2006-09-22 15:22:22 -07:00
|
|
|
/*
|
|
|
|
* IBM eServer eHCA Infiniband device driver for Linux on POWER
|
|
|
|
*
|
|
|
|
* internal queue handling
|
|
|
|
*
|
|
|
|
* Authors: Waleri Fomin <fomin@de.ibm.com>
|
|
|
|
* Reinhard Ernst <rernst@de.ibm.com>
|
|
|
|
* Christoph Raisch <raisch@de.ibm.com>
|
|
|
|
*
|
|
|
|
* Copyright (c) 2005 IBM Corporation
|
|
|
|
*
|
|
|
|
* All rights reserved.
|
|
|
|
*
|
|
|
|
* This source code is distributed under a dual license of GPL v2.0 and OpenIB
|
|
|
|
* BSD.
|
|
|
|
*
|
|
|
|
* OpenIB BSD License
|
|
|
|
*
|
|
|
|
* 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.
|
|
|
|
*
|
|
|
|
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
|
|
|
|
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
|
|
|
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
|
|
|
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
|
|
|
|
* LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
|
|
|
|
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
|
|
|
|
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
|
|
|
|
* BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER
|
|
|
|
* IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
|
|
|
|
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
|
|
|
|
* POSSIBILITY OF SUCH DAMAGE.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#ifndef __IPZ_PT_FN_H__
|
|
|
|
#define __IPZ_PT_FN_H__
|
|
|
|
|
|
|
|
#define EHCA_PAGESHIFT 12
|
|
|
|
#define EHCA_PAGESIZE 4096UL
|
|
|
|
#define EHCA_PAGEMASK (~(EHCA_PAGESIZE-1))
|
|
|
|
#define EHCA_PT_ENTRIES 512UL
|
|
|
|
|
|
|
|
#include "ehca_tools.h"
|
|
|
|
#include "ehca_qes.h"
|
|
|
|
|
2007-07-20 07:04:17 -07:00
|
|
|
struct ehca_pd;
|
|
|
|
struct ipz_small_queue_page;
|
|
|
|
|
2007-07-28 21:47:53 -07:00
|
|
|
extern struct kmem_cache *small_qp_cache;
|
|
|
|
|
2006-09-22 15:22:22 -07:00
|
|
|
/* struct generic ehca page */
|
|
|
|
struct ipz_page {
|
|
|
|
u8 entries[EHCA_PAGESIZE];
|
|
|
|
};
|
|
|
|
|
2007-07-20 07:04:17 -07:00
|
|
|
#define IPZ_SPAGE_PER_KPAGE (PAGE_SIZE / 512)
|
|
|
|
|
|
|
|
struct ipz_small_queue_page {
|
|
|
|
unsigned long page;
|
|
|
|
unsigned long bitmap[IPZ_SPAGE_PER_KPAGE / BITS_PER_LONG];
|
|
|
|
int fill;
|
|
|
|
void *mapped_addr;
|
|
|
|
u32 mmap_count;
|
|
|
|
struct list_head list;
|
|
|
|
};
|
|
|
|
|
2006-09-22 15:22:22 -07:00
|
|
|
/* struct generic queue in linux kernel virtual memory (kv) */
|
|
|
|
struct ipz_queue {
|
|
|
|
u64 current_q_offset; /* current queue entry */
|
|
|
|
|
|
|
|
struct ipz_page **queue_pages; /* array of pages belonging to queue */
|
|
|
|
u32 qe_size; /* queue entry size */
|
|
|
|
u32 act_nr_of_sg;
|
|
|
|
u32 queue_length; /* queue length allocated in bytes */
|
|
|
|
u32 pagesize;
|
|
|
|
u32 toggle_state; /* toggle flag - per page */
|
2007-07-20 07:04:17 -07:00
|
|
|
u32 offset; /* save offset within page for small_qp */
|
|
|
|
struct ipz_small_queue_page *small_page;
|
2006-09-22 15:22:22 -07:00
|
|
|
};
|
|
|
|
|
|
|
|
/*
|
|
|
|
* return current Queue Entry for a certain q_offset
|
|
|
|
* returns address (kv) of Queue Entry
|
|
|
|
*/
|
|
|
|
static inline void *ipz_qeit_calc(struct ipz_queue *queue, u64 q_offset)
|
|
|
