linux-kernel - Re: [PATCH] scsi: ses: fix some risks of out of bound access

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] [day] [month] [year] [list]

Date:   Wed, 27 Mar 2019 10:21:09 +0800
From:   "jianchao.wang" <jianchao.w.wang@...cle.com>
To:     "Ewan D. Milne" <emilne@...hat.com>, jejb@...ux.ibm.com,
        martin.petersen@...cle.com
Cc:     linux-scsi@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] scsi: ses: fix some risks of out of bound access

Hi Ewan

Thanks for your kindly response.

On 3/26/19 9:06 PM, Ewan D. Milne wrote:
> See below.
> 
> On Mon, 2019-03-25 at 15:40 +0800, Jianchao Wang wrote:
>> We have some places with risk of accessing out of bound of the
>> buffer allocated from slab, even it could corrupt the memory.
>>
>> Signed-off-by: Jianchao Wang <jianchao.w.wang@...cle.com>
>> ---
>>  drivers/scsi/ses.c | 27 ++++++++++++++++++++++-----
>>  1 file changed, 22 insertions(+), 5 deletions(-)
>>
>> diff --git a/drivers/scsi/ses.c b/drivers/scsi/ses.c
>> index 0fc3922..42e6a1f 100644
>> --- a/drivers/scsi/ses.c
>> +++ b/drivers/scsi/ses.c
>> @@ -520,6 +520,7 @@ static void ses_enclosure_data_process(struct enclosure_device *edev,
>>  	struct ses_device *ses_dev = edev->scratch;
>>  	int types = ses_dev->page1_num_types;
>>  	unsigned char *hdr_buf = kzalloc(INIT_ALLOC_SIZE, GFP_KERNEL);
>> +	unsigned char *page1_end = ses_dev->page1 + ses_dev->page1_len;
>>  
>>  	if (!hdr_buf)
>>  		goto simple_populate;
>> @@ -556,6 +557,11 @@ static void ses_enclosure_data_process(struct enclosure_device *edev,
>>  	type_ptr = ses_dev->page1_types;
>>  	components = 0;
>>  	for (i = 0; i < types; i++, type_ptr += 4) {
>> +		if (type_ptr > page1_end - 2) {
> 
> I think "if (type_ptr + 1 >= page1_end)" would be more consistent.

Yes, indeed.

> 
>> +			sdev_printk(KERN_ERR, sdev, "Access out of bound of page1"
>> +					"%p page1_end %p\n", page1_end, type_ptr);
> 
> This message is not helpful for someone with a device reporting
> invalid enclosure data.  It should be more generic, like
> "Enclosure data too short" or "invalid" or something.  And, the
> actual pointer values are irrelevant, it all depends upon the
> contents of the buffer.
> 
>> +			break;
>> +		}
>>  		for (j = 0; j < type_ptr[1]; j++) {
>>  			char *name = NULL;
>>  			struct enclosure_component *ecomp;
>> @@ -566,10 +572,15 @@ static void ses_enclosure_data_process(struct enclosure_device *edev,
>>  				} else {
>>  					len = (desc_ptr[2] << 8) + desc_ptr[3];
>>  					desc_ptr += 4;
>> -					/* Add trailing zero - pushes into
>> -					 * reserved space */
>> -					desc_ptr[len] = '\0';
>> -					name = desc_ptr;
>> +					if (desc_ptr + len >= buf + page7_len) {
>> +						desc_ptr = NULL;
>> +					} else {
>> +
>> +						/* Add trailing zero - pushes into
>> +						 * reserved space */
>> +						desc_ptr[len] = '\0';
>> +						name = desc_ptr;
>> +					}
>>  				}
>>  			}
>>  			if (type_ptr[0] == ENCLOSURE_COMPONENT_DEVICE ||
>> @@ -693,7 +704,13 @@ static int ses_intf_add(struct device *cdev,
>>  	/* begin at the enclosure descriptor */
>>  	type_ptr = buf + 8;
>>  	/* skip all the enclosure descriptors */
>> -	for (i = 0; i < num_enclosures && type_ptr < buf + len; i++) {
>> +	for (i = 0; i < num_enclosures; i++) {
>> +		if (type_ptr >= buf + len) {
>> +			sdev_printk(KERN_ERR, sdev, "Overflow the buf len = %d\n", len);
> 
> See above, the message is unhelpful.  The actual problem is that
> the Enclosure data is too short or invalid.

Yes, maybe we should dump all of the information of the page1 here.

> 
>> +			err = -EINVAL;
>> +			goto err_free;
>> +		}
>> +
>>  		types += type_ptr[2];
>>  		type_ptr += type_ptr[3] + 4;
>>  	}
> 
> This will still potentially leave type_ptr past the end of the
> buffer in the subsequent code, though, right?

Yes, type_ptr[3] is accessing type_ptr + 3 which has be beyond our checking point.

> 
> This might fix the problem for your malfunctioning device, but
> does not look like it would handle the general case.
> 

Actually, I really don't know much about this ses driver. I sent out this bad patch
because I want to push things forward as there is a serious issue on our customer
side. 

Thanks
Jianchao