2019-12-15 18:29:53

by Aditya Pakki

Subject: [PATCH] tpm/ppi: replace assertion code with recovery in tpm_eval_dsm

In tpm_eval_dsm, BUG_ON on ppi_handle is used as an assertion.
By returning NULL to the callers, instead of crashing, the error
can be better handled.

Signed-off-by: Aditya Pakki <[email protected]>
---
drivers/char/tpm/tpm_ppi.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/char/tpm/tpm_ppi.c b/drivers/char/tpm/tpm_ppi.c
index b2dab941cb7f..4b6f6a9c0b48 100644
--- a/drivers/char/tpm/tpm_ppi.c
+++ b/drivers/char/tpm/tpm_ppi.c
@@ -42,7 +42,9 @@ static inline union acpi_object *
tpm_eval_dsm(acpi_handle ppi_handle, int func, acpi_object_type type,
union acpi_object *argv4, u64 rev)
{
- BUG_ON(!ppi_handle);
+ if (!ppi_handle)
+ return NULL;
+
return acpi_evaluate_dsm_typed(ppi_handle, &tpm_ppi_guid,
rev, func, argv4, type);
}
--
2.20.1

2019-12-19 00:05:00

by Jarkko Sakkinen

Subject: Re: [PATCH] tpm/ppi: replace assertion code with recovery in tpm_eval_dsm

On Wed, 2019-12-18 at 09:45 -0400, Jason Gunthorpe wrote:
> On Sun, Dec 15, 2019 at 12:23:14PM -0600, Aditya Pakki wrote:
> > In tpm_eval_dsm, BUG_ON on ppi_handle is used as an assertion.
> > By returning NULL to the callers, instead of crashing, the error
> > can be better handled.
> >
> > Signed-off-by: Aditya Pakki <[email protected]>
> > drivers/char/tpm/tpm_ppi.c | 4 +++-
> > 1 file changed, 3 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/char/tpm/tpm_ppi.c b/drivers/char/tpm/tpm_ppi.c
> > index b2dab941cb7f..4b6f6a9c0b48 100644
> > +++ b/drivers/char/tpm/tpm_ppi.c
> > @@ -42,7 +42,9 @@ static inline union acpi_object *
> > tpm_eval_dsm(acpi_handle ppi_handle, int func, acpi_object_type type,
> > union acpi_object *argv4, u64 rev)
> > {
> > - BUG_ON(!ppi_handle);
> > + if (!ppi_handle)
> > + return NULL;
>
> If it can't happen the confusing if should either be omitted entirely
> or written as
>
> if (WARN_ON(!ppi_handle))
> return NULL;
>
> Leaving it as apparently operational code just creates confusion for
> the reader that now has the task to figure out why ppi_handle can be
> null.
>
> I favour not including tests for impossible conditions. The kernel
> will crash immediately if ppi_handle is null anyhow.
>
> Jason

Absolutely should be changed WARN_ON() as it never should happen. I'll
update the patch before sending PR to Linus since I have it already
applied.

Thanks Jason for the remark!

/Jarkko