@ -1644,7 +1644,10 @@ void fm10k_down(struct fm10k_intfc *interface)
 
 	/* Disable DMA engine for Tx/Rx */
 	err = hw->mac.ops.stop_hw(hw);
-	if (err)
+	if (err == FM10K_ERR_REQUESTS_PENDING)
+		dev_info(&interface->pdev->dev,
+			 "due to pending requests hw was not shut down gracefully\n");
+	else if (err)
 		dev_err(&interface->pdev->dev, "stop_hw failed: %d\n", err);
 
 	/* free any buffers still on the rings */
Signature · 1 line
2.9.0.rc1.405.g81f467e
#12 lore raw

PATCH v2 6/18 fm10k: only warn when stop_hw fails with FM10K_ERR_REQUESTS_PENDING

-----Original Message-----
From: Intel-wired-lan [mailto:intel-wired-lan-bounces at lists.osuosl.org] On Behalf Of Jacob Keller
Sent: Tuesday, June 7, 2016 4:09 PM
To: Intel Wired LAN <[email protected]>
Subject: [Intel-wired-lan] [PATCH v2 06/18] fm10k: only warn when stop_hw fails with FM10K_ERR_REQUESTS_PENDING

When stop_hw() routine fails with FM10K_ERR_REQUESTS_PENDING, this indicates that the Tx or Rx queues did not shutdown within the time limit. Print a more suitable message at the dev_info level instead of dev_err.

---

#13 lore raw

PATCH v2 7/18 fm10k: wait for queues to drain if stop_hw() fails once

In reply to Jacob Keller

It turns out that sometimes during a reset the Tx queues will be temporarily stuck longer than .stop_hw() expects. Work around this issue by attempting to .stop_hw() first. If it tails, wait a number of attempts until the Tx queues appear to be drained. After this, attempt stop_hw() again. This ensures that we avoid waiting if we don't need to, such as during the first initialization of a VF, and give the proper amount of time necessary to recover from most situations. It is possible that the hardware is actually stuck. For PFs, this is usually fixed by a datapath reset. Unfortunately the VF cannot request a similar reset for itself.

---
 drivers/net/ethernet/intel/fm10k/fm10k.h      |  1 +
 drivers/net/ethernet/intel/fm10k/fm10k_main.c |  2 +-
 drivers/net/ethernet/intel/fm10k/fm10k_pci.c  | 44 +++++++++++++++++++++++----
 3 files changed, 40 insertions(+), 7 deletions(-)
drivers/net/ethernet/intel/fm10k/fm10k.h +1 -0
index c8d0817766bf..c4cf08dcf5af 100644
@@ -458,6 +458,7 @@ __be16 fm10k_tx_encap_offload(struct sk_buff *skb);
 netdev_tx_t fm10k_xmit_frame_ring(struct sk_buff *skb,
 				  struct fm10k_ring *tx_ring);
 void fm10k_tx_timeout_reset(struct fm10k_intfc *interface);
+u64 fm10k_get_tx_pending(struct fm10k_ring *ring);
 bool fm10k_check_tx_hang(struct fm10k_ring *tx_ring);
 void fm10k_alloc_rx_buffers(struct fm10k_ring *rx_ring, u16 cleaned_count);
 
drivers/net/ethernet/intel/fm10k/fm10k_main.c +1 -1
index c6a464551577..c85fc98945fa 100644
@@ -1128,7 +1128,7 @@ static u64 fm10k_get_tx_completed(struct fm10k_ring *ring)
 	return ring->stats.packets;
 }
 
-static u64 fm10k_get_tx_pending(struct fm10k_ring *ring)
+u64 fm10k_get_tx_pending(struct fm10k_ring *ring)
 {
 	struct fm10k_intfc *interface = ring->q_vector->interface;
 	struct fm10k_hw *hw = &interface->hw;
drivers/net/ethernet/intel/fm10k/fm10k_pci.c +38 -6

65 lines not on this page · raw message

Signature · 1 line
2.9.0.rc1.405.g81f467e
#14 lore raw

PATCH v2 7/18 fm10k: wait for queues to drain if stop_hw() fails once

-----Original Message-----
From: Intel-wired-lan [mailto:intel-wired-lan-bounces at lists.osuosl.org] On Behalf Of Jacob Keller
Sent: Tuesday, June 7, 2016 4:09 PM
To: Intel Wired LAN <[email protected]>
Subject: [Intel-wired-lan] [PATCH v2 07/18] fm10k: wait for queues to drain if stop_hw() fails once

It turns out that sometimes during a reset the Tx queues will be temporarily stuck longer than .stop_hw() expects. Work around this issue by attempting to .stop_hw() first. If it tails, wait a number of attempts until the Tx queues appear to be drained. After this, attempt
stop_hw() again. This ensures that we avoid waiting if we don't need to, such as during the first initialization of a VF, and give the proper amount of time necessary to recover from most situations. It is possible that the hardware is actually stuck. For PFs, this is usually fixed by a datapath reset. Unfortunately the VF cannot request a similar reset for itself.