|
{
|
|
|
|
struct ipz_page *current_page;
|
|
|
|
if (q_offset >= queue->queue_length)
|
|
|
|
return NULL;
|
|
|
|
current_page = (queue->queue_pages)[q_offset >> EHCA_PAGESHIFT];
|
2007-02-15 09:06:33 -07:00
|
|
|
return ¤t_page->entries[q_offset & (EHCA_PAGESIZE - 1)];
|
2006-09-22 15:22:22 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* return current Queue Entry
|
|
|
|
* returns address (kv) of Queue Entry
|
|
|
|
*/
|
|
|
|
static inline void *ipz_qeit_get(struct ipz_queue *queue)
|
|
|
|
{
|
|
|
|
return ipz_qeit_calc(queue, queue->current_q_offset);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* return current Queue Page , increment Queue Page iterator from
|
|
|
|
* page to page in struct ipz_queue, last increment will return 0! and
|
|
|
|
* NOT wrap
|
|
|
|
* returns address (kv) of Queue Page
|
|
|
|
* warning don't use in parallel with ipz_QE_get_inc()
|
|
|
|
*/
|
|
|
|
void *ipz_qpageit_get_inc(struct ipz_queue *queue);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* return current Queue Entry, increment Queue Entry iterator by one
|
|
|
|
* step in struct ipz_queue, will wrap in ringbuffer
|
|
|
|
* returns address (kv) of Queue Entry BEFORE increment
|
|
|
|
* warning don't use in parallel with ipz_qpageit_get_inc()
|
|
|
|
*/
|
|
|
|
static inline void *ipz_qeit_get_inc(struct ipz_queue *queue)
|
|
|
|
{
|
|
|
|
void *ret = ipz_qeit_get(queue);
|
|
|
|
queue->current_q_offset += queue->qe_size;
|
|
|
|
if (queue->current_q_offset >= queue->queue_length) {
|
|
|
|
queue->current_q_offset = 0;
|
|
|
|
/* toggle the valid flag */
|
|
|
|
queue->toggle_state = (~queue->toggle_state) & 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2007-05-24 07:51:05 -07:00
|
|
|
/*
|
|
|
|
* return a bool indicating whether current Queue Entry is valid
|
|
|
|
*/
|
|
|
|
static inline int ipz_qeit_is_valid(struct ipz_queue *queue)
|
|
|
|
{
|
|
|
|
struct ehca_cqe *cqe = ipz_qeit_get(queue);
|
|
|
|
return ((cqe->cqe_flags >> 7) == (queue->toggle_state & 1));
|
|
|
|
}
|
|
|
|
|
2006-09-22 15:22:22 -07:00
|
|
|
/*
|
|
|
|
* return current Queue Entry, increment Queue Entry iterator by one
|
|
|
|
* step in struct ipz_queue, will wrap in ringbuffer
|
|
|
|
* returns address (kv) of Queue Entry BEFORE increment
|
|
|
|
* returns 0 and does not increment, if wrong valid state
|
|
|
|
* warning don't use in parallel with ipz_qpageit_get_inc()
|
|
|
|
*/
|
|
|
|
static inline void *ipz_qeit_get_inc_valid(struct ipz_queue *queue)
|
|
|
|
{
|
2007-05-24 07:51:05 -07:00
|
|
|
return ipz_qeit_is_valid(queue) ? ipz_qeit_get_inc(queue) : NULL;
|
IB: Return "maybe missed event" hint from ib_req_notify_cq()
The semantics defined by the InfiniBand specification say that
completion events are only generated when a completions is added to a
completion queue (CQ) after completion notification is requested. In
other words, this means that the following race is possible:
while (CQ is not empty)
ib_poll_cq(CQ);
// new completion is added after while loop is exited
ib_req_notify_cq(CQ);
// no event is generated for the existing completion
To close this race, the IB spec recommends doing another poll of the
CQ after requesting notification.
