diff mbox series

[iwl-net,v2,1/4] igc: Ensure the PTM cycle is reliably triggered

Message ID 20241023023040.111429-2-christopher.s.hall@intel.com (mailing list archive)
State Awaiting Upstream
Delegated to: Netdev Maintainers
Headers show
Series igc: Fix PTM timeout | expand

Checks

Context Check Description
netdev/series_format success Posting correctly formatted
netdev/tree_selection success Clearly marked for net
netdev/ynl success Generated files up to date; no warnings/errors; no diff in generated;
netdev/fixes_present success Fixes tag present in non-next series
netdev/header_inline success No static functions without inline keyword in header files
netdev/build_32bit success Errors and warnings before: 5 this patch: 5
netdev/build_tools success No tools touched, skip
netdev/cc_maintainers fail 1 blamed authors not CCed: anthony.l.nguyen@intel.com; 7 maintainers not CCed: anthony.l.nguyen@intel.com edumazet@google.com andrew+netdev@lunn.ch przemyslaw.kitszel@intel.com pabeni@redhat.com richardcochran@gmail.com kuba@kernel.org
netdev/build_clang success Errors and warnings before: 5 this patch: 5
netdev/verify_signedoff success Signed-off-by tag matches author and committer
netdev/deprecated_api success None detected
netdev/check_selftest success No net selftest shell script
netdev/verify_fixes success Fixes tag looks correct
netdev/build_allmodconfig_warn success Errors and warnings before: 4 this patch: 4
netdev/checkpatch warning WARNING: line length of 84 exceeds 80 columns WARNING: please, no space before tabs
netdev/build_clang_rust success No Rust files in patch. Skipping build
netdev/kdoc success Errors and warnings before: 6 this patch: 6
netdev/source_inline success Was 0 now: 0

Commit Message

Hall, Christopher S Oct. 23, 2024, 2:30 a.m. UTC 
From: Christopher S M Hall <christopher.s.hall@intel.com>

Writing to clear the PTM status 'valid' bit while the PTM cycle is
triggered results in unreliable PTM operation. To fix this, clear the
PTM 'trigger' and status after each PTM transaction.

The issue can be reproduced with the following:

$ sudo phc2sys -R 1000 -O 0 -i tsn0 -m

Note: 1000 Hz (-R 1000) is unrealistically large, but provides a way to
quickly reproduce the issue.

PHC2SYS exits with:

"ioctl PTP_OFFSET_PRECISE: Connection timed out" when the PTM transaction
  fails

Fixes: a90ec8483732 ("igc: Add support for PTP getcrosststamp()")
Signed-off-by: Christopher S M Hall <christopher.s.hall@intel.com>
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
---
 drivers/net/ethernet/intel/igc/igc_defines.h |  1 +
 drivers/net/ethernet/intel/igc/igc_ptp.c     | 70 ++++++++++++--------
 2 files changed, 42 insertions(+), 29 deletions(-)

Comments

Tony Nguyen Oct. 29, 2024, 9:21 p.m. UTC | #1 
On 10/22/2024 7:30 PM, Chris H wrote:

...

>   static int igc_phc_get_syncdevicetime(ktime_t *device,
>   				      struct system_counterval_t *system,
>   				      void *ctx)
>   {
> -	u32 stat, t2_curr_h, t2_curr_l, ctrl;
> +	u32 stat, t2_curr_h, t2_curr_l;

This needs to move a couple of lines down now for RCT.

Thanks,
Tony

>   	struct igc_adapter *adapter = ctx;
>   	struct igc_hw *hw = &adapter->hw;
>   	int err, count = 100;
Corinna Vinschen Oct. 30, 2024, 10:17 a.m. UTC | #2 
Hi Chris,

On Oct 23 02:30, Chris H wrote:
> From: Christopher S M Hall <christopher.s.hall@intel.com>
> 
> Writing to clear the PTM status 'valid' bit while the PTM cycle is
> triggered results in unreliable PTM operation. To fix this, clear the
> PTM 'trigger' and status after each PTM transaction.
> 
> The issue can be reproduced with the following:
> 
> $ sudo phc2sys -R 1000 -O 0 -i tsn0 -m
> 
> Note: 1000 Hz (-R 1000) is unrealistically large, but provides a way to
> quickly reproduce the issue.
> 
> PHC2SYS exits with:
> 
> "ioctl PTP_OFFSET_PRECISE: Connection timed out" when the PTM transaction
>   fails
> 
> Fixes: a90ec8483732 ("igc: Add support for PTP getcrosststamp()")
> Signed-off-by: Christopher S M Hall <christopher.s.hall@intel.com>
> Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>

In the v1 thread in August I asked for adding a description of the kdump
problem we observed to the commit message.  I came up with this text,
maybe we can still add it, or maybe a shhortened version of it?

