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]
Message-ID: <08ce58ab-3513-5d98-16a5-b197276f6bce@kunbus.com>
Date:   Tue, 9 Feb 2021 12:52:17 +0100
From:   Lino Sanfilippo <l.sanfilippo@...bus.com>
To:     Jason Gunthorpe <jgg@...pe.ca>,
        James Bottomley <James.Bottomley@...senPartnership.com>
Cc:     Jarkko Sakkinen <jarkko@...nel.org>,
        Lino Sanfilippo <LinoSanfilippo@....de>, peterhuewe@....de,
        stefanb@...ux.vnet.ibm.com, stable@...r.kernel.org,
        linux-integrity@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v3 2/2] tpm: in tpm2_del_space check if ops pointer is
 still valid

Hi Jason,

On 05.02.21 18:25, Jason Gunthorpe wrote:
> On Fri, Feb 05, 2021 at 08:48:11AM -0800, James Bottomley wrote:
>>> Thanks for pointing this out. I'd strongly support Jason's proposal:
>>>
>>> https://lore.kernel.org/linux-integrity/20201215175624.GG5487@ziepe.ca/
>>>
>>> It's the best long-term way to fix this.
>>
>> Really, no it's not.  It introduces extra mechanism we don't need.
> 
>> To recap the issue: character devices already have an automatic
>> mechanism which holds a reference to the struct device while the
>> character node is open so the default is to release resources on final
>> put of the struct device.
> 
> The refcount on the struct device only keeps the memory alive, it
> doesn't say anything about the ops. We still need to lock and check
> the ops each and every time they are used.
> 
> The fact cdev goes all the way till fput means we don't need the extra
> get/put I suggested to Lino at all.
> 
>> The practical consequence of this model is that if you allocate a chip
>> structure with tpm_chip_alloc() you have to release it again by doing a
>> put of *both* devices.
> 
> The final put of the devs should be directly after the
> cdev_device_del(), not in a devm. This became all confused because the
> devs was created during alloc, not register. Having a device that is
> initialized but will never be added is weird.
> 
> See sketch below.
> 
>> Stefan noticed the latter, so we got the bogus patch 8979b02aaf1d
>> ("tpm: Fix reference count to main device") applied which simply breaks
>> the master/slave model by not taking a reference on the master for the
>> slave.  I'm not sure why I didn't notice the problem with this fix at
>> the time, but attention must have been elsewhere.
> 
> Well, this is sort of OK because we never use the devs in TPM1, so we
> end up freeing the chip with a positive refcount on the devs, which is
> weird but not a functional bug.
> 
> Jason
> 
> diff --git a/drivers/char/tpm/tpm-chip.c b/drivers/char/tpm/tpm-chip.c
> index ddaeceb7e10910..e07193a0dd4438 100644
> --- a/drivers/char/tpm/tpm-chip.c
> +++ b/drivers/char/tpm/tpm-chip.c
> @@ -344,7 +344,6 @@ struct tpm_chip *tpm_chip_alloc(struct device *pdev,
>  	chip->dev_num = rc;
>  
>  	device_initialize(&chip->dev);
> -	device_initialize(&chip->devs);
>  
>  	chip->dev.class = tpm_class;
>  	chip->dev.class->shutdown_pre = tpm_class_shutdown;
> @@ -352,29 +351,12 @@ struct tpm_chip *tpm_chip_alloc(struct device *pdev,
>  	chip->dev.parent = pdev;
>  	chip->dev.groups = chip->groups;
>  
> -	chip->devs.parent = pdev;
> -	chip->devs.class = tpmrm_class;
> -	chip->devs.release = tpm_devs_release;
