Skip to contentSkip to conversation

bnxt_en: Bug fixes

PATCH Applied Michael Chan 2 participants 6 messages Nov 3, 2022 – Nov 5, 2022

Subsystems (3)
lore
Share & tools
lore

Michael Chan

author maintainer
#0 lore raw

cover bnxt_en: Bug fixes

This bug fix series includes fixes for PCIE AER, a crash that may occur when doing ethtool -C in the middle of error recovery, and aRFS.

Alex Barba (1):
bnxt_en: fix potentially incorrect return value for ndo_rx_flow_steer

Michael Chan (1):
bnxt_en: Fix possible crash in bnxt_hwrm_set_coal()

Vikas Gupta (2):
bnxt_en: refactor bnxt_cancel_reservations()
bnxt_en: fix the handling of PCIE-AER

 drivers/net/ethernet/broadcom/bnxt/bnxt.c     | 54 +++++++++++++++----
 drivers/net/ethernet/broadcom/bnxt/bnxt.h     |  1 +
 .../net/ethernet/broadcom/bnxt/bnxt_ethtool.c |  2 +-
 .../net/ethernet/broadcom/bnxt/bnxt_hwrm.c    |  3 +-
 4 files changed, 49 insertions(+), 11 deletions(-)
Signature · 1 line
2.18.1
author maintainer
#1 lore raw

PATCH 1/4 bnxt_en: refactor bnxt_cancel_reservations()

From: Vikas Gupta <[email protected]>

Introduce bnxt_clear_reservations() to clear the reserved attributes only. This will be used in the next patch to fix PCI AER handling.

---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 23 ++++++++++++++++-------
 1 file changed, 16 insertions(+), 7 deletions(-)
drivers/net/ethernet/broadcom/bnxt/bnxt.c +16 -7
index 04cf7684f1b0..3743d9755eb4 100644
@@ -9983,17 +9983,12 @@ static int bnxt_try_recover_fw(struct bnxt *bp)
 	return -ENODEV;
 }
 
-int bnxt_cancel_reservations(struct bnxt *bp, bool fw_reset)
+static void bnxt_clear_reservations(struct bnxt *bp, bool fw_reset)
 {
 	struct bnxt_hw_resc *hw_resc = &bp->hw_resc;
-	int rc;
 
 	if (!BNXT_NEW_RM(bp))
-		return 0; /* no resource reservations required */
-
-	rc = bnxt_hwrm_func_resc_qcaps(bp, true);
-	if (rc)
-		netdev_err(bp->dev, "resc_qcaps failed\n");
+		return; /* no resource reservations required */
 
 	hw_resc->resv_cp_rings = 0;
 	hw_resc->resv_stat_ctxs = 0;
@@ -10006,6 +10001,20 @@ int bnxt_cancel_reservations(struct bnxt *bp, bool fw_reset)
 		bp->tx_nr_rings = 0;
 		bp->rx_nr_rings = 0;
 	}
+}
+
+int bnxt_cancel_reservations(struct bnxt *bp, bool fw_reset)
+{
+	int rc;
+
+	if (!BNXT_NEW_RM(bp))
+		return 0; /* no resource reservations required */
+
+	rc = bnxt_hwrm_func_resc_qcaps(bp, true);
+	if (rc)
+		netdev_err(bp->dev, "resc_qcaps failed\n");
+
+	bnxt_clear_reservations(bp, fw_reset);
 
 	return rc;
 }
Signature · 1 line
2.18.1
author maintainer
#2 lore raw

PATCH 2/4 bnxt_en: fix the handling of PCIE-AER

In reply to Michael Chan

From: Vikas Gupta <[email protected]>

Fix the sequence required for PCIE-AER. While slot reset occurs, firmware might not be ready and the driver needs to check for its recovery. We also need to remap the health registers for some chips and clear the resource reservations. The resources will be allocated again during bnxt_io_resume().

Fixes fb1e6e562b37 ("bnxt_en: Fix AER recovery.")
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c     | 29 ++++++++++++++++++-
 drivers/net/ethernet/broadcom/bnxt/bnxt.h     |  1 +
 .../net/ethernet/broadcom/bnxt/bnxt_hwrm.c    |  3 +-
 3 files changed, 31 insertions(+), 2 deletions(-)
drivers/net/ethernet/broadcom/bnxt/bnxt.c +28 -1
index 3743d9755eb4..f44f9367b7fd 100644
@@ -13922,7 +13922,9 @@ static pci_ers_result_t bnxt_io_slot_reset(struct pci_dev *pdev)
 	pci_ers_result_t result = PCI_ERS_RESULT_DISCONNECT;
 	struct net_device *netdev = pci_get_drvdata(pdev);
 	struct bnxt *bp = netdev_priv(netdev);
-	int err = 0, off;
+	int retry = 0;
+	int err = 0;
+	int off;
 
 	netdev_info(bp->dev, "PCI Slot Reset\n");
 