  This patch also fixes a hang in igc_probe() when loading the igc
  driver in the kdump kernel on systems supporting PTM.

  The igc driver running in the base kernel enables PTM trigger in
  igc_probe().  Therefore the driver is always in PTM trigger mode,
  except in brief periods when manually triggering a PTM cycle.

  When a crash occurs, the NIC is reset while PTM trigger is enabled.
  Due to a hardware problem, the NIC is subsequently in a bad busmaster
  state and doesn't handle register reads/writes.  When running
  igc_probe() in the kdump kernel, the first register access to a NIC
  register hangs driver probing and ultimately breaks kdump.

  With this patch, igc has PTM trigger disabled most of the time,
  and the trigger is only enabled for very brief (10 - 100 us) periods
  when manually triggering a PTM cycle.  Chances that a crash occurs
  during a PTM trigger are not 0, but extremly reduced.


Thanks,
Corinna


> ---
>  drivers/net/ethernet/intel/igc/igc_defines.h |  1 +
>  drivers/net/ethernet/intel/igc/igc_ptp.c     | 70 ++++++++++++--------
>  2 files changed, 42 insertions(+), 29 deletions(-)
> 
> diff --git a/drivers/net/ethernet/intel/igc/igc_defines.h b/drivers/net/ethernet/intel/igc/igc_defines.h
> index 8e449904aa7d..afd0512dc9f8 100644
> --- a/drivers/net/ethernet/intel/igc/igc_defines.h
> +++ b/drivers/net/ethernet/intel/igc/igc_defines.h
> @@ -593,6 +593,7 @@
>  #define IGC_PTM_STAT_T4M1_OVFL		BIT(3) /* T4 minus T1 overflow */
>  #define IGC_PTM_STAT_ADJUST_1ST		BIT(4) /* 1588 timer adjusted during 1st PTM cycle */
>  #define IGC_PTM_STAT_ADJUST_CYC		BIT(5) /* 1588 timer adjusted during non-1st PTM cycle */
> +#define IGC_PTM_STAT_ALL        	GENMASK(5, 0) /* Used to clear all status */
>  
>  /* PCIe PTM Cycle Control */
>  #define IGC_PTM_CYCLE_CTRL_CYC_TIME(msec)	((msec) & 0x3ff) /* PTM Cycle Time (msec) */
> diff --git a/drivers/net/ethernet/intel/igc/igc_ptp.c b/drivers/net/ethernet/intel/igc/igc_ptp.c
> index 946edbad4302..00cc80d8d164 100644
> --- a/drivers/net/ethernet/intel/igc/igc_ptp.c
> +++ b/drivers/net/ethernet/intel/igc/igc_ptp.c
> @@ -974,11 +974,38 @@ static void igc_ptm_log_error(struct igc_adapter *adapter, u32 ptm_stat)
>  	}
>  }
>  
> +static void igc_ptm_trigger(struct igc_hw *hw)
> +{
> +	u32 ctrl;
> +
> +	/* To "manually" start the PTM cycle we need to set the
> +	 * trigger (TRIG) bit
> +	 */
> +	ctrl = rd32(IGC_PTM_CTRL);
> +	ctrl |= IGC_PTM_CTRL_TRIG;
> +	wr32(IGC_PTM_CTRL, ctrl);
> +	/* Perform flush after write to CTRL register otherwise
> +	 * transaction may not start
> +	 */
> +	wrfl();
> +}
> +
> +static void igc_ptm_reset(struct igc_hw *hw)
> +{
> +	u32 ctrl;
> +
> +	ctrl = rd32(IGC_PTM_CTRL);
> +	ctrl &= ~IGC_PTM_CTRL_TRIG;
> +	wr32(IGC_PTM_CTRL, ctrl);
> +	/* Write to clear all status */
> +	wr32(IGC_PTM_STAT, IGC_PTM_STAT_ALL);
> +}
> +
>  static int igc_phc_get_syncdevicetime(ktime_t *device,
>  				      struct system_counterval_t *system,
>  				      void *ctx)
>  {
> -	u32 stat, t2_curr_h, t2_curr_l, ctrl;
> +	u32 stat, t2_curr_h, t2_curr_l;
>  	struct igc_adapter *adapter = ctx;
>  	struct igc_hw *hw = &adapter->hw;
>  	int err, count = 100;
> @@ -994,25 +1021,13 @@ static int igc_phc_get_syncdevicetime(ktime_t *device,
>  		 * are transitory. Repeating the process returns valid
>  		 * data eventually.
>  		 */
> -
> -		/* To "manually" start the PTM cycle we need to clear and