---

#15 lore raw

PATCH v2 8/18 fm10k: split fm10k_reinit into two functions

In reply to Jacob Keller

There are several flows in the driver which perform the similar function of tearing down software and restoring software to recover from certain errors or PCIe events, including:

* fm10k_reinit
* fm10k_suspend/resume
* fm10k_io_error_detected/fm10k_io_resume

In addition, we want to implement a .reset_notify() handler as well which will also perform similar function.

Rework how the driver codes reset and resume flows by separating out the reinit logic into two functions "fm10k_prepare_for_reset" and "fm10k_handle_reset". This first step will allow us to re-use this functionality in the similar blocks of code instead of re-coding the same sequence of events slightly different.

The end result should be more maintainable and correct, fixing several inconsistencies with the work flow.

The new functions expect to take the rtnl_lock() themselves, and it does have the unfortunate side effect of having the reinit flow take then release then take the rtnl_lock. However, this minor downside is outweighted by the benefits of code reduction and reducing needless difference between these flows.

---
 drivers/net/ethernet/intel/fm10k/fm10k_pci.c | 33 +++++++++++++++++++++++-----
 1 file changed, 28 insertions(+), 5 deletions(-)
drivers/net/ethernet/intel/fm10k/fm10k_pci.c +28 -5
index 7c9b20c6b6c1..2963b41d01e2 100644
@@ -136,11 +136,9 @@ static void fm10k_detach_subtask(struct fm10k_intfc *interface)
 	rtnl_unlock();
 }
 
-static void fm10k_reinit(struct fm10k_intfc *interface)
+static void fm10k_prepare_for_reset(struct fm10k_intfc *interface)
 {
 	struct net_device *netdev = interface->netdev;
-	struct fm10k_hw *hw = &interface->hw;
-	int err;
 
 	WARN_ON(in_interrupt());
 
@@ -165,6 +163,17 @@ static void fm10k_reinit(struct fm10k_intfc *interface)
 	/* delay any future reset requests */
 	interface->last_reset = jiffies + (10 * HZ);
 
+	rtnl_unlock();
+}
+
+static int fm10k_handle_reset(struct fm10k_intfc *interface)
+{
+	struct net_device *netdev = interface->netdev;
+	struct fm10k_hw *hw = &interface->hw;
+	int err;
+
+	rtnl_lock();
+
 	/* reset and initialize the hardware so it is in a known state */
 	err = hw->mac.ops.reset_hw(hw);
 	if (err) {
@@ -185,7 +194,7 @@ static void fm10k_reinit(struct fm10k_intfc *interface)
 		goto reinit_err;
 	}
 
-	/* reassociate interrupts */
+	/* re-associate interrupts */
 	err = fm10k_mbx_request_irq(interface);
 	if (err)
 		goto err_mbx_irq;
@@ -219,7 +228,7 @@ static void fm10k_reinit(struct fm10k_intfc *interface)
 
 	clear_bit(__FM10K_RESETTING, &interface->state);
 
-	return;
+	return err;
 err_open:
 	fm10k_mbx_free_irq(interface);
 err_mbx_irq:
@@ -230,6 +239,20 @@ reinit_err:
 	rtnl_unlock();
 
 	clear_bit(__FM10K_RESETTING, &interface->state);
+
+	return err;
+}
+
+static void fm10k_reinit(struct fm10k_intfc *interface)
+{
+	int err;
+
+	fm10k_prepare_for_reset(interface);
+
+	err = fm10k_handle_reset(interface);
+	if (err)
+		dev_err(&interface->pdev->dev,
+			"fm10k_handle_reset failed: %d\n", err);
 }
 
 static void fm10k_reset_subtask(struct fm10k_intfc *interface)
Signature · 1 line
2.9.0.rc1.405.g81f467e
#16 lore raw

PATCH v2 8/18 fm10k: split fm10k_reinit into two functions

-----Original Message-----
From: Intel-wired-lan [mailto:intel-wired-lan-bounces at lists.osuosl.org] On Behalf Of Jacob Keller
Sent: Tuesday, June 7, 2016 4:09 PM
To: Intel Wired LAN <[email protected]>
Subject: [Intel-wired-lan] [PATCH v2 08/18] fm10k: split fm10k_reinit into two functions

There are several flows in the driver which perform the similar function of tearing down software and restoring software to recover from certain errors or PCIe events, including:

* fm10k_reinit
* fm10k_suspend/resume
* fm10k_io_error_detected/fm10k_io_resume

In addition, we want to implement a .reset_notify() handler as well which will also perform similar function.

Rework how the driver codes reset and resume flows by separating out the reinit logic into two functions "fm10k_prepare_for_reset" and "fm10k_handle_reset". This first step will allow us to re-use this functionality in the similar blocks of code instead of re-coding the same sequence of events slightly different.

The end result should be more maintainable and correct, fixing several inconsistencies with the work flow.

The new functions expect to take the rtnl_lock() themselves, and it does have the unfortunate side effect of having the reinit flow take then release then take the rtnl_lock. However, this minor downside is outweighted by the benefits of code reduction and reducing needless difference between these flows.

