Skip to contentSkip to conversation

bnxt_en: Misc. bug fixes

PATCH In review Michael Chan 2 participants 10 messages Apr 28, 2025 – May 1, 2025

Subsystems (3)
lore
Share & tools
lore

Michael Chan

author maintainer
#0 lore raw

cover bnxt_en: Misc. bug fixes

This series fixes a bug in the driver initialization path, MSIX setup sequencing issue in the FW error and AER paths, a missing skb_mark_for_recycle() in the VLAN error path, some ethtool coredump fixes, an ethtool selftest fix, and an ethtool register dump byte order fix.

Kalesh AP (1):
bnxt_en: Fix ethtool selftest output in one of the failure cases

Kashyap Desai (2):
bnxt_en: call pci_alloc_irq_vectors() after bnxt_reserve_rings()
bnxt_en: delay pci_alloc_irq_vectors() in the AER path

Michael Chan (1):
bnxt_en: Fix ethtool -d byte order for 32-bit values

Shravya KN (1):
bnxt_en: Fix error handling path in bnxt_init_chip()

Shruti Parab (2):
bnxt_en: Fix coredump logic to free allocated buffer
bnxt_en: Fix out-of-bound memcpy() during ethtool -w

Somnath Kotur (1):
bnxt_en: Add missing skb_mark_for_recycle() in bnxt_rx_vlan()

 drivers/net/ethernet/broadcom/bnxt/bnxt.c     | 28 +++++++------
 drivers/net/ethernet/broadcom/bnxt/bnxt.h     |  1 +
 .../ethernet/broadcom/bnxt/bnxt_coredump.c    | 20 +++++++---
 .../net/ethernet/broadcom/bnxt/bnxt_ethtool.c | 40 +++++++++++++++----
 4 files changed, 65 insertions(+), 24 deletions(-)
Signature · 1 line
2.30.1
author maintainer
#1 lore raw

PATCH 1/8 bnxt_en: Fix error handling path in bnxt_init_chip()

From: Shravya KN <[email protected]>

WARN_ON() is triggered in __flush_work() if bnxt_init_chip() fails because we call cancel_work_sync() on dim work that has not been initialized.

WARNING: CPU: 37 PID: 5223 at kernel/workqueue.c:4201 __flush_work.isra.0+0x212/0x230

The driver relies on the BNXT_STATE_NAPI_DISABLED bit to check if dim work has already been cancelled. But in the bnxt_open() path, BNXT_STATE_NAPI_DISABLED is not set and this causes the error path to think that it needs to cancel the uninitalized dim work. Fix it by setting BNXT_STATE_NAPI_DISABLED during initialization. The bit will be cleared when we enable NAPI and initialize dim work.

Fixes 40452969a506 ("bnxt_en: Fix DIM shutdown")
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 3 +++
 1 file changed, 3 insertions(+)
drivers/net/ethernet/broadcom/bnxt/bnxt.c +3 -0
index 2c8e2c19d854..c4bccc683597 100644
@@ -11602,6 +11602,9 @@ static void bnxt_init_napi(struct bnxt *bp)
 		poll_fn = bnxt_poll_p5;
 	else if (BNXT_CHIP_TYPE_NITRO_A0(bp))
 		cp_nr_rings--;
+
+	set_bit(BNXT_STATE_NAPI_DISABLED, &bp->state);
+
 	for (i = 0; i < cp_nr_rings; i++) {
 		bnapi = bp->bnapi[i];
 		netif_napi_add_config_locked(bp->dev, &bnapi->napi, poll_fn,
Signature · 1 line
2.30.1
author maintainer
#2 lore raw

PATCH 2/8 bnxt_en: Fix ethtool selftest output in one of the failure cases

In reply to Michael Chan

From: Kalesh AP <[email protected]>

When RDMA driver is loaded, running offline self test is not supported and driver returns failure early. But it is not clearing the input buffer and hence the application prints some junk characters for individual test results.

Fix it by clearing the buffer before returning.

