linux-kernel - Re: [PATCH v2] TCO watchdog pretimeout handler

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Tue, 16 Jun 2015 22:14:57 -0700
From:	Guenter Roeck <linux@...ck-us.net>
To:	Francois-Nicolas Muller <francois-nicolas.muller@...el.com>
CC:	wim@...ana.be, linux-watchdog@...r.kernel.org,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2] TCO watchdog pretimeout handler

On 06/16/2015 06:45 AM, Francois-Nicolas Muller wrote:
> Use TCO watchdog first timeout (pretimeout) to dump CPU backtraces
> and ease debug of watchdog expiration causes.
> TCO logic generates a SCI interrupt, then its handler dumps all CPU
> backtraces and calls panic (in order to execute registered panic
> callbacks).
> SCI interrupt number (GPE) is configured from ACPI tables.
>
> Signed-off-by: Francois-Nicolas Muller <francois-nicolas.muller@...el.com>
> ---
> Thanks Guenter for your review.
>
>> If I recall correctly, the iTCO watchdog can also generate an NMI.
>> Would it make sense to add support for handling this NMI as well ?
>
> As far as I know, there is no NMI option for TCO watchdog interrupt.
> Do you have any documentation about this ?
>

Actually that was a miscommunication, sorry. I confused it waith another watchdog.

I assume you took out all mention of SMI because it is not (yet) supported.
Would be interesting to know what systems out there actually use / configure.

  Here is a new version (v2) of the patch:
> - rebased on latest kernel
> - fixed coding style issues
>
> Francois-Nicolas
> ---
>   drivers/watchdog/iTCO_wdt.c | 50 +++++++++++++++++++++++++++++++++++++++++++++
>   1 file changed, 50 insertions(+)
>
> diff --git a/drivers/watchdog/iTCO_wdt.c b/drivers/watchdog/iTCO_wdt.c
> index 3c3fd41..cd2569a 100644
> --- a/drivers/watchdog/iTCO_wdt.c
> +++ b/drivers/watchdog/iTCO_wdt.c
> @@ -68,6 +68,8 @@
>   #include <linux/io.h>			/* For inb/outb/... */
>   #include <linux/mfd/core.h>
>   #include <linux/mfd/lpc_ich.h>
> +#include <linux/nmi.h>
> +#include <linux/acpi.h>
>
>   #include "iTCO_vendor.h"
>
> @@ -127,6 +129,12 @@ module_param(turn_SMI_watchdog_clear_off, int, 0);
>   MODULE_PARM_DESC(turn_SMI_watchdog_clear_off,
>   	"Turn off SMI clearing watchdog (depends on TCO-version)(default=1)");
>
> +#define DEFAULT_PRETIMEOUT 0
> +static bool pretimeout = DEFAULT_PRETIMEOUT;
> +module_param(pretimeout, bool, 0);
> +MODULE_PARM_DESC(pretimeout, "Enable watchdog pretimeout (default="
> +				__MODULE_STRING(DEFAULT_PRETIMEOUT) ")");
> +
>   /*
>    * Some TCO specific functions
>    */
> @@ -201,6 +209,45 @@ static int iTCO_wdt_unset_NO_REBOOT_bit(void)
>   	return ret; /* returns: 0 = OK, -EIO = Error */
>   }
>
> +static unsigned char *tco_hid = "8086229C";
> +

Do people understand what this means ? Is that some Intel magic string ?
Does this work for all instances of iTCO watchdogs, or only for a specific
system or iTCO version ?

Rafael asked this question as well, but I don't recall seeing an answer.

I see that it maps to a PCI ID for Intel Braswell, but I have no idea
how that translates to something useful for ACPI. Is this a well defined
(and allocated) ACPI HID ? How about other chips (non-Braswell)
which are supported by this driver ?

> +static u32 iTCO_wdt_pretimeout_handler(acpi_handle gpe_device, u32 gpe,
> +				       void *context)
> +{
> +	/* dump backtraces for all available cores */
> +	trigger_all_cpu_backtrace();
> +
> +	/* call panic notifiers */
> +	panic("Kernel Watchdog");
> +
> +	return ACPI_INTERRUPT_HANDLED;
> +}
> +
> +static acpi_status __init iTCO_wdt_register_gpe(acpi_handle handle,
> +					u32 lvl, void *context, void **rv)
> +{
> +	unsigned long long gpe;
> +	acpi_status status;
> +	union acpi_object object = { 0 };
> +	struct acpi_buffer buffer = { sizeof(union acpi_object), &object };
> +
> +	status = acpi_evaluate_object(handle, "_GPE", NULL, &buffer);
> +	if (ACPI_FAILURE(status))
> +		return status;
> +
> +	if (object.type != ACPI_TYPE_INTEGER)
> +		return AE_BAD_DATA;
> +
> +	gpe = object.integer.value;
> +	status = acpi_install_gpe_handler(NULL, gpe, ACPI_GPE_EDGE_TRIGGERED,
> +					  iTCO_wdt_pretimeout_handler, NULL);

Do we know for sure that _GPE is always associated with the watchdog ?
Is that because of tco_hid ?

Thanks,
Guenter

> +	if (ACPI_FAILURE(status))
> +		return status;
> +
> +	acpi_enable_gpe(NULL, gpe);
> +	return AE_OK;
> +}
> +
>   static int iTCO_wdt_start(struct watchdog_device *wd_dev)
>   {
>   	unsigned int val;
> @@ -641,6 +688,9 @@ static int __init iTCO_wdt_init_module(void)
>   	if (err)
>   		return err;
>
> +	if (pretimeout)
> +		acpi_get_devices(tco_hid, iTCO_wdt_register_gpe, NULL, NULL);
> +
>   	return 0;
>   }
>
>

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/