---

#17 lore raw

PATCH v2 9/18 fm10k: implement prepare_suspend and handle_resume

In reply to Jacob Keller

Implement fm10k_prepare_suspend and fm10k_handle_resume functions which abstract around the now existing fm10k_prepare_for_reset and fm10k_handle_reset. The new functions also handle stopping the service task, which is something that the original re-init flow does not need.

Every other location that does a suspend/resume type flow is expected to use these functions, because otherwise they may have conflicts with the running watchdog routines. This also has the effect of preventing possible surprise remove events during handling of FLR events and PCIe errors.

---
 drivers/net/ethernet/intel/fm10k/fm10k_pci.c | 38 ++++++++++++++++++++++++++++
 1 file changed, 38 insertions(+)
drivers/net/ethernet/intel/fm10k/fm10k_pci.c +38 -0
index 2963b41d01e2..a6ee046e9824 100644
@@ -2112,6 +2112,44 @@ static void fm10k_remove(struct pci_dev *pdev)
 	pci_disable_device(pdev);
 }
 
+static void fm10k_prepare_suspend(struct fm10k_intfc *interface)
+{
+	/* the watchdog task reads from registers, which might appear like
+	 * a surprise remove if the PCIe device is disabled while we're
+	 * stopped. We stop the watchdog task until after we resume software
+	 * activity.
+	 */
+	set_bit(__FM10K_SERVICE_DISABLE, &interface->state);
+	cancel_work_sync(&interface->service_task);
+
+	fm10k_prepare_for_reset(interface);
+}
+
+static int fm10k_handle_resume(struct fm10k_intfc *interface)
+{
+	struct fm10k_hw *hw = &interface->hw;
+	int err;
+
+	/* reset statistics starting values */
+	hw->mac.ops.rebind_hw_stats(hw, &interface->stats);
+
+	err = fm10k_handle_reset(interface);
+	if (err)
+		return err;
+
+	/* assume host is not ready, to prevent race with watchdog in case we
+	 * actually don't have connection to the switch
+	 */
+	interface->host_ready = false;
+	fm10k_watchdog_host_not_ready(interface);
+
+	/* clear the service task disable bit to allow service task to start */
+	clear_bit(__FM10K_SERVICE_DISABLE, &interface->state);
+	fm10k_service_event_schedule(interface);
+
+	return err;
+}
+
 #ifdef CONFIG_PM
 /**
  * fm10k_resume - Restore device to pre-sleep state
Signature · 1 line
2.9.0.rc1.405.g81f467e
#18 lore raw

PATCH v2 9/18 fm10k: implement prepare_suspend and handle_resume

-----Original Message-----
From: Intel-wired-lan [mailto:intel-wired-lan-bounces at lists.osuosl.org] On Behalf Of Jacob Keller
Sent: Tuesday, June 7, 2016 4:09 PM
To: Intel Wired LAN <[email protected]>
Subject: [Intel-wired-lan] [PATCH v2 09/18] fm10k: implement prepare_suspend and handle_resume

Implement fm10k_prepare_suspend and fm10k_handle_resume functions which abstract around the now existing fm10k_prepare_for_reset and fm10k_handle_reset. The new functions also handle stopping the service task, which is something that the original re-init flow does not need.

Every other location that does a suspend/resume type flow is expected to use these functions, because otherwise they may have conflicts with the running watchdog routines. This also has the effect of preventing possible surprise remove events during handling of FLR events and PCIe errors.

---

#19 lore raw

PATCH v2 10/18 fm10k: use common reset flow when handling io errors from PCI stack

In reply to Jacob Keller

Now that we have extracted the necessary steps for a split suspend/resume flow, re-use these functions instead of using the current open coded flow. This ensures that we don't miss any steps. It also ensures that we have the correct driver states set.

Since we'll be handling all of the reset flow ourselves, we no longer need to request a reset in the io_slot_reset() function.

---
 drivers/net/ethernet/intel/fm10k/fm10k_pci.c | 60 ++++------------------------
 1 file changed, 7 insertions(+), 53 deletions(-)
drivers/net/ethernet/intel/fm10k/fm10k_pci.c +7 -53
index a6ee046e9824..716a5c852143 100644
@@ -2312,17 +2312,7 @@ static pci_ers_result_t fm10k_io_error_detected(struct pci_dev *pdev,
 	if (state == pci_channel_io_perm_failure)
 		return PCI_ERS_RESULT_DISCONNECT;
 
-	rtnl_lock();
-
-	if (netif_running(netdev))
-		fm10k_close(netdev);
-
-	fm10k_mbx_free_irq(interface);
-
-	/* free interrupts */
-	fm10k_clear_queueing_scheme(interface);
-
-	rtnl_unlock();
+	fm10k_prepare_suspend(interface);
 
 	/* Request a slot reset. */
 	return PCI_ERS_RESULT_NEED_RESET;
@@ -2336,7 +2326,6 @@ static pci_ers_result_t fm10k_io_error_detected(struct pci_dev *pdev,
  */
 static pci_ers_result_t fm10k_io_slot_reset(struct pci_dev *pdev)
 {
-	struct fm10k_intfc *interface = pci_get_drvdata(pdev);
 	pci_ers_result_t result;
 
 	if (pci_enable_device_mem(pdev)) {
@@ -2354,12 +2343,6 @@ static pci_ers_result_t fm10k_io_slot_reset(struct pci_dev *pdev)
 
 		pci_wake_from_d3(pdev, false);
 
