Re: [PATCH v6 3/9] tpm: replace dynamically allocated bios_dir with a static array

[Jarkko Sakkinen <jarkko.sakkinen-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>]

On Tue, Nov 22, 2016 at 09:58:56AM -0700, Jason Gunthorpe wrote:
> On Tue, Nov 22, 2016 at 01:23:33PM +0200, Jarkko Sakkinen wrote:
> > On Mon, Nov 14, 2016 at 05:00:50AM -0500, Nayna Jain wrote:
> > > This commit is based on a commit by Nayna Jain. Replaced dynamically
> > > allocated bios_dir with a static array as the size is always constant.
> > > 
> > > Suggested-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
> > > Signed-off-by: Nayna Jain <nayna-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>
> > > Signed-off-by: Jarkko Sakkinen <jarkko.sakkinen-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
> > 
> > This commit remains unreviewed and tested. I'm in the author role here
> > so I cannot help with this. If that does not happen soon I cannot put
> > this into the pull request.
> 
> Nayna must have tested it, looks OK to me..
> 
> > > +err:
> > > +	chip->bios_dir[cnt] = NULL;
> > > +	tpm_bios_log_teardown(chip);
> > > +	return -EIO;
> 
> Except that return should ideally be PTR_ERR(chip->bios_dir[cnt])
> 
> .. and we still set ERR_PTR into bios_dir in the ENODEV case, so the
> overall series is still broken if securityfs is compiled out.
> 
> Lets fix this all like this - which is a good enough reason to leave the
> ENODEV detect alone - just squash this into your patch:

That was a great catch. Thank you.

> diff --git a/drivers/char/tpm/tpm_eventlog.c b/drivers/char/tpm/tpm_eventlog.c
> index 2a15b866ac257a..11bb1138a8282e 100644
> --- a/drivers/char/tpm/tpm_eventlog.c
> +++ b/drivers/char/tpm/tpm_eventlog.c
> @@ -356,15 +356,6 @@ static const struct file_operations tpm_bios_measurements_ops = {
>  	.release = tpm_bios_measurements_release,
>  };
>  
> -static int is_bad(void *p)
> -{
> -	if (!p)
> -		return 1;
> -	if (IS_ERR(p) && (PTR_ERR(p) != -ENODEV))
> -		return 1;
> -	return 0;
> -}
> -

This function is only confusing indirection anyway. Does not serve
really any justifiable purpose.

>  static int tpm_read_log(struct tpm_chip *chip)
>  {
>  	int rc;
> @@ -390,7 +381,8 @@ static int tpm_read_log(struct tpm_chip *chip)
>   * If an event log is found then the securityfs files are setup to
>   * export it to userspace, otherwise nothing is done.
>   *
> - * Returns -ENODEV if the firmware has no event log.
> + * Returns -ENODEV if the firmware has no event log or securityfs is not
> + * supported.
>   */
>  int tpm_bios_log_setup(struct tpm_chip *chip)
>  {
> @@ -407,7 +399,10 @@ int tpm_bios_log_setup(struct tpm_chip *chip)
>  
>  	cnt = 0;
>  	chip->bios_dir[cnt] = securityfs_create_dir(name, NULL);
> -	if (is_bad(chip->bios_dir[cnt]))
> +	/* NOTE: securityfs_create_dir can return ENODEV if securityfs is
> +	 * compiled out. The caller should ignore the ENODEV return code.
> +	 */
> +	if (IS_ERR(chip->bios_dir[cnt]))
>  		goto err;
>  	cnt++;
>  
> @@ -419,7 +414,7 @@ int tpm_bios_log_setup(struct tpm_chip *chip)
>  				   0440, chip->bios_dir[0],
>  				   (void *)&chip->bin_log_seqops,
>  				   &tpm_bios_measurements_ops);
> -	if (is_bad(chip->bios_dir[cnt]))
> +	if (IS_ERR(chip->bios_dir[cnt]))
>  		goto err;
>  	cnt++;
>  
> @@ -431,16 +426,17 @@ int tpm_bios_log_setup(struct tpm_chip *chip)
>  				   0440, chip->bios_dir[0],
>  				   (void *)&chip->ascii_log_seqops,
>  				   &tpm_bios_measurements_ops);
> -	if (is_bad(chip->bios_dir[cnt]))
> +	if (IS_ERR(chip->bios_dir[cnt]))
>  		goto err;
>  	cnt++;
>  
>  	return 0;
>  
>  err:
> +	rc = PTR_ERR(chip->bios_dir[cnt]);
>  	chip->bios_dir[cnt] = NULL;
>  	tpm_bios_log_teardown(chip);
> -	return -EIO;
> +	return rc;
>  }
>  
>  void tpm_bios_log_teardown(struct tpm_chip *chip)

I manually added the changes to:

  tpm: replace dynamically allocated bios_dir with a static array

The code was changed so radically after that patch so it was the
cleanest way.

I need to declare 'rc' in that patch.  I changed also this "int rc = 0;"
to "int rc;" as it does not need to be initialized in the declaration.
This affects:

  tpm: have event log use the tpm_chip

And finally I needed the update this patch too to accomadate the doc
change:

  tpm: Fix handling of missing event log

Could you check through those patches that I didn't blow things up,
which could easily happen given that I needed to update three patches
and give your final reviewed-by if it looks good to you?

In the meanwhile I'll start running sparse, coccicheck etc. for the
release content.

/Jarkko

------------------------------------------------------------------------------