Fixes 895621f1c816 ("bnxt_en: Don't support offline self test when RoCE driver is loaded")
---
 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 48dd5922e4dd..7be37976f3e4 100644
@@ -4991,6 +4991,7 @@ static void bnxt_self_test(struct net_device *dev, struct ethtool_test *etest,
 	if (!bp->num_tests || !BNXT_PF(bp))
 		return;
 
+	memset(buf, 0, sizeof(u64) * bp->num_tests);
 	if (etest->flags & ETH_TEST_FL_OFFLINE &&
 	    bnxt_ulp_registered(bp->edev)) {
 		etest->flags |= ETH_TEST_FL_FAILED;
@@ -4998,7 +4999,6 @@ static void bnxt_self_test(struct net_device *dev, struct ethtool_test *etest,
 		return;
 	}
 
-	memset(buf, 0, sizeof(u64) * bp->num_tests);
 	if (!netif_running(dev)) {
 		etest->flags |= ETH_TEST_FL_FAILED;
 		return;
Signature · 1 line
2.30.1
author maintainer
#3 lore raw

PATCH 3/8 bnxt_en: Add missing skb_mark_for_recycle() in bnxt_rx_vlan()

In reply to Michael Chan

From: Somnath Kotur <[email protected]>

If bnxt_rx_vlan() fails because the VLAN protocol ID is invalid, the SKB is freed but we're missing the call to recycle it. This may cause the warning:

"page_pool_release_retry() stalled pool shutdown"

Add the missing skb_mark_for_recycle() in bnxt_rx_vlan().

Fixes 86b05508f775 ("bnxt_en: Use the unified RX page pool buffers for XDP and non-XDP")
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 1 +
 1 file changed, 1 insertion(+)
drivers/net/ethernet/broadcom/bnxt/bnxt.c +1 -0
index c4bccc683597..cfc9ccab39bf 100644
@@ -2015,6 +2015,7 @@ static struct sk_buff *bnxt_rx_vlan(struct sk_buff *skb, u8 cmp_type,
 	}
 	return skb;
 vlan_err:
+	skb_mark_for_recycle(skb);
 	dev_kfree_skb(skb);
 	return NULL;
 }
Signature · 1 line
2.30.1
author maintainer
#4 lore raw

PATCH 4/8 bnxt_en: call pci_alloc_irq_vectors() after bnxt_reserve_rings()

In reply to Michael Chan

From: Kashyap Desai <[email protected]>

On some architectures (e.g. ARM), calling pci_alloc_irq_vectors() will immediately cause the MSIX table to be written. This will not work if we haven't called bnxt_reserve_rings() to properly map the MSIX table to the MSIX vectors reserved by FW.

Fix the FW error recovery path to delay the bnxt_init_int_mode() -> pci_alloc_irq_vectors() call by removing it from bnxt_hwrm_if_change(). bnxt_request_irq() later in the code path will call it and by then the MSIX table is properly mapped.

Fixes 4343838ca5eb ("bnxt_en: Replace deprecated PCI MSIX APIs")
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 7 +------
 1 file changed, 1 insertion(+), 6 deletions(-)
drivers/net/ethernet/broadcom/bnxt/bnxt.c +1 -6
index cfc9ccab39bf..3b2ea36f25a2 100644
@@ -12399,13 +12399,8 @@ static int bnxt_hwrm_if_change(struct bnxt *bp, bool up)
 				set_bit(BNXT_STATE_ABORT_ERR, &bp->state);
 				return rc;
 			}
+			/* IRQ will be initialized later in bnxt_request_irq()*/
 			bnxt_clear_int_mode(bp);
-			rc = bnxt_init_int_mode(bp);
-			if (rc) {
-				clear_bit(BNXT_STATE_FW_RESET_DET, &bp->state);
-				netdev_err(bp->dev, "init int mode failed\n");
-				return rc;
-			}
 		}
 		rc = bnxt_cancel_reservations(bp, fw_reset);
 	}
Signature · 1 line
2.30.1
author maintainer
#5 lore raw

PATCH 5/8 bnxt_en: delay pci_alloc_irq_vectors() in the AER path

In reply to Michael Chan

From: Kashyap Desai <[email protected]>

This patch is similar to the last patch to delay the pci_alloc_irq_vectors() call in the AER path until after calling bnxt_reserve_rings(). bnxt_reserve_rings() needs to properly map the MSIX table first before we call pci_alloc_irq_vectors() which may immediately write to the MSIX table in some architectures.