-		/* refresh hw_addr in case it was dropped */
-		interface->hw.hw_addr = interface->uc_addr;
-
-		interface->flags |= FM10K_FLAG_RESET_REQUESTED;
-		fm10k_service_event_schedule(interface);
-
 		result = PCI_ERS_RESULT_RECOVERED;
 	}
 
@@ -2379,44 +2362,15 @@ static void fm10k_io_resume(struct pci_dev *pdev)
 {
 	struct fm10k_intfc *interface = pci_get_drvdata(pdev);
 	struct net_device *netdev = interface->netdev;
-	struct fm10k_hw *hw = &interface->hw;
-	int err = 0;
+	int err;
 
-	/* reset hardware to known state */
-	err = hw->mac.ops.init_hw(&interface->hw);
-	if (err) {
-		dev_err(&pdev->dev, "init_hw failed: %d\n", err);
-		return;
-	}
+	err = fm10k_handle_resume(interface);
 
-	/* reset statistics starting values */
-	hw->mac.ops.rebind_hw_stats(hw, &interface->stats);
-
-	rtnl_lock();
-
-	err = fm10k_init_queueing_scheme(interface);
-	if (err) {
-		dev_err(&interface->pdev->dev,
-			"init_queueing_scheme failed: %d\n", err);
-		goto unlock;
-	}
-
-	/* reassociate interrupts */
-	fm10k_mbx_request_irq(interface);
-
-	rtnl_lock();
-	if (netif_running(netdev))
-		err = fm10k_open(netdev);
-	rtnl_unlock();
-
-	/* final check of hardware state before registering the interface */
-	err = err ? : fm10k_hw_ready(interface);
-
-	if (!err)
+	if (err)
+		dev_warn(&pdev->dev,
+			 "fm10k_io_resume failed: %d\n", err);
+	else
 		netif_device_attach(netdev);
-
-unlock:
-	rtnl_unlock();
 }
 
 static const struct pci_error_handlers fm10k_err_handler = {
Signature · 1 line
2.9.0.rc1.405.g81f467e
#20 lore raw

PATCH v2 10/18 fm10k: use common reset flow when handling io errors from PCI stack

-----Original Message-----
From: Intel-wired-lan [mailto:intel-wired-lan-bounces at lists.osuosl.org] On Behalf Of Jacob Keller
Sent: Tuesday, June 7, 2016 4:09 PM
To: Intel Wired LAN <[email protected]>
Subject: [Intel-wired-lan] [PATCH v2 10/18] fm10k: use common reset flow when handling io errors from PCI stack

Now that we have extracted the necessary steps for a split suspend/resume flow, re-use these functions instead of using the current open coded flow. This ensures that we don't miss any steps. It also ensures that we have the correct driver states set.

Since we'll be handling all of the reset flow ourselves, we no longer need to request a reset in the io_slot_reset() function.

---

#21 lore raw

PATCH v2 11/18 fm10k: implement reset_notify handler for PCIe FLR events

In reply to Jacob Keller

When a function level PCI reset is triggered using sysfs, it calls the driver's .reset_notify error handler. Implement a handler based on the now split fm10k_prepare_for_reset and fm10k_handle_reset functions, so that we fully reset the driver when the PCI function level reset occurs. This also ensures the reset is handled in a clean way by first disabling all the driver bits first and then restoring them after the function reset. Previously the stack simply performed a blind function reset and our driver didn't take any part in the process.

---
 drivers/net/ethernet/intel/fm10k/fm10k_pci.c | 33 ++++++++++++++++++++++++++++
 1 file changed, 33 insertions(+)
drivers/net/ethernet/intel/fm10k/fm10k_pci.c +33 -0
index 716a5c852143..c60c3b0f89e9 100644
@@ -2373,10 +2373,43 @@ static void fm10k_io_resume(struct pci_dev *pdev)
 		netif_device_attach(netdev);
 }
 
+/**
+ * fm10k_io_reset_notify - called when PCI function is reset
+ * @pdev: Pointer to PCI device
+ *
+ * This callback is called when the PCI function is reset such as from
+ * /sys/class/net/<enpX>/device/reset or similar. When prepare is true, it
+ * means we should prepare for a function reset. If prepare is false, it means
+ * the function reset just occurred.
+ */
+static void fm10k_io_reset_notify(struct pci_dev *pdev, bool prepare)
+{
+	struct fm10k_intfc *interface = pci_get_drvdata(pdev);
+	int err = 0;
+
+	if (prepare) {
+		/* warn incase we have any active VF devices */
+		if (pci_num_vf(pdev))
+			dev_warn(&pdev->dev,
+				 "PCIe FLR may cause issues for any active VF devices\n");
+
+		fm10k_prepare_suspend(interface);
+	} else {
+		err = fm10k_handle_resume(interface);
+	}
+
+	if (err) {
+		dev_warn(&pdev->dev,
+			 "fm10k_io_reset_notify failed: %d\n", err);
+		netif_device_detach(interface->netdev);
+	}
+}
+
 static const struct pci_error_handlers fm10k_err_handler = {
 	.error_detected = fm10k_io_error_detected,
 	.slot_reset = fm10k_io_slot_reset,
 	.resume = fm10k_io_resume,
+	.reset_notify = fm10k_io_reset_notify,
 };
 
 static struct pci_driver fm10k_driver = {
Signature · 1 line
2.9.0.rc1.405.g81f467e
#22 lore raw