> -		 * then set again the TRIG bit.
> -		 */
> -		ctrl = rd32(IGC_PTM_CTRL);
> -		ctrl &= ~IGC_PTM_CTRL_TRIG;
> -		wr32(IGC_PTM_CTRL, ctrl);
> -		ctrl |= IGC_PTM_CTRL_TRIG;
> -		wr32(IGC_PTM_CTRL, ctrl);
> -
> -		/* The cycle only starts "for real" when software notifies
> -		 * that it has read the registers, this is done by setting
> -		 * VALID bit.
> -		 */
> -		wr32(IGC_PTM_STAT, IGC_PTM_STAT_VALID);
> +		igc_ptm_trigger(hw);
>  
>  		err = readx_poll_timeout(rd32, IGC_PTM_STAT, stat,
>  					 stat, IGC_PTM_STAT_SLEEP,
>  					 IGC_PTM_STAT_TIMEOUT);
> +		igc_ptm_reset(hw);
> +
>  		if (err < 0) {
>  			netdev_err(adapter->netdev, "Timeout reading IGC_PTM_STAT register\n");
>  			return err;
> @@ -1021,15 +1036,7 @@ static int igc_phc_get_syncdevicetime(ktime_t *device,
>  		if ((stat & IGC_PTM_STAT_VALID) == IGC_PTM_STAT_VALID)
>  			break;
>  
> -		if (stat & ~IGC_PTM_STAT_VALID) {
> -			/* An error occurred, log it. */
> -			igc_ptm_log_error(adapter, stat);
> -			/* The STAT register is write-1-to-clear (W1C),
> -			 * so write the previous error status to clear it.
> -			 */
> -			wr32(IGC_PTM_STAT, stat);
> -			continue;
> -		}
> +		igc_ptm_log_error(adapter, stat);
>  	} while (--count);
>  
>  	if (!count) {
> @@ -1255,7 +1262,7 @@ void igc_ptp_stop(struct igc_adapter *adapter)
>  void igc_ptp_reset(struct igc_adapter *adapter)
>  {
>  	struct igc_hw *hw = &adapter->hw;
> -	u32 cycle_ctrl, ctrl;
> +	u32 cycle_ctrl, ctrl, stat;
>  	unsigned long flags;
>  	u32 timadj;
>  
> @@ -1290,14 +1297,19 @@ void igc_ptp_reset(struct igc_adapter *adapter)
>  		ctrl = IGC_PTM_CTRL_EN |
>  			IGC_PTM_CTRL_START_NOW |
>  			IGC_PTM_CTRL_SHRT_CYC(IGC_PTM_SHORT_CYC_DEFAULT) |
> -			IGC_PTM_CTRL_PTM_TO(IGC_PTM_TIMEOUT_DEFAULT) |
> -			IGC_PTM_CTRL_TRIG;
> +			IGC_PTM_CTRL_PTM_TO(IGC_PTM_TIMEOUT_DEFAULT);
>  
>  		wr32(IGC_PTM_CTRL, ctrl);
>  
>  		/* Force the first cycle to run. */
> -		wr32(IGC_PTM_STAT, IGC_PTM_STAT_VALID);
> +		igc_ptm_trigger(hw);
> +
> +		if (readx_poll_timeout_atomic(rd32, IGC_PTM_STAT, stat,
> +					      stat, IGC_PTM_STAT_SLEEP,
> +					      IGC_PTM_STAT_TIMEOUT))
> +			netdev_err(adapter->netdev, "Timeout reading IGC_PTM_STAT register\n");
>  
> +		igc_ptm_reset(hw);
>  		break;
>  	default:
>  		/* No work to do. */
> -- 
> 2.34.1
Avigail Dahan Oct. 31, 2024, 9:34 a.m. UTC | #3 
On 23/10/2024 5:30, Chris H wrote:
> From: Christopher S M Hall <christopher.s.hall@intel.com>
> 
> Writing to clear the PTM status 'valid' bit while the PTM cycle is
> triggered results in unreliable PTM operation. To fix this, clear the
> PTM 'trigger' and status after each PTM transaction.
> 
> The issue can be reproduced with the following:
> 
> $ sudo phc2sys -R 1000 -O 0 -i tsn0 -m
> 
> Note: 1000 Hz (-R 1000) is unrealistically large, but provides a way to
> quickly reproduce the issue.
> 
> PHC2SYS exits with:
> 
> "ioctl PTP_OFFSET_PRECISE: Connection timed out" when the PTM transaction
>    fails
> 
> Fixes: a90ec8483732 ("igc: Add support for PTP getcrosststamp()")
> Signed-off-by: Christopher S M Hall <christopher.s.hall@intel.com>
> Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
> ---
>   drivers/net/ethernet/intel/igc/igc_defines.h |  1 +
>   drivers/net/ethernet/intel/igc/igc_ptp.c     | 70 ++++++++++++--------
>   2 files changed, 42 insertions(+), 29 deletions(-)
> 
Tested-by: Avigail Dahan <avigailx.dahan@intel.com>
diff mbox series

