lkml.org 
[lkml]   [2022]   [Dec]   [21]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Date
SubjectRe: [PATCH] drivers/cros_ec: Handle CrOS EC panics
On Wed, Dec 21, 2022 at 12:23 PM Prashant Malani <pmalani@chromium.org> wrote:
>
> Hi Rob,
>
> I'd suggest using the commit title tag "platform/chrome: cros_ec: ..."
> for commits which are "ChromeOS EC" wide. That's in line with
> other recent commits in this directory.

Ack. Will do.

>
> On Wed, Dec 21, 2022 at 10:56 AM Rob Barnes <robbarnes@google.com> wrote:
> >
> > Add handler for CrOS EC panic events. When a panic is reported,
> > poll EC log then force an orderly shutdown.
> >
> > This will preserve the EC log leading up to the crash.
> >
> > Signed-off-by: Rob Barnes <robbarnes@google.com>
> > ---
> > drivers/platform/chrome/cros_ec_debugfs.c | 24 +++++++++++++++++++++
> > drivers/platform/chrome/cros_ec_lpc.c | 10 +++++++++
> > include/linux/platform_data/cros_ec_proto.h | 9 ++++++++
> > 3 files changed, 43 insertions(+)
> >
> > diff --git a/drivers/platform/chrome/cros_ec_debugfs.c b/drivers/platform/chrome/cros_ec_debugfs.c
> > index 21d973fc6be2..31637a4e4cf9 100644
> > --- a/drivers/platform/chrome/cros_ec_debugfs.c
> > +++ b/drivers/platform/chrome/cros_ec_debugfs.c
> > @@ -49,6 +49,7 @@ struct cros_ec_debugfs {
> > struct delayed_work log_poll_work;
> > /* EC panicinfo */
> > struct debugfs_blob_wrapper panicinfo_blob;
> > + struct notifier_block notifier_panic;
> > };
> >
> > /*
> > @@ -437,6 +438,23 @@ static int cros_ec_create_panicinfo(struct cros_ec_debugfs *debug_info)
> > return ret;
> > }
> >
> > +static int cros_ec_debugfs_panic_event(struct notifier_block *nb,
> > + unsigned long queued_during_suspend,
> > + void *_notify)
> > +{
> > + struct cros_ec_debugfs *debug_info =
> > + container_of(nb, struct cros_ec_debugfs, notifier_panic);
> > +
> > + if (debug_info->log_buffer.buf) {
> > + /* Force log poll work to run immediately */
> > + mod_delayed_work(debug_info->log_poll_work.wq, &debug_info->log_poll_work, 0);
> > + /* Block until log poll work finishes */
> > + flush_delayed_work(&debug_info->log_poll_work);
> > + }
> > +
> > + return NOTIFY_DONE;
> > +}
> > +
> > static int cros_ec_debugfs_probe(struct platform_device *pd)
> > {
> > struct cros_ec_dev *ec = dev_get_drvdata(pd->dev.parent);
> > @@ -473,6 +491,12 @@ static int cros_ec_debugfs_probe(struct platform_device *pd)
> > debugfs_create_u16("suspend_timeout_ms", 0664, debug_info->dir,
> > &ec->ec_dev->suspend_timeout_ms);
> >
> > + debug_info->notifier_panic.notifier_call = cros_ec_debugfs_panic_event;
> > + ret = blocking_notifier_chain_register(&ec->ec_dev->panic_notifier,
> > + &debug_info->notifier_panic);
> > + if (ret)
> > + goto remove_debugfs;
> > +
> > ec->debug_info = debug_info;
> >
> > dev_set_drvdata(&pd->dev, ec);
> > diff --git a/drivers/platform/chrome/cros_ec_lpc.c b/drivers/platform/chrome/cros_ec_lpc.c
> > index 7fc8f82280ac..21958c3b0c28 100644
> > --- a/drivers/platform/chrome/cros_ec_lpc.c
> > +++ b/drivers/platform/chrome/cros_ec_lpc.c
> > @@ -21,6 +21,7 @@
> > #include <linux/platform_data/cros_ec_proto.h>
> > #include <linux/platform_device.h>
> > #include <linux/printk.h>
> > +#include <linux/reboot.h>
> > #include <linux/suspend.h>
> >
> > #include "cros_ec.h"
> > @@ -332,6 +333,15 @@ static void cros_ec_lpc_acpi_notify(acpi_handle device, u32 value, void *data)
> >
> > if (value == ACPI_NOTIFY_DEVICE_WAKE)
> > pm_system_wakeup();
> > +
> > + if (value == ACPI_NOTIFY_CROS_EC_PANIC) {
> > + dev_err(ec_dev->dev,
> > + "CrOS EC Panic Reported. Shutdown is imminent!");
> > + blocking_notifier_call_chain(&ec_dev->panic_notifier, 0,
> > + ec_dev);
> > + /* Begin orderly shutdown. Force shutdown after 1 second. */
> > + hw_protection_shutdown("CrOS EC Panic", 1000);
>
> I feel like this patch is doing 2 things: pulling the logs, and then
> starting a shutdown.
> This should be split into 2 patches.

Ack. Will do.

>
> > + }
> > }
>
> Line limits are now 100 chars[1], so most of these lines can fit on 1 line.
>
> >
> > static int cros_ec_lpc_probe(struct platform_device *pdev)
> > diff --git a/include/linux/platform_data/cros_ec_proto.h b/include/linux/platform_data/cros_ec_proto.h
> > index e43107e0bee1..1c4487271836 100644
> > --- a/include/linux/platform_data/cros_ec_proto.h
> > +++ b/include/linux/platform_data/cros_ec_proto.h
> > @@ -41,6 +41,13 @@
> > #define EC_MAX_REQUEST_OVERHEAD 1
> > #define EC_MAX_RESPONSE_OVERHEAD 32
> >
> > +/*
> > + * EC panic is not covered by the standard (0-F) ACPI notify values.
> > + * Arbitrarily choosing B0 to notify ec panic, which is in the 84-BF
> > + * device specific ACPI notify range.
> > + */
> > +#define ACPI_NOTIFY_CROS_EC_PANIC 0xB0
>
> Can you provide a link (either in the commit, or here in the comment)
> to the coreboot/BIOS code which uses this value? I feel this should
> be documented in some form that correlates the caller and the callee.

Link: https://chromium-review.googlesource.com/c/chromiumos/third_party/coreboot/+/4023535

>
> > +
> > /*
> > * Command interface between EC and AP, for LPC, I2C and SPI interfaces.
> > */
> > @@ -176,6 +183,8 @@ struct cros_ec_device {
> > /* The platform devices used by the mfd driver */
> > struct platform_device *ec;
> > struct platform_device *pd;
> > +
> > + struct blocking_notifier_head panic_notifier;
>
> Any reason we cannot use the existing event_notifier (with value argument)?
> It's a system panic, so I doubt that computational overhead for other
> notifier block
> listeners is a concern.

The value field is already being used for "queued_during_suspend" in
event_notifier.
>
> BR,
>
> -Prashant
>
> [1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=bdc48fa11e46f867ea4d75fa59ee87a7f48be144

\
 
 \ /
  Last update: 2023-03-26 23:17    [W:0.094 / U:0.048 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site