PATCH v2 11/18 fm10k: implement reset_notify handler for PCIe FLR events

-----Original Message-----
From: Intel-wired-lan [mailto:intel-wired-lan-bounces at lists.osuosl.org] On Behalf Of Jacob Keller
Sent: Tuesday, June 7, 2016 4:09 PM
To: Intel Wired LAN <[email protected]>
Subject: [Intel-wired-lan] [PATCH v2 11/18] fm10k: implement reset_notify handler for PCIe FLR events

When a function level PCI reset is triggered using sysfs, it calls the driver's .reset_notify error handler. Implement a handler based on the now split fm10k_prepare_for_reset and fm10k_handle_reset functions, so that we fully reset the driver when the PCI function level reset occurs. This also ensures the reset is handled in a clean way by first disabling all the driver bits first and then restoring them after the function reset. Previously the stack simply performed a blind function reset and our driver didn't take any part in the process.

---

#23 lore raw

PATCH v2 12/18 fm10k: use common flow for suspend and resume

In reply to Jacob Keller

Continuing the effort to commonize the similar suspend/resume flows, finish up by using the new fm10k_handle_suspand and fm10k_handle_resume functions for the standard suspend/resume flow.

---
 drivers/net/ethernet/intel/fm10k/fm10k_pci.c | 73 ++--------------------------
 1 file changed, 3 insertions(+), 70 deletions(-)
drivers/net/ethernet/intel/fm10k/fm10k_pci.c +3 -70
index c60c3b0f89e9..b02361c42705 100644
@@ -2186,60 +2186,13 @@ static int fm10k_resume(struct pci_dev *pdev)
 	/* refresh hw_addr in case it was dropped */
 	hw->hw_addr = interface->uc_addr;
 
-	/* reset hardware to known state */
-	err = hw->mac.ops.init_hw(&interface->hw);
-	if (err) {
-		dev_err(&pdev->dev, "init_hw failed: %d\n", err);
+	err = fm10k_handle_resume(interface);
+	if (err)
 		return err;
-	}
-
-	/* reset statistics starting values */
-	hw->mac.ops.rebind_hw_stats(hw, &interface->stats);
-
-	rtnl_lock();
-
-	err = fm10k_init_queueing_scheme(interface);
-	if (err)
-		goto err_queueing_scheme;
-
-	err = fm10k_mbx_request_irq(interface);
-	if (err)
-		goto err_mbx_irq;
-
-	err = fm10k_hw_ready(interface);
-	if (err)
-		goto err_open;
-
-	err = netif_running(netdev) ? fm10k_open(netdev) : 0;
-	if (err)
-		goto err_open;
-
-	rtnl_unlock();
-
-	/* assume host is not ready, to prevent race with watchdog in case we
-	 * actually don't have connection to the switch
-	 */
-	interface->host_ready = false;
-	fm10k_watchdog_host_not_ready(interface);
-
-	/* clear the service task disable bit to allow service task to start */
-	clear_bit(__FM10K_SERVICE_DISABLE, &interface->state);
-	fm10k_service_event_schedule(interface);
-
-	/* restore SR-IOV interface */
-	fm10k_iov_resume(pdev);
 
 	netif_device_attach(netdev);
 
 	return 0;
-err_open:
-	fm10k_mbx_free_irq(interface);
-err_mbx_irq:
-	fm10k_clear_queueing_scheme(interface);
-err_queueing_scheme:
-	rtnl_unlock();
-
-	return err;
 }
 
 /**
@@ -2259,27 +2212,7 @@ static int fm10k_suspend(struct pci_dev *pdev,
 
 	netif_device_detach(netdev);
 
-	fm10k_iov_suspend(pdev);
-
-	/* the watchdog tasks may read registers, which will appear like a
-	 * surprise-remove event once the PCI device is disabled. This will
-	 * cause us to close the netdevice, so we don't retain the open/closed
-	 * state post-resume. Prevent this by disabling the service task while
-	 * suspended, until we actually resume.
-	 */
-	set_bit(__FM10K_SERVICE_DISABLE, &interface->state);
-	cancel_work_sync(&interface->service_task);
-
-	rtnl_lock();
-
-	if (netif_running(netdev))
-		fm10k_close(netdev);
-
-	fm10k_mbx_free_irq(interface);
-
-	fm10k_clear_queueing_scheme(interface);
-
-	rtnl_unlock();
+	fm10k_prepare_suspend(interface);
 
 	err = pci_save_state(pdev);
 	if (err)
Signature · 1 line
2.9.0.rc1.405.g81f467e
#24 lore raw

PATCH v2 12/18 fm10k: use common flow for suspend and resume

-----Original Message-----
From: Intel-wired-lan [mailto:intel-wired-lan-bounces at lists.osuosl.org] On Behalf Of Jacob Keller
Sent: Tuesday, June 7, 2016 4:09 PM
To: Intel Wired LAN <[email protected]>
Subject: [Intel-wired-lan] [PATCH v2 12/18] fm10k: use common flow for suspend and resume

Continuing the effort to commonize the similar suspend/resume flows, finish up by using the new fm10k_handle_suspand and fm10k_handle_resume functions for the standard suspend/resume flow.