@@ -13950,11 +13952,36 @@ static pci_ers_result_t bnxt_io_slot_reset(struct pci_dev *pdev)
 		pci_restore_state(pdev);
 		pci_save_state(pdev);
 
+		bnxt_inv_fw_health_reg(bp);
+		bnxt_try_map_fw_health_reg(bp);
+
+		/* In some PCIe AER scenarios, firmware may take up to
+		 * 10 seconds to become ready in the worst case.
+		 */
+		do {
+			err = bnxt_try_recover_fw(bp);
+			if (!err)
+				break;
+			retry++;
+		} while (retry < BNXT_FW_SLOT_RESET_RETRY);
+
+		if (err) {
+			dev_err(&pdev->dev, "Firmware not ready\n");
+			goto reset_exit;
+		}
+
 		err = bnxt_hwrm_func_reset(bp);
 		if (!err)
 			result = PCI_ERS_RESULT_RECOVERED;
+
+		bnxt_ulp_irq_stop(bp);
+		bnxt_clear_int_mode(bp);
+		err = bnxt_init_int_mode(bp);
+		bnxt_ulp_irq_restart(bp, err);
 	}
 
+reset_exit:
+	bnxt_clear_reservations(bp, true);
 	rtnl_unlock();
 
 	return result;
drivers/net/ethernet/broadcom/bnxt/bnxt.h +1 -0
index b1b17f911300..d5fa43cfe524 100644
@@ -1621,6 +1621,7 @@ struct bnxt_fw_health {
 
 #define BNXT_FW_RETRY			5
 #define BNXT_FW_IF_RETRY		10
+#define BNXT_FW_SLOT_RESET_RETRY	4
 
 enum board_idx {
 	BCM57301,
drivers/net/ethernet/broadcom/bnxt/bnxt_hwrm.c +2 -1
index b01d42928a53..132442f16fe6 100644
@@ -476,7 +476,8 @@ static int __hwrm_send(struct bnxt *bp, struct bnxt_hwrm_ctx *ctx)
 		memset(ctx->resp, 0, PAGE_SIZE);
 
 	req_type = le16_to_cpu(ctx->req->req_type);
-	if (BNXT_NO_FW_ACCESS(bp) && req_type != HWRM_FUNC_RESET) {
+	if (BNXT_NO_FW_ACCESS(bp) &&
+	    (req_type != HWRM_FUNC_RESET && req_type != HWRM_VER_GET)) {
 		netdev_dbg(bp->dev, "hwrm req_type 0x%x skipped, FW channel down\n",
 			   req_type);
 		goto exit;
Signature · 1 line
2.18.1
author maintainer
#3 lore raw

PATCH 3/4 bnxt_en: Fix possible crash in bnxt_hwrm_set_coal()

In reply to Michael Chan

During the error recovery sequence, the rtnl_lock is not held for the entire duration and some datastructures may be freed during the sequence. Check for the BNXT_STATE_OPEN flag instead of netif_running() to ensure that the device is fully operational before proceeding to reconfigure the coalescing settings.

This will fix a possible crash like this:

BUG: unable to handle kernel NULL pointer dereference at 0000000000000000
PGD 0 P4D 0
Oops: 0000 [#1] SMP NOPTI
CPU: 10 PID: 181276 Comm: ethtool Kdump: loaded Tainted: G          IOE    --------- -  - 4.18.0-348.el8.x86_64 #1
Hardware name: Dell Inc. PowerEdge R740/0F9N89, BIOS 2.3.10 08/15/2019
RIP: 0010:bnxt_hwrm_set_coal+0x1fb/0x2a0 [bnxt_en]
Code: c2 66 83 4e 22 08 66 89 46 1c e8 10 cb 00 00 41 83 c6 01 44 39 b3 68 01 00 00 0f 8e a3 00 00 00 48 8b 93 c8 00 00 00 49 63 c6 <48> 8b 2c c2 48 8b 85 b8 02 00 00 48 85 c0 74 2e 48 8b 74 24 08 f6
RSP: 0018:ffffb11c8dcaba50 EFLAGS: 00010246
RAX: 0000000000000000 RBX: ffff8d168a8b0ac0 RCX: 00000000000000c5
RDX: 0000000000000000 RSI: ffff8d162f72c000 RDI: ffff8d168a8b0b28
RBP: 0000000000000000 R08: b6e1f68a12e9a7eb R09: 0000000000000000
R10: 0000000000000001 R11: 0000000000000037 R12: ffff8d168a8b109c
R13: ffff8d168a8b10aa R14: 0000000000000000 R15: ffffffffc01ac4e0
FS:  00007f3852e4c740(0000) GS:ffff8d24c0080000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000000000000 CR3: 000000041b3ee003 CR4: 00000000007706e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
PKRU: 55555554
Call Trace:
 ethnl_set_coalesce+0x3ce/0x4c0
 genl_family_rcv_msg_doit.isra.15+0x10f/0x150
 genl_family_rcv_msg+0xb3/0x160
 ? coalesce_fill_reply+0x480/0x480
 genl_rcv_msg+0x47/0x90
 ? genl_family_rcv_msg+0x160/0x160
 netlink_rcv_skb+0x4c/0x120
 genl_rcv+0x24/0x40
 netlink_unicast+0x196/0x230
 netlink_sendmsg+0x204/0x3d0
 sock_sendmsg+0x4c/0x50
 __sys_sendto+0xee/0x160
 ? syscall_trace_enter+0x1d3/0x2c0
 ? __audit_syscall_exit+0x249/0x2a0
 __x64_sys_sendto+0x24/0x30
 do_syscall_64+0x5b/0x1a0
 entry_SYSCALL_64_after_hwframe+0x65/0xca
RIP: 0033:0x7f38524163bb
Fixes 2151fe0830fd ("bnxt_en: Handle RESET_NOTIFY async event from firmware.")
---
 drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c +1 -1
index f57e524c7e30..8cad15c458b3 100644
@@ -162,7 +162,7 @@ static int bnxt_set_coalesce(struct net_device *dev,
 	}
 
 reset_coalesce:
-	if (netif_running(dev)) {
+	if (test_bit(BNXT_STATE_OPEN, &bp->state)) {
 		if (update_stats) {
 			rc = bnxt_close_nic(bp, true, false);
 			if (!rc)
Signature · 1 line
2.18.1
author maintainer
#4 lore raw

PATCH 4/4 bnxt_en: fix potentially incorrect return value for ndo_rx_flow_steer

In reply to Michael Chan

From: Alex Barba <[email protected]>

In the bnxt_en driver ndo_rx_flow_steer returns '0' whenever an entry that we are attempting to steer is already found. This is not the correct behavior. The return code should be the value/index that corresponds to the entry. Returning zero all the time causes the RFS records to be incorrect unless entry '0' is the correct one. As flows migrate to different cores this can create entries that are not correct.

Fixes c0c050c58d84 ("bnxt_en: New Broadcom ethernet driver.")
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/ethernet/broadcom/bnxt/bnxt.c +1 -1
index f44f9367b7fd..c78b6e9dea2c 100644
@@ -12903,8 +12903,8 @@ static int bnxt_rx_flow_steer(struct net_device *dev, const struct sk_buff *skb,
 	rcu_read_lock();
 	hlist_for_each_entry_rcu(fltr, head, hash) {
 		if (bnxt_fltr_match(fltr, new_fltr)) {
+			rc = fltr->sw_id;
 			rcu_read_unlock();
-			rc = 0;
 			goto err_free;
 		}
 	}
Signature · 1 line
2.18.1
#5 lore raw
In reply to Michael Chan
This series was applied to netdev/net.git (master) by Jakub Kicinski <[email protected]>:

Hello:

This series was applied to netdev/net.git (master) by Jakub Kicinski <[email protected]>:

Quoted 10 lines · On Thu, 3 Nov 2022 19:33:23 -0400 you wrote:

This bug fix series includes fixes for PCIE AER, a crash that may occur when doing ethtool -C in the middle of error recovery, and aRFS.

Alex Barba (1):
bnxt_en: fix potentially incorrect return value for ndo_rx_flow_steer

Michael Chan (1):
bnxt_en: Fix possible crash in bnxt_hwrm_set_coal()

[...]

Here is the summary with links:
- [net,1/4] bnxt_en: refactor bnxt_cancel_reservations()

    https://git.kernel.org/netdev/net/c/b4c66425771d
  - [net,2/4] bnxt_en: fix the handling of PCIE-AER
    https://git.kernel.org/netdev/net/c/0cf736a18a1e
  - [net,3/4] bnxt_en: Fix possible crash in bnxt_hwrm_set_coal()
    https://git.kernel.org/netdev/net/c/6d81ea3765df
  - [net,4/4] bnxt_en: fix potentially incorrect return value for ndo_rx_flow_steer
    https://git.kernel.org/netdev/net/c/02597d39145b

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html