However, it is not always possible to arrange code this way (for
example, we have found that NAPI for IPoIB cannot poll after
requesting notification). Also, some hardware (eg Mellanox HCAs)
actually will generate an event for completions added before the call
to ib_req_notify_cq() -- which is allowed by the spec, since there's
no way for any upper-layer consumer to know exactly when a completion
was really added -- so the extra poll of the CQ is just a waste.
Motivated by this, we add a new flag "IB_CQ_REPORT_MISSED_EVENTS" for
ib_req_notify_cq() so that it can return a hint about whether the a
completion may have been added before the request for notification.
The return value of ib_req_notify_cq() is extended so:
< 0 means an error occurred while requesting notification
== 0 means notification was requested successfully, and if
IB_CQ_REPORT_MISSED_EVENTS was passed in, then no
events were missed and it is safe to wait for another
event.
> 0 is only returned if IB_CQ_REPORT_MISSED_EVENTS was
passed in. It means that the consumer must poll the
CQ again to make sure it is empty to avoid the race
described above.
We add a flag to enable this behavior rather than turning it on
unconditionally, because checking for missed events may incur
significant overhead for some low-level drivers, and consumers that
don't care about the results of this test shouldn't be forced to pay
for the test.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-05-06 21:02:48 -07:00
|
|
|
}
|
|
|
|
|
2006-09-22 15:22:22 -07:00
|
|
|
/*
|
|
|
|
* returns and resets Queue Entry iterator
|
|
|
|
* returns address (kv) of first Queue Entry
|
|
|
|
*/
|
|
|
|
static inline void *ipz_qeit_reset(struct ipz_queue *queue)
|
|
|
|
{
|
|
|
|
queue->current_q_offset = 0;
|
|
|
|
return ipz_qeit_get(queue);
|
|
|
|
}
|
|
|
|
|
2006-11-20 15:54:12 -07:00
|
|
|
/*
|
|
|
|
* return the q_offset corresponding to an absolute address
|
|
|
|
*/
|
|
|
|
int ipz_queue_abs_to_offset(struct ipz_queue *queue, u64 addr, u64 *q_offset);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* return the next queue offset. don't modify the queue.
|
|
|
|
*/
|
|
|
|
static inline u64 ipz_queue_advance_offset(struct ipz_queue *queue, u64 offset)
|
|
|
|
{
|
|
|
|
offset += queue->qe_size;
|
|
|
|
if (offset >= queue->queue_length) offset = 0;
|
|
|
|
return offset;
|
|
|
|
}
|
|
|
|
|
2006-09-22 15:22:22 -07:00
|
|
|
/* struct generic page table */
|
|
|
|
struct ipz_pt {
|
|
|
|
u64 entries[EHCA_PT_ENTRIES];
|
|
|
|
};
|
|
|
|
|
|
|
|
/* struct page table for a queue, only to be used in pf */
|
|
|
|
struct ipz_qpt {
|
|
|
|
/* queue page tables (kv), use u64 because we know the element length */
|
|
|
|
u64 *qpts;
|
|
|
|
u32 n_qpts;
|
|
|
|
u32 n_ptes; /* number of page table entries */
|
|
|
|
u64 *current_pte_addr;
|
|
|
|
};
|
|
|
|
|
|
|
|
/*
|
|
|
|
* constructor for a ipz_queue_t, placement new for ipz_queue_t,
|
|
|
|
* new for all dependent datastructors
|
|
|
|
* all QP Tables are the same
|
|
|
|
* flow:
|
|
|
|
* allocate+pin queue
|
|
|
|
* see ipz_qpt_ctor()
|
|
|
|
* returns true if ok, false if out of memory
|
|
|
|
*/
|
2007-07-20 07:04:17 -07:00
|
|
|
int ipz_queue_ctor(struct ehca_pd *pd, struct ipz_queue *queue,
|
|
|
|
const u32 nr_of_pages, const u32 pagesize,
|
|
|
|
const u32 qe_size, const u32 nr_of_sg,
