Skip to contentSkip to conversation

octeon_ep: reset firmware ready status

PATCH Applied Vimlesh Kumar 3 participants 3 messages Jan 15 – Jan 19

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH v4 1/1 octeon_ep: reset firmware ready status

Add support to reset firmware ready status when the driver is removed(either in unload or unbind)

---
V4: Use static inline function instead of macro for readability.

V3:
- Reformat code to less than 80 columns wide.
- Use #defines for register constants.

V2: Use recommended bit manipulation macros.

V1: https://lore.kernel.org/all/[email protected]/

 .../marvell/octeon_ep/octep_cn9k_pf.c         | 26 ++++++++++++++++
 .../marvell/octeon_ep/octep_cnxk_pf.c         |  2 +-
 .../marvell/octeon_ep/octep_regs_cn9k_pf.h    | 30 +++++++++++++++++++
 .../marvell/octeon_ep/octep_regs_cnxk_pf.h    |  1 +
 4 files changed, 58 insertions(+), 1 deletion(-)
drivers/net/ethernet/marvell/octeon_ep/octep_cn9k_pf.c +26 -0
index b5805969404f..686a3259ccb8 100644
@@ -637,6 +637,19 @@ static int octep_soft_reset_cn93_pf(struct octep_device *oct)
 
 	octep_write_csr64(oct, CN93_SDP_WIN_WR_MASK_REG, 0xFF);
 
+	/* Firmware status CSR is supposed to be cleared by
+	 * core domain reset, but due to a hw bug, it is not.
+	 * Set it to RUNNING right before reset so that it is not
+	 * left in READY (1) state after a reset.  This is required
+	 * in addition to the early setting to handle the case where
+	 * the OcteonTX is unexpectedly reset, reboots, and then
+	 * the module is removed.
+	 */
+	OCTEP_PCI_WIN_WRITE(oct,
+			    CN9K_PEMX_PFX_CSX_PFCFGX(0,
+						     0, CN9K_PCIEEP_VSECST_CTL),
+			    FW_STATUS_DOWNING);
+
 	/* Set core domain reset bit */
 	OCTEP_PCI_WIN_WRITE(oct, CN93_RST_CORE_DOMAIN_W1S, 1);
 	/* Wait for 100ms as Octeon resets. */
@@ -894,4 +907,17 @@ void octep_device_setup_cn93_pf(struct octep_device *oct)
 
 	octep_init_config_cn93_pf(oct);
 	octep_configure_ring_mapping_cn93_pf(oct);
+
+	if (oct->chip_id == OCTEP_PCI_DEVICE_ID_CN98_PF)
+		return;
+
+	/* Firmware status CSR is supposed to be cleared by
+	 * core domain reset, but due to IPBUPEM-38842, it is not.
+	 * Set it to RUNNING early in boot, so that unexpected resets
+	 * leave it in a state that is not READY (1).
+	 */
+	OCTEP_PCI_WIN_WRITE(oct,
+			    CN9K_PEMX_PFX_CSX_PFCFGX(0,
+						     0, CN9K_PCIEEP_VSECST_CTL),
+			    FW_STATUS_RUNNING);
 }
drivers/net/ethernet/marvell/octeon_ep/octep_cnxk_pf.c +1 -1
index 5de0b5ecbc5f..e07264b3dbf8 100644
@@ -660,7 +660,7 @@ static int octep_soft_reset_cnxk_pf(struct octep_device *oct)
 	 * the module is removed.
 	 */
 	OCTEP_PCI_WIN_WRITE(oct, CNXK_PEMX_PFX_CSX_PFCFGX(0, 0, CNXK_PCIEEP_VSECST_CTL),
-			    FW_STATUS_RUNNING);
+			    FW_STATUS_DOWNING);
 
 	/* Set chip domain reset bit */
 	OCTEP_PCI_WIN_WRITE(oct, CNXK_RST_CHIP_DOMAIN_W1S, 1);
drivers/net/ethernet/marvell/octeon_ep/octep_regs_cn9k_pf.h +30 -0
index ca473502d7a0..23fd0a697992 100644
@@ -5,6 +5,8 @@
  *
  */
 
+#include <linux/bitfield.h>
+
 #ifndef _OCTEP_REGS_CN9K_PF_H_
 #define _OCTEP_REGS_CN9K_PF_H_
 