---

#25 lore raw

PATCH v2 13/18 fm10k: enable bus master after every reset

In reply to Jacob Keller

If an FLR occurs, VF devices will be knocked out of bus master mode, and the driver will be unable to recover from the reset properly, resulting in malicious driver events and an infinite reset loop. In the normal case, the bus master mode will already be enabled and this call will essentially be a no-op. Since we're doing this every reset, it is possible we could remove the other calls to pci_set_master() but it seems not harmful to just leave them in place.

---
 drivers/net/ethernet/intel/fm10k/fm10k_pci.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/net/ethernet/intel/fm10k/fm10k_pci.c +2 -0
index b02361c42705..5e40460b041d 100644
@@ -174,6 +174,8 @@ static int fm10k_handle_reset(struct fm10k_intfc *interface)
 
 	rtnl_lock();
 
+	pci_set_master(interface->pdev);
+
 	/* reset and initialize the hardware so it is in a known state */
 	err = hw->mac.ops.reset_hw(hw);
 	if (err) {
Signature · 1 line
2.9.0.rc1.405.g81f467e
#26 lore raw

PATCH v2 13/18 fm10k: enable bus master after every reset

-----Original Message-----
From: Intel-wired-lan [mailto:intel-wired-lan-bounces at lists.osuosl.org] On Behalf Of Jacob Keller
Sent: Tuesday, June 7, 2016 4:09 PM
To: Intel Wired LAN <[email protected]>
Subject: [Intel-wired-lan] [PATCH v2 13/18] fm10k: enable bus master after every reset

If an FLR occurs, VF devices will be knocked out of bus master mode, and the driver will be unable to recover from the reset properly, resulting in malicious driver events and an infinite reset loop. In the normal case, the bus master mode will already be enabled and this call will essentially be a no-op. Since we're doing this every reset, it is possible we could remove the other calls to pci_set_master() but it seems not harmful to just leave them in place.

---

#27 lore raw

PATCH v2 14/18 fm10k: check if PCIe link is restored

In reply to Jacob Keller

Sometimes, a VF driver will lose PCIe address access, such as due to a PF FLR event. In fm10k_detach_subtask, poll and check whether the PCIe register space is active again and restore the device when it has.

---
 drivers/net/ethernet/intel/fm10k/fm10k_pci.c | 13 +++++++++++++
 1 file changed, 13 insertions(+)
drivers/net/ethernet/intel/fm10k/fm10k_pci.c +13 -0
index 5e40460b041d..d4ccb2a0c585 100644
@@ -123,11 +123,24 @@ static void fm10k_service_timer(unsigned long data)
 static void fm10k_detach_subtask(struct fm10k_intfc *interface)
 {
 	struct net_device *netdev = interface->netdev;
+	u32 __iomem *hw_addr;
+	u32 value;
 
 	/* do nothing if device is still present or hw_addr is set */
 	if (netif_device_present(netdev) || interface->hw.hw_addr)
 		return;
 
+	/* check the real address space to see if we've recovered */
+	hw_addr = READ_ONCE(interface->uc_addr);
+	value = readl(hw_addr);
+	if ((~value)) {
+		interface->hw.hw_addr = interface->uc_addr;
+		netif_device_attach(netdev);
+		interface->flags |= FM10K_FLAG_RESET_REQUESTED;
+		netdev_warn(netdev, "PCIe link restored, device now attached\n");
+		return;
+	}
+
 	rtnl_lock();
 
 	if (netif_running(netdev))
Signature · 1 line
2.9.0.rc1.405.g81f467e
#28 lore raw

PATCH v2 14/18 fm10k: check if PCIe link is restored

-----Original Message-----
From: Intel-wired-lan [mailto:intel-wired-lan-bounces at lists.osuosl.org] On Behalf Of Jacob Keller
Sent: Tuesday, June 7, 2016 4:09 PM
To: Intel Wired LAN <[email protected]>
Subject: [Intel-wired-lan] [PATCH v2 14/18] fm10k: check if PCIe link is restored

Sometimes, a VF driver will lose PCIe address access, such as due to a PF FLR event. In fm10k_detach_subtask, poll and check whether the PCIe register space is active again and restore the device when it has.