Move the bnxt_init_int_mode() call from bnxt_io_slot_reset() to bnxt_io_resume() after calling bnxt_reserve_rings().

With this change, the AER path may call bnxt_open() -> bnxt_hwrm_if_change() with bp->irq_tbl set to NULL. bp->irq_tbl is cleared when we call bnxt_clear_int_mode() in bnxt_io_slot_reset(). So we cannot use !bp->irq_tbl to detect aborted FW reset. Add a new BNXT_FW_RESET_STATE_ABORT to detect aborted FW reset in bnxt_hwrm_if_change().

---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 17 +++++++++++------
 drivers/net/ethernet/broadcom/bnxt/bnxt.h |  1 +
 2 files changed, 12 insertions(+), 6 deletions(-)
drivers/net/ethernet/broadcom/bnxt/bnxt.c +11 -6
index 3b2ea36f25a2..78e496b0ec26 100644
@@ -12325,12 +12325,15 @@ static int bnxt_hwrm_if_change(struct bnxt *bp, bool up)
 {
 	struct hwrm_func_drv_if_change_output *resp;
 	struct hwrm_func_drv_if_change_input *req;
-	bool fw_reset = !bp->irq_tbl;
 	bool resc_reinit = false;
 	bool caps_change = false;
 	int rc, retry = 0;
+	bool fw_reset;
 	u32 flags = 0;
 
+	fw_reset = (bp->fw_reset_state == BNXT_FW_RESET_STATE_ABORT);
+	bp->fw_reset_state = 0;
+
 	if (!(bp->fw_cap & BNXT_FW_CAP_IF_CHANGE))
 		return 0;
 
@@ -14833,7 +14836,7 @@ static void bnxt_fw_reset_abort(struct bnxt *bp, int rc)
 	clear_bit(BNXT_STATE_IN_FW_RESET, &bp->state);
 	if (bp->fw_reset_state != BNXT_FW_RESET_STATE_POLL_VF)
 		bnxt_dl_health_fw_status_update(bp, false);
-	bp->fw_reset_state = 0;
+	bp->fw_reset_state = BNXT_FW_RESET_STATE_ABORT;
 	netif_close(bp->dev);
 }
 
@@ -16931,10 +16934,9 @@ static pci_ers_result_t bnxt_io_slot_reset(struct pci_dev *pdev)
 		if (!err)
 			result = PCI_ERS_RESULT_RECOVERED;
 
+		/* IRQ will be initialized later in bnxt_io_resume */
 		bnxt_ulp_irq_stop(bp);
 		bnxt_clear_int_mode(bp);
-		err = bnxt_init_int_mode(bp);
-		bnxt_ulp_irq_restart(bp, err);
 	}
 
 reset_exit:
@@ -16963,10 +16965,13 @@ static void bnxt_io_resume(struct pci_dev *pdev)
 
 	err = bnxt_hwrm_func_qcaps(bp);
 	if (!err) {
-		if (netif_running(netdev))
+		if (netif_running(netdev)) {
 			err = bnxt_open(netdev);
-		else
+		} else {
 			err = bnxt_reserve_rings(bp, true);
+			if (!err)
+				err = bnxt_init_int_mode(bp);
+		}
 	}
 
 	if (!err)
drivers/net/ethernet/broadcom/bnxt/bnxt.h +1 -0
index 21726cf56586..bc8b3b7e915d 100644
@@ -2614,6 +2614,7 @@ struct bnxt {
 #define BNXT_FW_RESET_STATE_POLL_FW	4
 #define BNXT_FW_RESET_STATE_OPENING	5
 #define BNXT_FW_RESET_STATE_POLL_FW_DOWN	6
+#define BNXT_FW_RESET_STATE_ABORT	7
 
 	u16			fw_reset_min_dsecs;
 #define BNXT_DFLT_FW_RST_MIN_DSECS	20
Signature · 1 line
2.30.1
author maintainer
#6 lore raw

PATCH 6/8 bnxt_en: Fix coredump logic to free allocated buffer

In reply to Michael Chan

From: Shruti Parab <[email protected]>

When handling HWRM_DBG_COREDUMP_LIST FW command in bnxt_hwrm_dbg_dma_data(), the allocated buffer info->dest_buf is not freed in the error path. In the normal path, info->dest_buf is assigned to coredump->data and it will eventually be freed after the coredump is collected.