@@ -383,6 +385,34 @@
 /* bit 1 for firmware heartbeat interrupt */
 #define CN93_SDP_EPF_OEI_RINT_DATA_BIT_HBEAT	BIT_ULL(1)
 
+#define FW_STATUS_DOWNING      0ULL
+#define FW_STATUS_RUNNING      2ULL
+
+#define CN9K_PEM_GENMASK BIT_ULL(36)
+#define CN9K_PF_GENMASK GENMASK_ULL(21, 18)
+#define CN9K_PFX_CSX_PFCFGX_SHADOW_BIT BIT_ULL(16)
+#define CN9K_PFX_CSX_PFCFGX_BASE_ADDR (0x8e0000008000ULL)
+#define CN9K_4BYTE_ALIGNED_ADDRESS_OFFSET(offset) ((offset) & BIT_ULL(2))
+#define CN9K_PEMX_PFX_CSX_PFCFGX cn9k_pemx_pfx_csx_pfcfgx
+
+static inline u64 cn9k_pemx_pfx_csx_pfcfgx(u64 pem, u32 pf, u32 offset)
+{
+	u32 shadow_addr_bit, pf_addr_bits, aligned_offset;
+	u64 pem_addr_bits;
+
+	pem_addr_bits = FIELD_PREP(CN9K_PEM_GENMASK, pem);
+	pf_addr_bits = FIELD_PREP(CN9K_PF_GENMASK, pf);
+	shadow_addr_bit = CN9K_PFX_CSX_PFCFGX_SHADOW_BIT & (offset);
+	aligned_offset = rounddown((offset), 8);
+
+	return (CN9K_PFX_CSX_PFCFGX_BASE_ADDR | pem_addr_bits
+		| pf_addr_bits | shadow_addr_bit | aligned_offset)
+		+ CN9K_4BYTE_ALIGNED_ADDRESS_OFFSET(offset);
+}
+
+/* Register defines for use with CN9K_PEMX_PFX_CSX_PFCFGX */
+#define CN9K_PCIEEP_VSECST_CTL  0x4D0
+
 #define CN93_PEM_BAR4_INDEX            7
 #define CN93_PEM_BAR4_INDEX_SIZE       0x400000ULL
 #define CN93_PEM_BAR4_INDEX_OFFSET     (CN93_PEM_BAR4_INDEX * CN93_PEM_BAR4_INDEX_SIZE)
drivers/net/ethernet/marvell/octeon_ep/octep_regs_cnxk_pf.h +1 -0
index e637d7c8224d..a6b6c9f356de 100644
@@ -396,6 +396,7 @@
 #define CNXK_SDP_EPF_OEI_RINT_DATA_BIT_MBOX	BIT_ULL(0)
 /* bit 1 for firmware heartbeat interrupt */
 #define CNXK_SDP_EPF_OEI_RINT_DATA_BIT_HBEAT	BIT_ULL(1)
+#define FW_STATUS_DOWNING      0ULL
 #define FW_STATUS_RUNNING      2ULL
 #define CNXK_PEMX_PFX_CSX_PFCFGX(pem, pf, offset)      ({ typeof(offset) _off = (offset); \
 							  ((0x8e0000008000 | \
Signature · 1 line
2.47.0
reviewer Reviewed-by on v4 1/1
#1 lore raw
Quoted 10 lines · Vimlesh Kumar

Add support to reset firmware ready status when the driver is removed(either in unload or unbind)

---
V4: Use static inline function instead of macro for readability.

Quoted 7 lines

V3:
- Reformat code to less than 80 columns wide.
- Use #defines for register constants.

V2: Use recommended bit manipulation macros.

V1: https://lore.kernel.org/all/[email protected]/

Thanks for hte updates.

#2 lore raw
In reply to Vimlesh Kumar
This patch was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:

Hello:

This patch was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:

Quoted 9 lines · On Thu, 15 Jan 2026 09:20:47 +0000 you wrote:

Add support to reset firmware ready status when the driver is removed(either in unload or unbind)

[...]

Here is the summary with links:
- [net-next,v4] octeon_ep: reset firmware ready status

    https://git.kernel.org/netdev/net-next/c/3b85d5f8562c

You are awesome, thank you!

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