android_kernel_xiaomi_sm8350/drivers/net/cxgb3/cxgb3_offload.h
Steve Wise e998f245c4 RDMA/cxgb3: Doorbell overflow avoidance and recovery
T3 hardware doorbell FIFO overflows can cause application stalls due
to lost doorbell ring events.  This has been seen when running large
NP IMB alltoall MPI jobs.  The T3 hardware supports an xon/xoff-type
flow control mechanism to help avoid overflowing the HW doorbell FIFO.

This patch uses these interrupts to disable RDMA QP doorbell rings
when we near an overflow condition, and then turn them back on (and
ring all the active QP doorbells) when when the doorbell FIFO empties
out.  In addition if an doorbell ring is dropped by the hardware, the
code will now recover.

Design:

cxgb3:
- enable these DB interrupts
- in the interrupt handler, schedule work tasks to call the ULPs event
  handlers with the new events.
- ring all the qset txqs when an overflow is detected.

iw_cxgb3:
- disable db ringing on all active qps when we get the DB_FULL event
- enable db ringing on all active qps and ring all active dbs when we get
  the DB_EMPTY event
- On DB_DROP event:
       - disable db rings in the event handler
       - delay-schedule a work task which rings and enables the dbs on
         all active qps.
- in post_send and post_recv logic, don't ring the db if it's disabled.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
2010-02-24 10:40:28 -08:00

