Re: [Xen-devel] [RFC PATCH 3/3] xen/netback: Fix grant copy across page boundary with KASAN

[Sergey Dyasli <sergey.dyasli@citrix.com>]

On 17/12/2019 15:14, Durrant, Paul wrote:
>> -----Original Message-----
>> From: Xen-devel <xen-devel-bounces@lists.xenproject.org> On Behalf Of
>> Sergey Dyasli
>> Sent: 17 December 2019 14:08
>> To: xen-devel@lists.xen.org; kasan-dev@googlegroups.com; linux-
>> kernel@vger.kernel.org
>> Cc: Juergen Gross <jgross@suse.com>; Sergey Dyasli
>> <sergey.dyasli@citrix.com>; Stefano Stabellini <sstabellini@kernel.org>;
>> George Dunlap <george.dunlap@citrix.com>; Ross Lagerwall
>> <ross.lagerwall@citrix.com>; Alexander Potapenko <glider@google.com>;
>> Andrey Ryabinin <aryabinin@virtuozzo.com>; Boris Ostrovsky
>> <boris.ostrovsky@oracle.com>; Dmitry Vyukov <dvyukov@google.com>
>> Subject: [Xen-devel] [RFC PATCH 3/3] xen/netback: Fix grant copy across
>> page boundary with KASAN
>>
>> From: Ross Lagerwall <ross.lagerwall@citrix.com>
>>
>> When KASAN (or SLUB_DEBUG) is turned on, the normal expectation that
>> allocations are aligned to the next power of 2 of the size does not
>> hold. Therefore, handle grant copies that cross page boundaries.
>>
>> Signed-off-by: Ross Lagerwall <ross.lagerwall@citrix.com>
>> Signed-off-by: Sergey Dyasli <sergey.dyasli@citrix.com>
>
> Would have been nice to cc netback maintainers...

Sorry, I'll try to be more careful next time.

>
>> ---
>>  drivers/net/xen-netback/common.h  |  2 +-
>>  drivers/net/xen-netback/netback.c | 55 ++++++++++++++++++++++++-------
>>  2 files changed, 45 insertions(+), 12 deletions(-)
>>
>> diff --git a/drivers/net/xen-netback/common.h b/drivers/net/xen-
>> netback/common.h
>> index 05847eb91a1b..e57684415edd 100644
>> --- a/drivers/net/xen-netback/common.h
>> +++ b/drivers/net/xen-netback/common.h
>> @@ -155,7 +155,7 @@ struct xenvif_queue { /* Per-queue data for xenvif */
>>  	struct pending_tx_info pending_tx_info[MAX_PENDING_REQS];
>>  	grant_handle_t grant_tx_handle[MAX_PENDING_REQS];
>>
>> -	struct gnttab_copy tx_copy_ops[MAX_PENDING_REQS];
>> +	struct gnttab_copy tx_copy_ops[MAX_PENDING_REQS * 2];
>>  	struct gnttab_map_grant_ref tx_map_ops[MAX_PENDING_REQS];
>>  	struct gnttab_unmap_grant_ref tx_unmap_ops[MAX_PENDING_REQS];
>>  	/* passed to gnttab_[un]map_refs with pages under (un)mapping */
>> diff --git a/drivers/net/xen-netback/netback.c b/drivers/net/xen-
>> netback/netback.c
>> index 0020b2e8c279..1541b6e0cc62 100644
>> --- a/drivers/net/xen-netback/netback.c
>> +++ b/drivers/net/xen-netback/netback.c
>> @@ -320,6 +320,7 @@ static int xenvif_count_requests(struct xenvif_queue
>> *queue,
>>
>>  struct xenvif_tx_cb {
>>  	u16 pending_idx;
>> +	u8 copies;
>>  };
>
> I know we're a way off the limit (48 bytes) but I wonder if we ought to have a compile time check here that we're not overflowing skb->cb.

I will add a BUILD_BUG_ON()

