Re: [PATCH 4/4] xen/scsifront: harden driver against malicious backend

From: Juergen Gross <jgross@suse.com>
To: Boris Ostrovsky <boris.ostrovsky@oracle.com>,
	xen-devel@lists.xenproject.org, linux-scsi@vger.kernel.org,
	linux-kernel@vger.kernel.org
Cc: Stefano Stabellini <sstabellini@kernel.org>,
	"James E.J. Bottomley" <jejb@linux.ibm.com>,
	"Martin K. Petersen" <martin.petersen@oracle.com>
Subject: Re: [PATCH 4/4] xen/scsifront: harden driver against malicious backend
Date: Thu, 21 Apr 2022 17:29:27 +0200	[thread overview]
Message-ID: <15c89971-aefb-e71c-3798-c2e3820601f8@suse.com> (raw)
In-Reply-To: <9d8c453b-7147-d80c-3d4f-666a3b530929@suse.com>

[-- Attachment #1.1.1: Type: text/plain, Size: 1830 bytes --]

On 21.04.22 12:13, Juergen Gross wrote:
> On 20.04.22 18:13, Boris Ostrovsky wrote:
>> Just a couple of nits.
>>
>>
>> On 4/20/22 5:25 AM, Juergen Gross wrote:
>>> -static int scsifront_ring_drain(struct vscsifrnt_info *info)
>>> +static int scsifront_ring_drain(struct vscsifrnt_info *info,
>>> +                unsigned int *eoiflag)
>>>   {
>>> -    struct vscsiif_response *ring_rsp;
>>> +    struct vscsiif_response ring_rsp;
>>>       RING_IDX i, rp;
>>>       int more_to_do = 0;
>>> -    rp = info->ring.sring->rsp_prod;
>>> -    rmb();    /* ordering required respective to dom0 */
>>> +    rp = READ_ONCE(info->ring.sring->rsp_prod);
>>> +    virt_rmb();    /* ordering required respective to backend */
>>> +    if (RING_RESPONSE_PROD_OVERFLOW(&info->ring, rp)) {
>>> +        scsifront_set_error(info, "illegal number of responses");
>>
>>
>> In net and block drivers we report number of such responses. (But not in usb)
> I'm not sure the specific value is of any interest.
> 
>>> +        return 0;
>>> +    }
>>>       for (i = info->ring.rsp_cons; i != rp; i++) {
>>> -        ring_rsp = RING_GET_RESPONSE(&info->ring, i);
>>> -        scsifront_do_response(info, ring_rsp);
>>> +        RING_COPY_RESPONSE(&info->ring, i, &ring_rsp);
>>> +        scsifront_do_response(info, &ring_rsp);
>>> +        if (info->host_active == STATE_ERROR)
>>> +            return 0;
>>> +        *eoiflag = 0;
>>
>>
>> *eoiflags &= ~XEN_EOI_FLAG_SPURIOUS; ?
> 
> Yes, probably better.
> 
>> We also use eoi_flags name in other instances in this file.
> 
> I'll unify the name.

Oh, eoi_flags is used in the backend driver. So nothing to unify.

Juergen

[-- Attachment #1.1.2: OpenPGP public key --]
[-- Type: application/pgp-keys, Size: 3149 bytes --]

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 495 bytes --]