Free info->dest_buf immediately inside bnxt_hwrm_dbg_dma_data() in the error path.

Fixes c74751f4c392 ("bnxt_en: Return error if FW returns more data than dump length")
---
 drivers/net/ethernet/broadcom/bnxt/bnxt_coredump.c | 5 +++++
 1 file changed, 5 insertions(+)
drivers/net/ethernet/broadcom/bnxt/bnxt_coredump.c +5 -0
index 5576e7cf8463..9a74793648f6 100644
@@ -116,6 +116,11 @@ static int bnxt_hwrm_dbg_dma_data(struct bnxt *bp, void *msg,
 				memcpy(info->dest_buf + off, dma_buf, len);
 			} else {
 				rc = -ENOBUFS;
+				if (cmn_req->req_type ==
+				    cpu_to_le16(HWRM_DBG_COREDUMP_LIST)) {
+					kfree(info->dest_buf);
+					info->dest_buf = NULL;
+				}
 				break;
 			}
 		}
Signature · 1 line
2.30.1
author maintainer
#7 lore raw

PATCH 7/8 bnxt_en: Fix out-of-bound memcpy() during ethtool -w

In reply to Michael Chan

From: Shruti Parab <[email protected]>

When retrieving the FW coredump using ethtool, it can sometimes cause memory corruption:

BUG: KFENCE: memory corruption in __bnxt_get_coredump+0x3ef/0x670 [bnxt_en]
Corrupted memory at 0x000000008f0f30e8 [ ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ] (in kfence-#45):
__bnxt_get_coredump+0x3ef/0x670 [bnxt_en]
ethtool_get_dump_data+0xdc/0x1a0
__dev_ethtool+0xa1e/0x1af0
dev_ethtool+0xa8/0x170
dev_ioctl+0x1b5/0x580
sock_do_ioctl+0xab/0xf0
sock_ioctl+0x1ce/0x2e0
__x64_sys_ioctl+0x87/0xc0
do_syscall_64+0x5c/0xf0
entry_SYSCALL_64_after_hwframe+0x78/0x80

...

This happens when copying the coredump segment list in bnxt_hwrm_dbg_dma_data() with the HWRM_DBG_COREDUMP_LIST FW command. The info->dest_buf buffer is allocated based on the number of coredump segments returned by the FW. The segment list is then DMA'ed by the FW and the length of the DMA is returned by FW. The driver then copies this DMA'ed segment list to info->dest_buf.

In some cases, this DMA length may exceed the info->dest_buf length and cause the above BUG condition. Fix it by capping the copy length to not exceed the length of info->dest_buf. The extra DMA data contains no useful information.

This code path is shared for the HWRM_DBG_COREDUMP_LIST and the HWRM_DBG_COREDUMP_RETRIEVE FW commands. The buffering is different for these 2 FW commands. To simplify the logic, we need to move the line to adjust the buffer length for HWRM_DBG_COREDUMP_RETRIEVE up, so that the new check to cap the copy length will work for both commands.

Fixes c74751f4c392 ("bnxt_en: Return error if FW returns more data than dump length")
---
 .../net/ethernet/broadcom/bnxt/bnxt_coredump.c    | 15 ++++++++++-----
 1 file changed, 10 insertions(+), 5 deletions(-)
drivers/net/ethernet/broadcom/bnxt/bnxt_coredump.c +10 -5
index 9a74793648f6..a000d3f630bd 100644
@@ -110,10 +110,19 @@ static int bnxt_hwrm_dbg_dma_data(struct bnxt *bp, void *msg,
 			}
 		}
 
+		if (cmn_req->req_type ==
+				cpu_to_le16(HWRM_DBG_COREDUMP_RETRIEVE))
+			info->dest_buf_size += len;
+
 		if (info->dest_buf) {
 			if ((info->seg_start + off + len) <=
 			    BNXT_COREDUMP_BUF_LEN(info->buf_len)) {
-				memcpy(info->dest_buf + off, dma_buf, len);
+				u16 copylen = min_t(u16, len,
+						    info->dest_buf_size - off);
+
+				memcpy(info->dest_buf + off, dma_buf, copylen);
+				if (copylen < len)
+					break;
 			} else {
 				rc = -ENOBUFS;
 				if (cmn_req->req_type ==
@@ -125,10 +134,6 @@ static int bnxt_hwrm_dbg_dma_data(struct bnxt *bp, void *msg,
 			}
 		}
 
