Re: [PATCH v3 2/2] IB/mad: Use IDR for agent IDs

From: Jason Gunthorpe
Date: Mon Jun 18 2018 - 13:26:13 EST


On Wed, Jun 13, 2018 at 05:34:03AM -0700, willy@xxxxxxxxxxxxx wrote:
> Allocate agent IDs from a global IDR instead of an atomic variable.
> This eliminates the possibility of reusing an ID which is already in
> use after 4 billion registrations. We limit the assigned ID to be less
> than 2^24 as the mlx4 driver uses the most significant byte of the agent
> ID to store the slave number. Users unlucky enough to see a collision
> between agent numbers and slave numbers see messages like:
>
> mlx4_ib: egress mad has non-null tid msb:1 class:4 slave:0
>
> and the MAD layer stops working.
>
> We look up the agent under protection of the RCU lock, which means we
> have to free the agent using kfree_rcu, and only increment the reference
> counter if it is not 0.
>
> Signed-off-by: Matthew Wilcox <mawilcox@xxxxxxxxxxxxx>
> Reported-by: Hans Westgaard Ry <hans.westgaard.ry@xxxxxxxxxx>
> Acked-by: Jack Morgenstein <jackm@xxxxxxxxxxxxxxxxxx>
> Tested-by: Jack Morgenstein <jackm@xxxxxxxxxxxxxxxxxx>
> ---
> drivers/infiniband/core/mad.c | 83 ++++++++++++++++++------------
> drivers/infiniband/core/mad_priv.h | 7 +--
> 2 files changed, 55 insertions(+), 35 deletions(-)

Applied to for-next, thanks everyone

Jason