Re: [ewg] [PATCH]mlx4_ib XRC RCV: Fix mlx4_ib_reg_xrc_rcv_qp() locking

2011-02-16 Thread Jack Morgenstein
You are correct! Good catch. We will add this to OFED. (P.S., I would rather leave irqsave -- it is used everywhere else for this spinlock). -Jack On Monday 14 February 2011 09:32, sebastien dugue wrote: Resending to the proper ML (sorry). In mlx4_ib_reg_xrc_rcv_qp(), we need to

Re: [ewg] [PATCH]mlx4_ib XRC RCV: Fix mlx4_ib_reg_xrc_rcv_qp() locking

2011-02-16 Thread sebastien dugue
On Wed, 16 Feb 2011 14:50:02 +0200 Jack Morgenstein ja...@dev.mellanox.co.il wrote: You are correct! Good catch. We will add this to OFED. Thanks, (P.S., I would rather leave irqsave -- it is used everywhere else for this spinlock). Right, but everywhere you know for sure you're in

Re: [ewg] [PATCH]mlx4_ib XRC RCV: Fix mlx4_ib_reg_xrc_rcv_qp() locking

2011-02-16 Thread Jack Morgenstein
On Wednesday 16 February 2011 15:02, sebastien dugue wrote:   Also, one thing I noticed in that same function: why allocate ctx_entry before knowing if it's going to be of any use? The allocation could be done right before the first use. I did it just to gather all the error returns at

[ewg] [PATCH]mlx4_ib XRC RCV: Fix mlx4_ib_reg_xrc_rcv_qp() locking

2011-02-13 Thread sebastien dugue
Resending to the proper ML (sorry). In mlx4_ib_reg_xrc_rcv_qp(), we need to take the xrc_reg_list_lock spinlock when walking the xrc_reg_list. We've been hit by this on 2 customer sites. Also, I guess spin_lock_irqsave() could be replaced by spin_lock_irq() in that function as we know