[ofa-general] Re: [PATCH] IB/ipoib - partial error clean up unmaps wrong address

[Michael S. Tsirkin]

> Quoting Roland Dreier <rdreier at cisco.com>:
> Subject: Re: [PATCH] IB/ipoib - partial error clean up unmaps wrong address
> 
>  > -	for (; i >= 0; --i)
>  > -		ib_dma_unmap_single(priv->ca, mapping[i + 1], PAGE_SIZE, DMA_FROM_DEVICE);
>  > +	for (; i > 0; --i)
>  > +		ib_dma_unmap_single(priv->ca, mapping[i], PAGE_SIZE, DMA_FROM_DEVICE);
> 
> Michael -- this looks rather clearly correct to me.  Any objection to
> applying it?

Yes, the patch looks clearly correct to me.
I recently saw a crash on one system which looks like it could be related:

Call Trace:<IRQ> <ffffffffa01ff3f0>{:ib_ipoib:ipoib_cm_alloc_rx_skb+796}
       <ffffffffa01fefdf>{:ib_ipoib:ipoib_cm_post_receive+119}
       <ffffffff801d5729>{selinux_socket_sock_rcv_skb+530}
       <ffffffffa01ffc04>{:ib_ipoib:ipoib_cm_handle_rx_wc+477}
       <ffffffff802aa3af>{sock_def_readable+16} <ffffffff802e9b29>{udp_queue_rcv_skb+827}
       <ffffffff802ea016>{udp_rcv+1153} <ffffffffa01fafa0>{:ib_ipoib:ipoib_ib_completion+144}
       <ffffffff801321e3>{activate_task+124} <ffffffffa023a69f>{:ib_mthca:mthca_eq_int+215}
       <ffffffffa023adea>{:ib_mthca:mthca_arbel_interrupt+56}
       <ffffffff80112f4a>{handle_IRQ_event+41} <ffffffff801131c4>{do_IRQ+197}
       <ffffffff80110833>{ret_from_intr+0}  <EOI>

I hoped to get this patch stress tested and report whether it helps before Acking.
But it seems this won't happen soon since that system is busy.

It's probably best to apply this patch.

Acked-by: Michael S. Tsirkin <mst at dev.mellanox.co.il>


-- 
MST