---

#29 lore raw

PATCH v2 15/18 fm10k: implement request_lport_map pointer

In reply to Jacob Keller

If the fm10k interface is brought up, but the switch manager software is not running, the driver will continuously request the lport map every few seconds in the base driver watchdog routine. Eventually after several minutes the switch mailbox Tx fifo will fill up and the mailbox will timeout, resulting in a reset. This reset will appear as if for no reason, and occurs regularly every few minutes until the switch manager software is loaded.

Prevent this from happening by only requesting the lport map after we've verified the switch mailbox is tx_ready. In order to simplify code logic and reduce code duplication, implement this as a new function pointer "mac.ops.request_lport_map" which the VF will not implement. Otherwise, we have to duplicate the tx_ready check outside of fm10k_get_host_state_generic, or re-implement most of fm10k_get_host_state_generic in the pf version.

The resulting code is simpler and easier to understand, and prevents the PF from continuously requesting lport map and filling the Tx fifo of a switch mailbox that isn't ready.

---
 drivers/net/ethernet/intel/fm10k/fm10k_common.c |  6 +++++-
 drivers/net/ethernet/intel/fm10k/fm10k_pf.c     | 15 +++------------
 drivers/net/ethernet/intel/fm10k/fm10k_type.h   |  1 +
 3 files changed, 9 insertions(+), 13 deletions(-)
drivers/net/ethernet/intel/fm10k/fm10k_common.c +5 -1
index 5bbf19cfe29b..d6baaea8bc7c 100644
@@ -519,8 +519,12 @@ s32 fm10k_get_host_state_generic(struct fm10k_hw *hw, bool *host_ready)
 		goto out;
 
 	/* interface cannot receive traffic without logical ports */
-	if (mac->dglort_map == FM10K_DGLORTMAP_NONE)
+	if (mac->dglort_map == FM10K_DGLORTMAP_NONE) {
+		if (hw->mac.ops.request_lport_map)
+			ret_val = hw->mac.ops.request_lport_map(hw);
+
 		goto out;
+	}
 
 	/* if we passed all the tests above then the switch is ready and we no
 	 * longer need to check for link
drivers/net/ethernet/intel/fm10k/fm10k_pf.c +3 -12
index 23f3566b17fc..682299dd0ce4 100644
@@ -1622,25 +1622,15 @@ static s32 fm10k_request_lport_map_pf(struct fm10k_hw *hw)
  **/
 static s32 fm10k_get_host_state_pf(struct fm10k_hw *hw, bool *switch_ready)
 {
-	s32 ret_val = 0;
 	u32 dma_ctrl2;
 
 	/* verify the switch is ready for interaction */
 	dma_ctrl2 = fm10k_read_reg(hw, FM10K_DMA_CTRL2);
 	if (!(dma_ctrl2 & FM10K_DMA_CTRL2_SWITCH_READY))
-		goto out;
+		return 0;
 
 	/* retrieve generic host state info */
-	ret_val = fm10k_get_host_state_generic(hw, switch_ready);
-	if (ret_val)
-		goto out;
-
-	/* interface cannot receive traffic without logical ports */
-	if (hw->mac.dglort_map == FM10K_DGLORTMAP_NONE)
-		ret_val = fm10k_request_lport_map_pf(hw);
-
-out:
-	return ret_val;
+	return fm10k_get_host_state_generic(hw, switch_ready);
 }
 
 /* This structure defines the attibutes to be parsed below */
@@ -1816,6 +1806,7 @@ static const struct fm10k_mac_ops mac_ops_pf = {
 	.set_dma_mask		= fm10k_set_dma_mask_pf,
 	.get_fault		= fm10k_get_fault_pf,
 	.get_host_state		= fm10k_get_host_state_pf,
+	.request_lport_map	= fm10k_request_lport_map_pf,
 };
 
 static const struct fm10k_iov_ops iov_ops_pf = {
drivers/net/ethernet/intel/fm10k/fm10k_type.h +1 -0
index 1d65ad85d72e..f4e75c498287 100644
@@ -526,6 +526,7 @@ struct fm10k_mac_ops {
 	s32 (*stop_hw)(struct fm10k_hw *);
 	s32 (*get_bus_info)(struct fm10k_hw *);
 	s32 (*get_host_state)(struct fm10k_hw *, bool *);
+	s32 (*request_lport_map)(struct fm10k_hw *);
 	s32 (*update_vlan)(struct fm10k_hw *, u32, u8, bool);
 	s32 (*read_mac_addr)(struct fm10k_hw *);
 	s32 (*update_uc_addr)(struct fm10k_hw *, u16, const u8 *,
Signature · 1 line
2.9.0.rc1.405.g81f467e
#30 lore raw

PATCH v2 15/18 fm10k: implement request_lport_map pointer

-----Original Message-----
From: Intel-wired-lan [mailto:intel-wired-lan-bounces at lists.osuosl.org] On Behalf Of Jacob Keller
Sent: Tuesday, June 7, 2016 4:09 PM
To: Intel Wired LAN <[email protected]>
Subject: [Intel-wired-lan] [PATCH v2 15/18] fm10k: implement request_lport_map pointer

If the fm10k interface is brought up, but the switch manager software is not running, the driver will continuously request the lport map every few seconds in the base driver watchdog routine. Eventually after several minutes the switch mailbox Tx fifo will fill up and the mailbox will timeout, resulting in a reset. This reset will appear as if for no reason, and occurs regularly every few minutes until the switch manager software is loaded.

