Re: [PATCH v4 RESEND] tpm,tpm_tis: Disable interrupts after 1000 unhandled IRQs

["Jarkko Sakkinen" <jarkko@kernel.org>]

On Thu Jul 13, 2023 at 7:01 PM UTC, Lino Sanfilippo wrote:
> After activation of interrupts for TPM TIS drivers 0-day reports an
> interrupt storm on an Inspur NF5180M6 server.
>
> Fix this by detecting the storm and falling back to polling:
> Count the number of unhandled interrupts within a 10 ms time interval. In
> case that more than 1000 were unhandled deactivate interrupts entirely,
> deregister the handler and use polling instead.
>
> Also print a note to point to the tpm_tis_dmi_table.
>
> Since the interrupt deregistration function devm_free_irq() waits for all
> interrupt handlers to finish, only trigger a worker in the interrupt
> handler and do the unregistration in the worker to avoid a deadlock.
>
> Note: the storm detection logic equals the implementation in
> note_interrupt() which uses timestamps and counters stored in struct
> irq_desc. Since this structure is private to the generic interrupt core
> the TPM TIS core uses its own timestamps and counters. Furthermore the TPM
> interrupt handler always returns IRQ_HANDLED to prevent the generic
> interrupt core from processing the interrupt storm.
>
> Fixes: e644b2f498d2 ("tpm, tpm_tis: Enable interrupt test")
> Reported-by: kernel test robot <yujie.liu@intel.com>
> Closes: https://lore.kernel.org/oe-lkp/202305041325.ae8b0c43-yujie.liu@intel.com/
> Suggested-by: Lukas Wunner <lukas@wunner.de>
> Signed-off-by: Lino Sanfilippo <l.sanfilippo@kunbus.com>
> ---
>  drivers/char/tpm/tpm_tis_core.c | 103 +++++++++++++++++++++++++++-----
>  drivers/char/tpm/tpm_tis_core.h |   4 ++
>  2 files changed, 92 insertions(+), 15 deletions(-)
>
> Resending the patch due to several bounce messages for the first attempt.
>
> Changes to v3 (all requested by Jarko):
> - remove all inline comments 
> - rename tpm_tis_reenable_polling() to tpm_tis_revert_interrupts()
> - rename tpm_tis_check_for_interrupt_storm() to tpm_tis_update_unhandled_irqs()
> - rename label "unhandled" to "err"
> - add Fixes: tag
>
> Changes to v2:
> - use define for max number of unhandles irqs(requested by Jarko)
> - rename intmask to int_mask (requested by Jarko)
> - rephrased short summary (requested by Jarko)
> - rename disable_interrupts to tpm_tis_disable_interrupts (requested by Jarko)
> - print info message concerning adding an entry to tpm_tis_dmi_table
>   (suggested by Jerry)
> - amended commit message
> - handle failure of locality request by returning IRQ_NONE
> - dont take and release locality in __tpm_tis_disable_interrupts but in its
> caller
>
> diff --git a/drivers/char/tpm/tpm_tis_core.c b/drivers/char/tpm/tpm_tis_core.c
> index 558144fa707a..88a5384c09c0 100644
> --- a/drivers/char/tpm/tpm_tis_core.c
> +++ b/drivers/char/tpm/tpm_tis_core.c
> @@ -24,9 +24,12 @@
>  #include <linux/wait.h>
>  #include <linux/acpi.h>
>  #include <linux/freezer.h>
> +#include <linux/dmi.h>
>  #include "tpm.h"
>  #include "tpm_tis_core.h"
>  
> +#define TPM_TIS_MAX_UNHANDLED_IRQS	1000
> +
>  static void tpm_tis_clkrun_enable(struct tpm_chip *chip, bool value);
>  
>  static bool wait_for_tpm_stat_cond(struct tpm_chip *chip, u8 mask,
> @@ -468,25 +471,29 @@ static int tpm_tis_send_data(struct tpm_chip *chip, const u8 *buf, size_t len)
>  	return rc;
>  }