-		if (cmn_req->req_type ==
-				cpu_to_le16(HWRM_DBG_COREDUMP_RETRIEVE))
-			info->dest_buf_size += len;
-
 		if (!(cmn_resp->flags & HWRM_DBG_CMN_FLAGS_MORE))
 			break;
 
Signature · 1 line
2.30.1
author maintainer
#8 lore raw

PATCH 8/8 bnxt_en: Fix ethtool -d byte order for 32-bit values

In reply to Michael Chan

For version 1 register dump that includes the PCIe stats, the existing code incorrectly assumes that all PCIe stats are 64-bit values. Fix it by using an array containing the starting and ending index of the 32-bit values. The loop in bnxt_get_regs() will use the array to do proper endian swap for the 32-bit values.

Fixes b5d600b027eb ("bnxt_en: Add support for 'ethtool -d'")
---
 .../net/ethernet/broadcom/bnxt/bnxt_ethtool.c | 38 ++++++++++++++++---
 1 file changed, 32 insertions(+), 6 deletions(-)
drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c +32 -6
index 7be37976f3e4..f5d490bf997e 100644
@@ -2062,6 +2062,17 @@ static int bnxt_get_regs_len(struct net_device *dev)
 	return reg_len;
 }
 
+#define BNXT_PCIE_32B_ENTRY(start, end)			\
+	 { offsetof(struct pcie_ctx_hw_stats, start),	\
+	   offsetof(struct pcie_ctx_hw_stats, end) }
+
+static const struct {
+	u16 start;
+	u16 end;
+} bnxt_pcie_32b_entries[] = {
+	BNXT_PCIE_32B_ENTRY(pcie_ltssm_histogram[0], pcie_ltssm_histogram[3]),
+};
+
 static void bnxt_get_regs(struct net_device *dev, struct ethtool_regs *regs,
 			  void *_p)
 {
@@ -2094,12 +2105,27 @@ static void bnxt_get_regs(struct net_device *dev, struct ethtool_regs *regs,
 	req->pcie_stat_host_addr = cpu_to_le64(hw_pcie_stats_addr);
 	rc = hwrm_req_send(bp, req);
 	if (!rc) {
-		__le64 *src = (__le64 *)hw_pcie_stats;
-		u64 *dst = (u64 *)(_p + BNXT_PXP_REG_LEN);
-		int i;
-
-		for (i = 0; i < sizeof(*hw_pcie_stats) / sizeof(__le64); i++)
-			dst[i] = le64_to_cpu(src[i]);
+		u8 *dst = (u8 *)(_p + BNXT_PXP_REG_LEN);
+		u8 *src = (u8 *)hw_pcie_stats;
+		int i, j;
+
+		for (i = 0, j = 0; i < sizeof(*hw_pcie_stats); ) {
+			if (i >= bnxt_pcie_32b_entries[j].start &&
+			    i <= bnxt_pcie_32b_entries[j].end) {
+				u32 *dst32 = (u32 *)(dst + i);
+
+				*dst32 = le32_to_cpu(*(__le32 *)(src + i));
+				i += 4;
+				if (i > bnxt_pcie_32b_entries[j].end &&
+				    j < ARRAY_SIZE(bnxt_pcie_32b_entries) - 1)
+					j++;
+			} else {
+				u64 *dst64 = (u64 *)(dst + i);
+
+				*dst64 = le64_to_cpu(*(__le64 *)(src + i));
+				i += 8;
+			}
+		}
 	}
 	hwrm_req_drop(bp, req);
 }
Signature · 1 line
2.30.1
maintainer
#9 lore raw
In reply to Michael Chan
Quoted 5 lines · On Mon, 28 Apr 2025 15:58:55 -0700 Michael Chan wrote:

This series fixes a bug in the driver initialization path, MSIX setup sequencing issue in the FW error and AER paths, a missing skb_mark_for_recycle() in the VLAN error path, some ethtool coredump fixes, an ethtool selftest fix, and an ethtool register dump byte order fix.

Applied by Dave last night, thanks!