Prevent this from happening by only requesting the lport map after we've verified the switch mailbox is tx_ready. In order to simplify code logic and reduce code duplication, implement this as a new function pointer "mac.ops.request_lport_map" which the VF will not implement. Otherwise, we have to duplicate the tx_ready check outside of fm10k_get_host_state_generic, or re-implement most of fm10k_get_host_state_generic in the pf version.

The resulting code is simpler and easier to understand, and prevents the PF from continuously requesting lport map and filling the Tx fifo of a switch mailbox that isn't ready.

---

#31 lore raw

PATCH v2 16/18 fm10k: force link to remain down for at least a second on resume events

In reply to Jacob Keller

When we resume from an AER recovery with many active VFs, the PF sees many spurious link up and link down events. Prevent this by delaying link down for at least one second after the resume event.

---
 drivers/net/ethernet/intel/fm10k/fm10k_pci.c | 4 ++++
 1 file changed, 4 insertions(+)
drivers/net/ethernet/intel/fm10k/fm10k_pci.c +4 -0
index d4ccb2a0c585..b8245c734c96 100644
@@ -2158,6 +2158,10 @@ static int fm10k_handle_resume(struct fm10k_intfc *interface)
 	interface->host_ready = false;
 	fm10k_watchdog_host_not_ready(interface);
 
+	/* force link to stay down for a second to prevent link flutter */
+	interface->link_down_event = jiffies + (HZ);
+	set_bit(__FM10K_LINK_DOWN, &interface->state);
+
 	/* clear the service task disable bit to allow service task to start */
 	clear_bit(__FM10K_SERVICE_DISABLE, &interface->state);
 	fm10k_service_event_schedule(interface);
Signature · 1 line
2.9.0.rc1.405.g81f467e
#32 lore raw

PATCH v2 16/18 fm10k: force link to remain down for at least a second on resume events

-----Original Message-----
From: Intel-wired-lan [mailto:intel-wired-lan-bounces at lists.osuosl.org] On Behalf Of Jacob Keller
Sent: Tuesday, June 7, 2016 4:09 PM
To: Intel Wired LAN <[email protected]>
Subject: [Intel-wired-lan] [PATCH v2 16/18] fm10k: force link to remain down for at least a second on resume events

When we resume from an AER recovery with many active VFs, the PF sees many spurious link up and link down events. Prevent this by delaying link down for at least one second after the resume event.

---

#33 lore raw

PATCH v2 17/18 fm10k: return proper error code when pci_enable_msix_range fails

In reply to Jacob Keller

The pci_enable_msix_range() function returns a positive value of the number of allocated vectors if it succeeds. On failure it returns a negative error code. Return this code properly so that the error message printed by the driver will show the actual error code instead of being masked by -ENOMEM.

---
 drivers/net/ethernet/intel/fm10k/fm10k_main.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/ethernet/intel/fm10k/fm10k_main.c +1 -1
index c85fc98945fa..f72d1cae10ff 100644
@@ -1858,7 +1858,7 @@ static int fm10k_init_msix_capability(struct fm10k_intfc *interface)
 	if (v_budget < 0) {
 		kfree(interface->msix_entries);
 		interface->msix_entries = NULL;
-		return -ENOMEM;
+		return v_budget;
 	}
 
 	/* record the number of queues available for q_vectors */
Signature · 1 line
2.9.0.rc1.405.g81f467e
#34 lore raw

PATCH v2 17/18 fm10k: return proper error code when pci_enable_msix_range fails

-----Original Message-----
From: Intel-wired-lan [mailto:intel-wired-lan-bounces at lists.osuosl.org] On Behalf Of Jacob Keller
Sent: Tuesday, June 7, 2016 4:09 PM
To: Intel Wired LAN <[email protected]>
Subject: [Intel-wired-lan] [PATCH v2 17/18] fm10k: return proper error code when pci_enable_msix_range fails

The pci_enable_msix_range() function returns a positive value of the number of allocated vectors if it succeeds. On failure it returns a negative error code. Return this code properly so that the error message printed by the driver will show the actual error code instead of being masked by -ENOMEM.

---

#35 lore raw

PATCH v2 18/18 fm10k: bump version number

In reply to Jacob Keller
---
 drivers/net/ethernet/intel/fm10k/fm10k_main.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/ethernet/intel/fm10k/fm10k_main.c +1 -1
index f72d1cae10ff..e9767b6366a8 100644
@@ -28,7 +28,7 @@
 
 #include "fm10k.h"
 
-#define DRV_VERSION	"0.19.3-k"
+#define DRV_VERSION	"0.21.2-k"
 #define DRV_SUMMARY	"Intel(R) Ethernet Switch Host Interface Driver"
 const char fm10k_driver_version[] = DRV_VERSION;
 char fm10k_driver_name[] = "fm10k";
Signature · 1 line
2.9.0.rc1.405.g81f467e