>  
> -static void disable_interrupts(struct tpm_chip *chip)
> +static void __tpm_tis_disable_interrupts(struct tpm_chip *chip)
> +{
> +	struct tpm_tis_data *priv = dev_get_drvdata(&chip->dev);
> +	u32 int_mask = 0;
> +
> +	tpm_tis_read32(priv, TPM_INT_ENABLE(priv->locality), &int_mask);
> +	int_mask &= ~TPM_GLOBAL_INT_ENABLE;
> +	tpm_tis_write32(priv, TPM_INT_ENABLE(priv->locality), int_mask);
> +
> +	chip->flags &= ~TPM_CHIP_FLAG_IRQ;
> +}
> +
> +static void tpm_tis_disable_interrupts(struct tpm_chip *chip)
>  {
>  	struct tpm_tis_data *priv = dev_get_drvdata(&chip->dev);
> -	u32 intmask;
> -	int rc;
>  
>  	if (priv->irq == 0)
>  		return;
>  
> -	rc = tpm_tis_read32(priv, TPM_INT_ENABLE(priv->locality), &intmask);
> -	if (rc < 0)
> -		intmask = 0;
> -
> -	intmask &= ~TPM_GLOBAL_INT_ENABLE;
> -	rc = tpm_tis_write32(priv, TPM_INT_ENABLE(priv->locality), intmask);
> +	__tpm_tis_disable_interrupts(chip);
>  
>  	devm_free_irq(chip->dev.parent, priv->irq, chip);
>  	priv->irq = 0;
> -	chip->flags &= ~TPM_CHIP_FLAG_IRQ;
>  }
>  
>  /*
> @@ -552,7 +559,7 @@ static int tpm_tis_send(struct tpm_chip *chip, u8 *buf, size_t len)
>  	if (!test_bit(TPM_TIS_IRQ_TESTED, &priv->flags))
>  		tpm_msleep(1);
>  	if (!test_bit(TPM_TIS_IRQ_TESTED, &priv->flags))
> -		disable_interrupts(chip);
> +		tpm_tis_disable_interrupts(chip);
>  	set_bit(TPM_TIS_IRQ_TESTED, &priv->flags);
>  	return rc;
>  }
> @@ -752,6 +759,57 @@ static bool tpm_tis_req_canceled(struct tpm_chip *chip, u8 status)
>  	return status == TPM_STS_COMMAND_READY;
>  }
>  
> +static irqreturn_t tpm_tis_revert_interrupts(struct tpm_chip *chip)
> +{
> +	struct tpm_tis_data *priv = dev_get_drvdata(&chip->dev);
> +	const char *product;
> +	const char *vendor;
> +
> +	dev_warn(&chip->dev, FW_BUG
> +		 "TPM interrupt storm detected, polling instead\n");
> +
> +	vendor = dmi_get_system_info(DMI_SYS_VENDOR);
> +	product = dmi_get_system_info(DMI_PRODUCT_VERSION);
> +
> +	if (vendor && product) {
> +		dev_info(&chip->dev,
> +			"Consider adding the following entry to tpm_tis_dmi_table:\n");
> +		dev_info(&chip->dev, "\tDMI_SYS_VENDOR: %s\n", vendor);
> +		dev_info(&chip->dev, "\tDMI_PRODUCT_VERSION: %s\n", product);
> +	}
> +
> +	if (tpm_tis_request_locality(chip, 0) != 0)
> +		return IRQ_NONE;
> +
> +	__tpm_tis_disable_interrupts(chip);
> +	tpm_tis_relinquish_locality(chip, 0);
> +
> +	schedule_work(&priv->free_irq_work);
> +
> +	return IRQ_HANDLED;
> +}
> +
> +static irqreturn_t tpm_tis_update_unhandled_irqs(struct tpm_chip *chip)
> +{
> +	struct tpm_tis_data *priv = dev_get_drvdata(&chip->dev);
> +	irqreturn_t irqret = IRQ_HANDLED;
> +
> +	if (!(chip->flags & TPM_CHIP_FLAG_IRQ))
> +		return IRQ_HANDLED;
> +
> +	if (time_after(jiffies, priv->last_unhandled_irq + HZ/10))
> +		priv->unhandled_irqs = 1;
> +	else
> +		priv->unhandled_irqs++;
> +
> +	priv->last_unhandled_irq = jiffies;
> +
> +	if (priv->unhandled_irqs > TPM_TIS_MAX_UNHANDLED_IRQS)
> +		irqret = tpm_tis_revert_interrupts(chip);
> +
> +	return irqret;
> +}
> +
>  static irqreturn_t tis_int_handler(int dummy, void *dev_id)
>  {
>  	struct tpm_chip *chip = dev_id;
> @@ -761,10 +819,10 @@ static irqreturn_t tis_int_handler(int dummy, void *dev_id)
>  
>  	rc = tpm_tis_read32(priv, TPM_INT_STATUS(priv->locality), &interrupt);
>  	if (rc < 0)
> -		return IRQ_NONE;
> +		goto err;
>  
