linux-kernel - Re: [PATCH] scsi: Let scsi_execute_cmd() mark args->sshdr as invalid

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: <85a7dc28-74ec-f4d6-b5c3-ca456ce9d380@suse.com>
Date:   Thu, 11 May 2023 15:17:11 +0200
From:   Juergen Gross <jgross@...e.com>
To:     Martin Wilck <mwilck@...e.com>, linux-kernel@...r.kernel.org,
        linux-scsi@...r.kernel.org
Cc:     "James E.J. Bottomley" <jejb@...ux.ibm.com>,
        "Martin K. Petersen" <martin.petersen@...cle.com>,
        stable@...r.kernel.org
Subject: Re: [PATCH] scsi: Let scsi_execute_cmd() mark args->sshdr as invalid

On 11.05.23 15:10, Martin Wilck wrote:
> On Thu, 2023-05-11 at 14:34 +0200, Juergen Gross wrote:
>> Some callers of scsi_execute_cmd() (like e.g. sd_spinup_disk()) are
>> passing an uninitialized struct sshdr and don't look at the return
>> value of scsi_execute_cmd() before looking at the contents of that
>> struct.
>>
>> This can result in false positives when looking for specific error
>> conditions.
>>
>> In order to fix that let scsi_execute_cmd() zero sshdr-
>>> response_code,
>> resulting in scsi_sense_valid() returning false.
>>
>> Cc: stable@...r.kernel.org
>> Fixes: 3949e2f04262 ("scsi: simplify scsi_execute_req_flags")
>> Signed-off-by: Juergen Gross <jgross@...e.com>
>> ---
>> I'm not aware of any real error having happened due to this problem,
>> but I thought it should be fixed anyway.
>> I _think_ 3949e2f04262 was introducing the problem, but I'm not 100%
>> sure it is really the commit to be blamed.
>> ---
>>   drivers/scsi/scsi_lib.c | 14 ++++++++++----
>>   1 file changed, 10 insertions(+), 4 deletions(-)
> 
> One nitpick below, otherwise it looks good to me.
> 
>>
>> diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
>> index b7c569a42aa4..923336620bff 100644
>> --- a/drivers/scsi/scsi_lib.c
>> +++ b/drivers/scsi/scsi_lib.c
>> @@ -209,11 +209,17 @@ int scsi_execute_cmd(struct scsi_device *sdev,
>> const unsigned char *cmd,
>>          struct scsi_cmnd *scmd;
>>          int ret;
>>   
>> -       if (!args)
>> +       if (!args) {
>>                  args = &default_args;
>> -       else if (WARN_ON_ONCE(args->sense &&
>> -                             args->sense_len !=
>> SCSI_SENSE_BUFFERSIZE))
>> -               return -EINVAL;
>> +       } else {
>> +               /* Mark sense data to be invalid. */
>> +               if (args->sshdr)
>> +                       args->sshdr->response_code = 0;
> 
> We know for certain that sizeof(*sshdr) is 8 bytes, and will most
> probably remain so. Thus
> 
>      memset(sshdr, 0, sizeof(*sshdr))
> 
> would result in more efficient code.

I fail to see why zeroing a single byte would be less efficient than zeroing
a possibly unaligned 8-byte area.


Juergen

Download attachment "OpenPGP_0xB0DE9DD628BF132F.asc" of type "application/pgp-keys" (3099 bytes)

Download attachment "OpenPGP_signature" of type "application/pgp-signature" (496 bytes)