linux-kernel - Re: [PATCH] tpm: add support for partial reads

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 for Android: free password hash cracker in your pocket

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

Date:   Thu, 19 Jul 2018 14:01:16 -0700
From:   Tadeusz Struk <tadeusz.struk@...el.com>
To:     James Bottomley <James.Bottomley@...senPartnership.com>,
        jarkko.sakkinen@...ux.intel.com
Cc:     jgg@...pe.ca, linux-integrity@...r.kernel.org,
        linux-security-module@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] tpm: add support for partial reads

On 07/19/2018 01:27 PM, James Bottomley wrote:
> On Thu, 2018-07-19 at 13:12 -0700, Tadeusz Struk wrote:
>> On 07/19/2018 12:52 PM, James Bottomley wrote:
>>> The ABI break is the error case as I outlined above.  We can't
>>> assume everyone uses the current interface without getting an error
>>> and one error and your hosed is a nasty failure case to change the
>>> interface to. 
>>
>> Well, if there is a broken application out there that doesn't work
>> today it will not work after this change neither.
> 
> It doesn't have to be broken ... it could be using EFAULT to probe the
> buffer size for instance.  That's the point of not breaking the ABI:
> you don't second guess what applications are doing.
> 

Looking at the existing implementation again:
https://git.kernel.org/pub/scm/linux/kernel/git/jmorris/linux-security.git/tree/drivers/char/tpm/tpm-dev-common.c?h=next-tpm#n56

EFAULT is returned only if the copy_to_user() fails.
So today, if an application wants read 1 byte of a response, and provides
1 byte buffer for it, then only 1 byte of the response will be copied,
no error code will be returned, and the rest of the response will be gone.
I don't really see how and why would anyone use EFAULT err to probe for
the buffer size. That would really be a broken application.

Thanks,
-- 
Tadeusz