>  	if (interrupt == 0)
> -		return IRQ_NONE;
> +		goto err;
>  
>  	set_bit(TPM_TIS_IRQ_TESTED, &priv->flags);
>  	if (interrupt & TPM_INTF_DATA_AVAIL_INT)
> @@ -780,10 +838,13 @@ static irqreturn_t tis_int_handler(int dummy, void *dev_id)
>  	rc = tpm_tis_write32(priv, TPM_INT_STATUS(priv->locality), interrupt);
>  	tpm_tis_relinquish_locality(chip, 0);
>  	if (rc < 0)
> -		return IRQ_NONE;
> +		goto err;
>  
>  	tpm_tis_read32(priv, TPM_INT_STATUS(priv->locality), &interrupt);
>  	return IRQ_HANDLED;
> +
> +err:
> +	return tpm_tis_update_unhandled_irqs(chip);
>  }
>  
>  static void tpm_tis_gen_interrupt(struct tpm_chip *chip)
> @@ -804,6 +865,15 @@ static void tpm_tis_gen_interrupt(struct tpm_chip *chip)
>  		chip->flags &= ~TPM_CHIP_FLAG_IRQ;
>  }
>  
> +static void tpm_tis_free_irq_func(struct work_struct *work)
> +{
> +	struct tpm_tis_data *priv = container_of(work, typeof(*priv), free_irq_work);
> +	struct tpm_chip *chip = priv->chip;
> +
> +	devm_free_irq(chip->dev.parent, priv->irq, chip);
> +	priv->irq = 0;
> +}
> +
>  /* Register the IRQ and issue a command that will cause an interrupt. If an
>   * irq is seen then leave the chip setup for IRQ operation, otherwise reverse
>   * everything and leave in polling mode. Returns 0 on success.
> @@ -816,6 +886,7 @@ static int tpm_tis_probe_irq_single(struct tpm_chip *chip, u32 intmask,
>  	int rc;
>  	u32 int_status;
>  
> +	INIT_WORK(&priv->free_irq_work, tpm_tis_free_irq_func);
>  
>  	rc = devm_request_threaded_irq(chip->dev.parent, irq, NULL,
>  				       tis_int_handler, IRQF_ONESHOT | flags,
> @@ -918,6 +989,7 @@ void tpm_tis_remove(struct tpm_chip *chip)
>  		interrupt = 0;
>  
>  	tpm_tis_write32(priv, reg, ~TPM_GLOBAL_INT_ENABLE & interrupt);
> +	flush_work(&priv->free_irq_work);
>  
>  	tpm_tis_clkrun_enable(chip, false);
>  
> @@ -1021,6 +1093,7 @@ int tpm_tis_core_init(struct device *dev, struct tpm_tis_data *priv, int irq,
>  	chip->timeout_b = msecs_to_jiffies(TIS_TIMEOUT_B_MAX);
>  	chip->timeout_c = msecs_to_jiffies(TIS_TIMEOUT_C_MAX);
>  	chip->timeout_d = msecs_to_jiffies(TIS_TIMEOUT_D_MAX);
> +	priv->chip = chip;
>  	priv->timeout_min = TPM_TIMEOUT_USECS_MIN;
>  	priv->timeout_max = TPM_TIMEOUT_USECS_MAX;
>  	priv->phy_ops = phy_ops;
> @@ -1179,7 +1252,7 @@ int tpm_tis_core_init(struct device *dev, struct tpm_tis_data *priv, int irq,
>  			rc = tpm_tis_request_locality(chip, 0);
>  			if (rc < 0)
>  				goto out_err;
> -			disable_interrupts(chip);
> +			tpm_tis_disable_interrupts(chip);
>  			tpm_tis_relinquish_locality(chip, 0);
>  		}
>  	}
> diff --git a/drivers/char/tpm/tpm_tis_core.h b/drivers/char/tpm/tpm_tis_core.h
> index 610bfadb6acf..b1a169d7d1ca 100644
> --- a/drivers/char/tpm/tpm_tis_core.h
> +++ b/drivers/char/tpm/tpm_tis_core.h
> @@ -91,11 +91,15 @@ enum tpm_tis_flags {
>  };
>  
>  struct tpm_tis_data {
> +	struct tpm_chip *chip;
>  	u16 manufacturer_id;
>  	struct mutex locality_count_mutex;
>  	unsigned int locality_count;
>  	int locality;
>  	int irq;
> +	struct work_struct free_irq_work;
> +	unsigned long last_unhandled_irq;
> +	unsigned int unhandled_irqs;
>  	unsigned int int_mask;
>  	unsigned long flags;
>  	void __iomem *ilb_base_addr;
>
> base-commit: 6995e2de6891c724bfeb2db33d7b87775f913ad1
> -- 
> 2.25.1

Reviewed-by: Jarkko Sakkinen <jarkko@kernel.org>

Looks very good, thank you.

BR, Jarkko