Patch

diff --git a/drivers/net/ethernet/intel/igc/igc_defines.h b/drivers/net/ethernet/intel/igc/igc_defines.h
index 8e449904aa7d..afd0512dc9f8 100644
--- a/drivers/net/ethernet/intel/igc/igc_defines.h
+++ b/drivers/net/ethernet/intel/igc/igc_defines.h
@@ -593,6 +593,7 @@ 
 #define IGC_PTM_STAT_T4M1_OVFL		BIT(3) /* T4 minus T1 overflow */
 #define IGC_PTM_STAT_ADJUST_1ST		BIT(4) /* 1588 timer adjusted during 1st PTM cycle */
 #define IGC_PTM_STAT_ADJUST_CYC		BIT(5) /* 1588 timer adjusted during non-1st PTM cycle */
+#define IGC_PTM_STAT_ALL        	GENMASK(5, 0) /* Used to clear all status */
 
 /* PCIe PTM Cycle Control */
 #define IGC_PTM_CYCLE_CTRL_CYC_TIME(msec)	((msec) & 0x3ff) /* PTM Cycle Time (msec) */
diff --git a/drivers/net/ethernet/intel/igc/igc_ptp.c b/drivers/net/ethernet/intel/igc/igc_ptp.c
index 946edbad4302..00cc80d8d164 100644
--- a/drivers/net/ethernet/intel/igc/igc_ptp.c
+++ b/drivers/net/ethernet/intel/igc/igc_ptp.c
@@ -974,11 +974,38 @@  static void igc_ptm_log_error(struct igc_adapter *adapter, u32 ptm_stat)
 	}
 }
 