> -	/* get extra reference on main device to hold on
> -	 * behalf of devs.  This holds the chip structure
> -	 * while cdevs is in use.  The corresponding put
> -	 * is in the tpm_devs_release (TPM2 only)
> -	 */
> -	if (chip->flags & TPM_CHIP_FLAG_TPM2)
> -		get_device(&chip->dev);
> -
>  	if (chip->dev_num == 0)
>  		chip->dev.devt = MKDEV(MISC_MAJOR, TPM_MINOR);
>  	else
>  		chip->dev.devt = MKDEV(MAJOR(tpm_devt), chip->dev_num);
>  
> -	chip->devs.devt =
> -		MKDEV(MAJOR(tpm_devt), chip->dev_num + TPM_NUM_DEVICES);
> -
>  	rc = dev_set_name(&chip->dev, "tpm%d", chip->dev_num);
> -	if (rc)
> -		goto out;
> -	rc = dev_set_name(&chip->devs, "tpmrm%d", chip->dev_num);
>  	if (rc)
>  		goto out;
>  
> @@ -382,9 +364,7 @@ struct tpm_chip *tpm_chip_alloc(struct device *pdev,
>  		chip->flags |= TPM_CHIP_FLAG_VIRTUAL;
>  
>  	cdev_init(&chip->cdev, &tpm_fops);
> -	cdev_init(&chip->cdevs, &tpmrm_fops);
>  	chip->cdev.owner = THIS_MODULE;
> -	chip->cdevs.owner = THIS_MODULE;
>  
>  	rc = tpm2_init_space(&chip->work_space, TPM2_SPACE_BUFFER_SIZE);
>  	if (rc) {
> @@ -396,7 +376,6 @@ struct tpm_chip *tpm_chip_alloc(struct device *pdev,
>  	return chip;
>  
>  out:
> -	put_device(&chip->devs);
>  	put_device(&chip->dev);
>  	return ERR_PTR(rc);
>  }
> @@ -445,13 +424,33 @@ static int tpm_add_char_device(struct tpm_chip *chip)
>  	}
>  
>  	if (chip->flags & TPM_CHIP_FLAG_TPM2) {
> +		device_initialize(&chip->devs);
> +		chip->devs.parent = pdev;
> +		chip->devs.class = tpmrm_class;
> +		rc = dev_set_name(&chip->devs, "tpmrm%d", chip->dev_num);
> +		if (rc)
> +			goto out_put_devs;
> +
> +		/*
> +                 * get extra reference on main device to hold on behalf of devs.
> +                 * This holds the chip structure while cdevs is in use. The
> +		 * corresponding put is in the tpm_devs_release.
> +		 */
> +		get_device(&chip->dev);
> +		chip->devs.release = tpm_devs_release;
> +
> +		chip->devs.devt =
> +			MKDEV(MAJOR(tpm_devt), chip->dev_num + TPM_NUM_DEVICES);
> +		cdev_init(&chip->cdevs, &tpmrm_fops);
> +		chip->cdevs.owner = THIS_MODULE;
> +
>  		rc = cdev_device_add(&chip->cdevs, &chip->devs);
>  		if (rc) {
>  			dev_err(&chip->devs,
>  				"unable to cdev_device_add() %s, major %d, minor %d, err=%d\n",
>  				dev_name(&chip->devs), MAJOR(chip->devs.devt),
>  				MINOR(chip->devs.devt), rc);
> -			return rc;
> +			goto out_put_devs;
>  		}
>  	}
>  
> @@ -460,6 +459,10 @@ static int tpm_add_char_device(struct tpm_chip *chip)
>  	idr_replace(&dev_nums_idr, chip, chip->dev_num);
>  	mutex_unlock(&idr_lock);
>  
> +out_put_devs:
> +	put_device(&chip->devs);
> +out_del_dev:
> +	cdev_device_del(&chip->cdev);
>  	return rc;
>  }
>  
> @@ -640,8 +643,10 @@ void tpm_chip_unregister(struct tpm_chip *chip)
>  	if (IS_ENABLED(CONFIG_HW_RANDOM_TPM))
>  		hwrng_unregister(&chip->hwrng);
>  	tpm_bios_log_teardown(chip);
> -	if (chip->flags & TPM_CHIP_FLAG_TPM2)
> +	if (chip->flags & TPM_CHIP_FLAG_TPM2) {
>  		cdev_device_del(&chip->cdevs, &chip->devs);
> +		put_device(&chip->devs);
> +	}
>  	tpm_del_char_device(chip);
>  }
>  EXPORT_SYMBOL_GPL(tpm_chip_unregister);
> 

I tested the solution you scetched and it fixes the issue for me. Will you send a (real) patch for this?

Best regards,
Lino

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