lkml.org 
[lkml]   [2018]   [Jun]   [12]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 2/2] IB/mad: Use IDR for agent IDs
On Fri, Jun 08, 2018 at 10:42:18AM -0700, Matthew Wilcox wrote:
> From: Matthew Wilcox <mawilcox@microsoft.com>
>
> 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, and we can also limit the assigned
> ID to be less than 2^24, which fixes a bug in the mlx4 device.
>
> 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@microsoft.com>
> drivers/infiniband/core/mad.c | 78 ++++++++++++++++++------------
> drivers/infiniband/core/mad_priv.h | 7 +--
> include/linux/idr.h | 9 ++++
> 3 files changed, 59 insertions(+), 35 deletions(-)
>
> diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c
> index 68f4dda916c8..62384a3dd3ec 100644
> +++ b/drivers/infiniband/core/mad.c
> @@ -38,6 +38,7 @@
> #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
>
> #include <linux/dma-mapping.h>
> +#include <linux/idr.h>
> #include <linux/slab.h>
> #include <linux/module.h>
> #include <linux/security.h>
> @@ -58,8 +59,8 @@ MODULE_PARM_DESC(send_queue_size, "Size of send queue in number of work requests
> module_param_named(recv_queue_size, mad_recvq_size, int, 0444);
> MODULE_PARM_DESC(recv_queue_size, "Size of receive queue in number of work requests");
>
> +static DEFINE_IDR(ib_mad_clients);
> static struct list_head ib_mad_port_list;
> -static atomic_t ib_mad_client_id = ATOMIC_INIT(0);
>
> /* Port list lock */
> static DEFINE_SPINLOCK(ib_mad_port_list_lock);
> @@ -377,13 +378,24 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device,
> goto error4;
> }
>
> - spin_lock_irq(&port_priv->reg_lock);
> - mad_agent_priv->agent.hi_tid = atomic_inc_return(&ib_mad_client_id);
> + idr_preload(GFP_KERNEL);
> + idr_lock(&ib_mad_clients);
> + ret2 = idr_alloc_cyclic(&ib_mad_clients, mad_agent_priv, 0,
> + (1 << 24), GFP_ATOMIC);

I like this series, my only concern is this magic number here, at the
very least it deserves a big comment explaining why it
exists..

Let me see if I can get someone to give it some test time, I assume
you haven't been able to test it it?

> diff --git a/include/linux/idr.h b/include/linux/idr.h
> index e856f4e0ab35..bef0df8600e2 100644
> +++ b/include/linux/idr.h
> @@ -81,6 +81,15 @@ static inline void idr_set_cursor(struct idr *idr, unsigned int val)
> WRITE_ONCE(idr->idr_next, val);
> }
>
> +#define idr_lock(idr) xa_lock(&(idr)->idr_rt)
> +#define idr_unlock(idr) xa_unlock(&(idr)->idr_rt)
> +#define idr_lock_irq(idr) xa_lock_irq(&(idr)->idr_rt)
> +#define idr_unlock_irq(idr) xa_unlock_irq(&(idr)->idr_rt)
> +#define idr_lock_irqsave(idr, flags) \
> + xa_lock_irqsave(&(idr)->idr_rt, flags)
> +#define idr_unlock_irqrestore(idr, flags) \
> + xa_unlock_irqrestore(&(idr)->idr_rt, flags)
> +

And you are Ok to take these through the rdma tree?

Thanks,
Jason

\
 
 \ /
  Last update: 2018-06-12 22:35    [W:0.164 / U:33.616 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site