>
>>
>>  #define XENVIF_TX_CB(skb) ((struct xenvif_tx_cb *)(skb)->cb)
>> @@ -439,6 +440,7 @@ static int xenvif_tx_check_gop(struct xenvif_queue
>> *queue,
>>  {
>>  	struct gnttab_map_grant_ref *gop_map = *gopp_map;
>>  	u16 pending_idx = XENVIF_TX_CB(skb)->pending_idx;
>> +	u8 copies = XENVIF_TX_CB(skb)->copies;
>>  	/* This always points to the shinfo of the skb being checked, which
>>  	 * could be either the first or the one on the frag_list
>>  	 */
>> @@ -450,23 +452,27 @@ static int xenvif_tx_check_gop(struct xenvif_queue
>> *queue,
>>  	int nr_frags = shinfo->nr_frags;
>>  	const bool sharedslot = nr_frags &&
>>  				frag_get_pending_idx(&shinfo->frags[0]) ==
>> pending_idx;
>> -	int i, err;
>> +	int i, err = 0;
>>
>> -	/* Check status of header. */
>> -	err = (*gopp_copy)->status;
>> -	if (unlikely(err)) {
>> -		if (net_ratelimit())
>> -			netdev_dbg(queue->vif->dev,
>> +	while (copies) {
>> +		/* Check status of header. */
>> +		int newerr = (*gopp_copy)->status;
>> +		if (unlikely(newerr)) {
>> +			if (net_ratelimit())
>> +				netdev_dbg(queue->vif->dev,
>>  				   "Grant copy of header failed! status: %d
>> pending_idx: %u ref: %u\n",
>>  				   (*gopp_copy)->status,
>>  				   pending_idx,
>>  				   (*gopp_copy)->source.u.ref);
>> -		/* The first frag might still have this slot mapped */
>> -		if (!sharedslot)
>> -			xenvif_idx_release(queue, pending_idx,
>> -					   XEN_NETIF_RSP_ERROR);
>> +			/* The first frag might still have this slot mapped */
>> +			if (!sharedslot && !err)
>> +				xenvif_idx_release(queue, pending_idx,
>> +						   XEN_NETIF_RSP_ERROR);
>
> Can't this be done after the loop, if there is an accumulated err? I think it would make the code slightly neater.

Looks like xenvif_idx_release() indeed wants to be just after the loop.

>
>> +			err = newerr;
>> +		}
>> +		(*gopp_copy)++;
>> +		copies--;
>>  	}
>> -	(*gopp_copy)++;
>>
>>  check_frags:
>>  	for (i = 0; i < nr_frags; i++, gop_map++) {
>> @@ -910,6 +916,7 @@ static void xenvif_tx_build_gops(struct xenvif_queue
>> *queue,
>>  			xenvif_tx_err(queue, &txreq, extra_count, idx);
>>  			break;
>>  		}
>> +		XENVIF_TX_CB(skb)->copies = 0;
>>
>>  		skb_shinfo(skb)->nr_frags = ret;
>>  		if (data_len < txreq.size)
>> @@ -933,6 +940,7 @@ static void xenvif_tx_build_gops(struct xenvif_queue
>> *queue,
>>  						   "Can't allocate the frag_list
>> skb.\n");
>>  				break;
>>  			}
>> +			XENVIF_TX_CB(nskb)->copies = 0;
>>  		}
>>
>>  		if (extras[XEN_NETIF_EXTRA_TYPE_GSO - 1].type) {
>> @@ -990,6 +998,31 @@ static void xenvif_tx_build_gops(struct xenvif_queue
>> *queue,
>>
>>  		queue->tx_copy_ops[*copy_ops].len = data_len;
>
> If offset_in_page(skb->data)+ data_len can exceed XEN_PAGE_SIZE, does this not need to be truncated?

It is performed as the first thing inside the if condition below.

>>  		queue->tx_copy_ops[*copy_ops].flags = GNTCOPY_source_gref;
>> +		XENVIF_TX_CB(skb)->copies++;
>> +
>> +		if (offset_in_page(skb->data) + data_len > XEN_PAGE_SIZE) {
>> +			unsigned int extra_len = offset_in_page(skb->data) +
>> +					     data_len - XEN_PAGE_SIZE;
>> +
>> +			queue->tx_copy_ops[*copy_ops].len -= extra_len;
>> +			(*copy_ops)++;
>> +
>> +			queue->tx_copy_ops[*copy_ops].source.u.ref = txreq.gref;
>> +			queue->tx_copy_ops[*copy_ops].source.domid =
>> +				queue->vif->domid;
>> +			queue->tx_copy_ops[*copy_ops].source.offset =
>> +				txreq.offset + data_len - extra_len;
>> +
>> +			queue->tx_copy_ops[*copy_ops].dest.u.gmfn =
>> +				virt_to_gfn(skb->data + data_len - extra_len);
>> +			queue->tx_copy_ops[*copy_ops].dest.domid = DOMID_SELF;
>> +			queue->tx_copy_ops[*copy_ops].dest.offset = 0;
>> +
>> +			queue->tx_copy_ops[*copy_ops].len = extra_len;
>> +			queue->tx_copy_ops[*copy_ops].flags =
>> GNTCOPY_source_gref;
>> +
>> +			XENVIF_TX_CB(skb)->copies++;
>> +		}
>>
>>  		(*copy_ops)++;
>>
>> --
>> 2.17.1
>>
>>
>> _______________________________________________
>> Xen-devel mailing list
>> Xen-devel@lists.xenproject.org
>> https://lists.xenproject.org/mailman/listinfo/xen-devel