|
|
|
|
int is_small);
|
2006-09-22 15:22:22 -07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* destructor for a ipz_queue_t
|
|
|
|
* -# free queue
|
|
|
|
* see ipz_queue_ctor()
|
|
|
|
* returns true if ok, false if queue was NULL-ptr of free failed
|
|
|
|
*/
|
2007-07-20 07:04:17 -07:00
|
|
|
int ipz_queue_dtor(struct ehca_pd *pd, struct ipz_queue *queue);
|
2006-09-22 15:22:22 -07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* constructor for a ipz_qpt_t,
|
|
|
|
* placement new for struct ipz_queue, new for all dependent datastructors
|
|
|
|
* all QP Tables are the same,
|
|
|
|
* flow:
|
|
|
|
* -# allocate+pin queue
|
|
|
|
* -# initialise ptcb
|
|
|
|
* -# allocate+pin PTs
|
|
|
|
* -# link PTs to a ring, according to HCA Arch, set bit62 id needed
|
|
|
|
* -# the ring must have room for exactly nr_of_PTEs
|
|
|
|
* see ipz_qpt_ctor()
|
|
|
|
*/
|
|
|
|
void ipz_qpt_ctor(struct ipz_qpt *qpt,
|
|
|
|
const u32 nr_of_qes,
|
|
|
|
const u32 pagesize,
|
|
|
|
const u32 qe_size,
|
|
|
|
const u8 lowbyte, const u8 toggle,
|
|
|
|
u32 * act_nr_of_QEs, u32 * act_nr_of_pages);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* return current Queue Entry, increment Queue Entry iterator by one
|
|
|
|
* step in struct ipz_queue, will wrap in ringbuffer
|
|
|
|
* returns address (kv) of Queue Entry BEFORE increment
|
|
|
|
* warning don't use in parallel with ipz_qpageit_get_inc()
|
|
|
|
* warning unpredictable results may occur if steps>act_nr_of_queue_entries
|
|
|
|
* fix EQ page problems
|
|
|
|
*/
|
|
|
|
void *ipz_qeit_eq_get_inc(struct ipz_queue *queue);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* return current Event Queue Entry, increment Queue Entry iterator
|
|
|
|
* by one step in struct ipz_queue if valid, will wrap in ringbuffer
|
|
|
|
* returns address (kv) of Queue Entry BEFORE increment
|
|
|
|
* returns 0 and does not increment, if wrong valid state
|
|
|
|
* warning don't use in parallel with ipz_queue_QPageit_get_inc()
|
|
|
|
* warning unpredictable results may occur if steps>act_nr_of_queue_entries
|
|
|
|
*/
|
|
|
|
static inline void *ipz_eqit_eq_get_inc_valid(struct ipz_queue *queue)
|
|
|
|
{
|
|
|
|
void *ret = ipz_qeit_get(queue);
|
2007-07-12 08:53:47 -07:00
|
|
|
u32 qe = *(u8 *)ret;
|
2006-09-22 15:22:22 -07:00
|
|
|
if ((qe >> 7) != (queue->toggle_state & 1))
|
|
|
|
return NULL;
|
|
|
|
ipz_qeit_eq_get_inc(queue); /* this is a good one */
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2007-02-15 09:06:33 -07:00
|
|
|
static inline void *ipz_eqit_eq_peek_valid(struct ipz_queue *queue)
|
|
|
|
{
|
|
|
|
void *ret = ipz_qeit_get(queue);
|
2007-07-12 08:53:47 -07:00
|
|
|
u32 qe = *(u8 *)ret;
|
2007-02-15 09:06:33 -07:00
|
|
|
if ((qe >> 7) != (queue->toggle_state & 1))
|
|
|
|
return NULL;
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2006-09-22 15:22:22 -07:00
|
|
|
/* returns address (GX) of first queue entry */
|
|
|
|
static inline u64 ipz_qpt_get_firstpage(struct ipz_qpt *qpt)
|
|
|
|
{
|
|
|
|
return be64_to_cpu(qpt->qpts[0]);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* returns address (kv) of first page of queue page table */
|
|
|
|
static inline void *ipz_qpt_get_qpt(struct ipz_qpt *qpt)
|
|
|
|
{
|
|
|
|
return qpt->qpts;
|
|
|
|
}
|
|
|
|
|
|
|
|
#endif /* __IPZ_PT_FN_H__ */
|