+static void igc_ptm_trigger(struct igc_hw *hw)
+{
+	u32 ctrl;
+
+	/* To "manually" start the PTM cycle we need to set the
+	 * trigger (TRIG) bit
+	 */
+	ctrl = rd32(IGC_PTM_CTRL);
+	ctrl |= IGC_PTM_CTRL_TRIG;
+	wr32(IGC_PTM_CTRL, ctrl);
+	/* Perform flush after write to CTRL register otherwise
+	 * transaction may not start
+	 */
+	wrfl();
+}
+
+static void igc_ptm_reset(struct igc_hw *hw)
+{
+	u32 ctrl;
+
+	ctrl = rd32(IGC_PTM_CTRL);
+	ctrl &= ~IGC_PTM_CTRL_TRIG;
+	wr32(IGC_PTM_CTRL, ctrl);
+	/* Write to clear all status */
+	wr32(IGC_PTM_STAT, IGC_PTM_STAT_ALL);
+}
+
 static int igc_phc_get_syncdevicetime(ktime_t *device,
 				      struct system_counterval_t *system,
 				      void *ctx)
 {
-	u32 stat, t2_curr_h, t2_curr_l, ctrl;
+	u32 stat, t2_curr_h, t2_curr_l;
 	struct igc_adapter *adapter = ctx;
 	struct igc_hw *hw = &adapter->hw;
 	int err, count = 100;
@@ -994,25 +1021,13 @@  static int igc_phc_get_syncdevicetime(ktime_t *device,
 		 * are transitory. Repeating the process returns valid
 		 * data eventually.
 		 */
-
-		/* To "manually" start the PTM cycle we need to clear and
-		 * then set again the TRIG bit.
-		 */
-		ctrl = rd32(IGC_PTM_CTRL);
-		ctrl &= ~IGC_PTM_CTRL_TRIG;
-		wr32(IGC_PTM_CTRL, ctrl);
-		ctrl |= IGC_PTM_CTRL_TRIG;
-		wr32(IGC_PTM_CTRL, ctrl);
-
-		/* The cycle only starts "for real" when software notifies
-		 * that it has read the registers, this is done by setting
-		 * VALID bit.
-		 */
-		wr32(IGC_PTM_STAT, IGC_PTM_STAT_VALID);
+		igc_ptm_trigger(hw);
 
 		err = readx_poll_timeout(rd32, IGC_PTM_STAT, stat,
 					 stat, IGC_PTM_STAT_SLEEP,
 					 IGC_PTM_STAT_TIMEOUT);
+		igc_ptm_reset(hw);
+
 		if (err < 0) {
 			netdev_err(adapter->netdev, "Timeout reading IGC_PTM_STAT register\n");
 			return err;
@@ -1021,15 +1036,7 @@  static int igc_phc_get_syncdevicetime(ktime_t *device,
 		if ((stat & IGC_PTM_STAT_VALID) == IGC_PTM_STAT_VALID)
 			break;
 
-		if (stat & ~IGC_PTM_STAT_VALID) {
-			/* An error occurred, log it. */
-			igc_ptm_log_error(adapter, stat);
-			/* The STAT register is write-1-to-clear (W1C),
-			 * so write the previous error status to clear it.
-			 */
-			wr32(IGC_PTM_STAT, stat);
-			continue;
-		}
+		igc_ptm_log_error(adapter, stat);
 	} while (--count);
 
 	if (!count) {
@@ -1255,7 +1262,7 @@  void igc_ptp_stop(struct igc_adapter *adapter)
 void igc_ptp_reset(struct igc_adapter *adapter)
 {
 	struct igc_hw *hw = &adapter->hw;
-	u32 cycle_ctrl, ctrl;
+	u32 cycle_ctrl, ctrl, stat;
 	unsigned long flags;
 	u32 timadj;
 
@@ -1290,14 +1297,19 @@  void igc_ptp_reset(struct igc_adapter *adapter)
 		ctrl = IGC_PTM_CTRL_EN |
 			IGC_PTM_CTRL_START_NOW |
 			IGC_PTM_CTRL_SHRT_CYC(IGC_PTM_SHORT_CYC_DEFAULT) |
-			IGC_PTM_CTRL_PTM_TO(IGC_PTM_TIMEOUT_DEFAULT) |
-			IGC_PTM_CTRL_TRIG;
+			IGC_PTM_CTRL_PTM_TO(IGC_PTM_TIMEOUT_DEFAULT);
 
 		wr32(IGC_PTM_CTRL, ctrl);
 
 		/* Force the first cycle to run. */
-		wr32(IGC_PTM_STAT, IGC_PTM_STAT_VALID);
+		igc_ptm_trigger(hw);
+
+		if (readx_poll_timeout_atomic(rd32, IGC_PTM_STAT, stat,
+					      stat, IGC_PTM_STAT_SLEEP,
+					      IGC_PTM_STAT_TIMEOUT))
+			netdev_err(adapter->netdev, "Timeout reading IGC_PTM_STAT register\n");
 
+		igc_ptm_reset(hw);
 		break;
 	default:
 		/* No work to do. */