210 lines
6.2 KiB
C

/*
* Copyright (c) 2006-2008 Chelsio, Inc. All rights reserved.
*
* This software is available to you under a choice of one of two
* licenses. You may choose to be licensed under the terms of the GNU
* General Public License (GPL) Version 2, available from the file
* COPYING in the main directory of this source tree, or the
* OpenIB.org BSD license below:
*
* Redistribution and use in source and binary forms, with or
* without modification, are permitted provided that the following
* conditions are met:
*
* - Redistributions of source code must retain the above
* copyright notice, this list of conditions and the following
* disclaimer.
*
* - Redistributions in binary form must reproduce the above
* copyright notice, this list of conditions and the following
* disclaimer in the documentation and/or other materials
* provided with the distribution.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
#ifndef _CXGB3_OFFLOAD_H
#define _CXGB3_OFFLOAD_H
#include <linux/list.h>
#include <linux/skbuff.h>
#include "l2t.h"
#include "t3cdev.h"
#include "t3_cpl.h"
struct adapter;
void cxgb3_offload_init(void);
void cxgb3_adapter_ofld(struct adapter *adapter);
void cxgb3_adapter_unofld(struct adapter *adapter);
int cxgb3_offload_activate(struct adapter *adapter);
void cxgb3_offload_deactivate(struct adapter *adapter);
void cxgb3_set_dummy_ops(struct t3cdev *dev);
struct t3cdev *dev2t3cdev(struct net_device *dev);
/*
* Client registration. Users of T3 driver must register themselves.
* The T3 driver will call the add function of every client for each T3
* adapter activated, passing up the t3cdev ptr. Each client fills out an
* array of callback functions to process CPL messages.
*/
void cxgb3_register_client(struct cxgb3_client *client);
void cxgb3_unregister_client(struct cxgb3_client *client);
void cxgb3_add_clients(struct t3cdev *tdev);
void cxgb3_remove_clients(struct t3cdev *tdev);
void cxgb3_event_notify(struct t3cdev *tdev, u32 event, u32 port);
typedef int (*cxgb3_cpl_handler_func)(struct t3cdev *dev,
struct sk_buff *skb, void *ctx);
enum {
OFFLOAD_STATUS_UP,
OFFLOAD_STATUS_DOWN,
OFFLOAD_PORT_DOWN,
OFFLOAD_PORT_UP,
OFFLOAD_DB_FULL,
OFFLOAD_DB_EMPTY,
OFFLOAD_DB_DROP
};
struct cxgb3_client {
char *name;
void (*add) (struct t3cdev *);
void (*remove) (struct t3cdev *);
cxgb3_cpl_handler_func *handlers;
int (*redirect)(void *ctx, struct dst_entry *old,
struct dst_entry *new, struct l2t_entry *l2t);
struct list_head client_list;
void (*event_handler)(struct t3cdev *tdev, u32 event, u32 port);
};
/*
* TID allocation services.
*/
int cxgb3_alloc_atid(struct t3cdev *dev, struct cxgb3_client *client,
void *ctx);
int cxgb3_alloc_stid(struct t3cdev *dev, struct cxgb3_client *client,
void *ctx);
void *cxgb3_free_atid(struct t3cdev *dev, int atid);
void cxgb3_free_stid(struct t3cdev *dev, int stid);
void cxgb3_insert_tid(struct t3cdev *dev, struct cxgb3_client *client,
void *ctx, unsigned int tid);
void cxgb3_queue_tid_release(struct t3cdev *dev, unsigned int tid);
void cxgb3_remove_tid(struct t3cdev *dev, void *ctx, unsigned int tid);
struct t3c_tid_entry {
struct cxgb3_client *client;
void *ctx;
};
/* CPL message priority levels */
enum {
CPL_PRIORITY_DATA = 0, /* data messages */
CPL_PRIORITY_SETUP = 1, /* connection setup messages */
CPL_PRIORITY_TEARDOWN = 0, /* connection teardown messages */
CPL_PRIORITY_LISTEN = 1, /* listen start/stop messages */
CPL_PRIORITY_ACK = 1, /* RX ACK messages */
CPL_PRIORITY_CONTROL = 1 /* offload control messages */
};
/* Flags for return value of CPL message handlers */
enum {
CPL_RET_BUF_DONE = 1, /* buffer processing done, buffer may be freed */
CPL_RET_BAD_MSG = 2, /* bad CPL message (e.g., unknown opcode) */
CPL_RET_UNKNOWN_TID = 4 /* unexpected unknown TID */
};
typedef int (*cpl_handler_func)(struct t3cdev *dev, struct sk_buff *skb);
/*
* Returns a pointer to the first byte of the CPL header in an sk_buff that
* contains a CPL message.
*/
static inline void *cplhdr(struct sk_buff *skb)
{
return skb->data;
}
void t3_register_cpl_handler(unsigned int opcode, cpl_handler_func h);
union listen_entry {
struct t3c_tid_entry t3c_tid;
union listen_entry *next;
};
union active_open_entry {
struct t3c_tid_entry t3c_tid;
union active_open_entry *next;
};
/*
* Holds the size, base address, free list start, etc of the TID, server TID,
* and active-open TID tables for a offload device.
* The tables themselves are allocated dynamically.
*/
struct tid_info {
struct t3c_tid_entry *tid_tab;
unsigned int ntids;
atomic_t tids_in_use;
union listen_entry *stid_tab;
unsigned int nstids;
unsigned int stid_base;
union active_open_entry *atid_tab;
unsigned int natids;
unsigned int atid_base;
/*
* The following members are accessed R/W so we put them in their own
* cache lines.
*
* XXX We could combine the atid fields above with the lock here since
* atids are use once (unlike other tids). OTOH the above fields are
* usually in cache due to tid_tab.
*/
spinlock_t atid_lock ____cacheline_aligned_in_smp;
union active_open_entry *afree;
unsigned int atids_in_use;
spinlock_t stid_lock ____cacheline_aligned;
union listen_entry *sfree;
unsigned int stids_in_use;
};
struct t3c_data {
struct list_head list_node;
struct t3cdev *dev;
unsigned int tx_max_chunk; /* max payload for TX_DATA */
unsigned int max_wrs; /* max in-flight WRs per connection */
unsigned int nmtus;
const unsigned short *mtus;
struct tid_info tid_maps;
struct t3c_tid_entry *tid_release_list;
spinlock_t tid_release_lock;
struct work_struct tid_release_task;
struct sk_buff *nofail_skb;
unsigned int release_list_incomplete;
};
/*
* t3cdev -> t3c_data accessor
*/
#define T3C_DATA(dev) (*(struct t3c_data **)&(dev)->l4opt)
#endif