* [RFC PATCH 00/15] staging/rdma/hfi1: Initial patches to add rdmavt support in HFI1
@ 2015-12-14 17:27 Dennis Dalessandro
2015-12-14 17:28 ` [RFC PATCH 06/15] IB/hfi1: Remove driver specific members from hfi1 qp type Dennis Dalessandro
` (3 more replies)
0 siblings, 4 replies; 25+ messages in thread
From: Dennis Dalessandro @ 2015-12-14 17:27 UTC (permalink / raw)
To: devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-rdma-u79uwXL29TY76Z2rM5mHXA
Cc: gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r, dledford-H+wXaHxf7aLQT0dZR+AlfA
This patch series is being submitted as a Request For Comment only. It depends
on code submitted to the rdma subsystem [1].
This work is the first submission aimed to satisfy the TODO item for removing
duplicated code in hfi1. At the time of submission hfi1 and qib contained alot
of duplicated verbs processing code. The qib driver is having similar changes
made to use rdmavt. This will result in a common code base that both drivers and
future drivers such as soft-roce can use.
Note that due to the ongoing submission of hfi1 improvement patches, there will
likely be a number of conflicts which will still need to be resolved.
We also are still faced with the issue of separate trees for this work as was
discussed previously [2]. The result of that conversation was to keep the
drivers in separate trees until the 4.5 merge window. We are hoping that after
this merge window a single maintainer can take control of hfi1, qib, and rdmavt
so that these patches can move forward and be applied.
For now though we would like to get feedback on these patches with more to
follow.
[1] https://www.mail-archive.com/linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org/msg30074.html
[2] https://www.mail-archive.com/linux-rdma%40vger.kernel.org/msg29360.html
---
Dennis Dalessandro (15):
IB/hfi1: Begin to use rdmavt for verbs
IB/hfi1: Add basic rdmavt capability flags for hfi1
IB/hfi1: Consolidate dma ops for hfi1
IB/hfi1: Use rdmavt protection domain
IB/hfi1: Remove MR data structures from hfi1
IB/hfi1: Remove driver specific members from hfi1 qp type
IB/hfi1: Add device specific info prints
IB/hfi1: Use correct rdmavt header files after move.
IB/hfi1: Use address handle in rdmavt and remove from hfi1
IB/hfi1: Implement hfi1 support for AH notification
IB/hfi1: Remove hfi1 MR and hfi1 specific qp type
IB/hfi1: Remove srq from hfi1
IB/hfi1: Remove ibport and use rdmavt version
IB/hfi1: Remove mmap from hfi1
IB/hfi1: Use rdmavt pkey verbs function
drivers/staging/rdma/hfi1/Kconfig | 2
drivers/staging/rdma/hfi1/Makefile | 4
drivers/staging/rdma/hfi1/chip.c | 36 +-
drivers/staging/rdma/hfi1/common.h | 2
drivers/staging/rdma/hfi1/cq.c | 20 +
drivers/staging/rdma/hfi1/diag.c | 13 -
drivers/staging/rdma/hfi1/driver.c | 31 +-
drivers/staging/rdma/hfi1/hfi.h | 27 +-
drivers/staging/rdma/hfi1/init.c | 5
drivers/staging/rdma/hfi1/intr.c | 2
drivers/staging/rdma/hfi1/keys.c | 356 ---------------------
drivers/staging/rdma/hfi1/mad.c | 163 +++++-----
drivers/staging/rdma/hfi1/mmap.c | 192 -----------
drivers/staging/rdma/hfi1/mr.c | 522 ------------------------------
drivers/staging/rdma/hfi1/pio.c | 10 -
drivers/staging/rdma/hfi1/qp.c | 214 +++++++-----
drivers/staging/rdma/hfi1/qp.h | 44 +--
drivers/staging/rdma/hfi1/rc.c | 155 +++++----
drivers/staging/rdma/hfi1/ruc.c | 161 +++++----
drivers/staging/rdma/hfi1/sdma.h | 8
drivers/staging/rdma/hfi1/srq.c | 58 ++-
drivers/staging/rdma/hfi1/sysfs.c | 18 +
drivers/staging/rdma/hfi1/trace.h | 22 +
drivers/staging/rdma/hfi1/uc.c | 19 +
drivers/staging/rdma/hfi1/ud.c | 91 +++--
drivers/staging/rdma/hfi1/verbs.c | 526 +++++++++++--------------------
drivers/staging/rdma/hfi1/verbs.h | 531 ++++---------------------------
drivers/staging/rdma/hfi1/verbs_mcast.c | 36 +-
28 files changed, 843 insertions(+), 2425 deletions(-)
delete mode 100644 drivers/staging/rdma/hfi1/keys.c
delete mode 100644 drivers/staging/rdma/hfi1/mmap.c
delete mode 100644 drivers/staging/rdma/hfi1/mr.c
--
-Denny
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 25+ messages in thread
* [RFC PATCH 01/15] IB/hfi1: Begin to use rdmavt for verbs
[not found] ` <20151214172509.11576.19722.stgit-K+u1se/DcYrLESAwzcoQNrvm/XP+8Wra@public.gmane.org>
@ 2015-12-14 17:27 ` Dennis Dalessandro
2015-12-14 17:28 ` [RFC PATCH 02/15] IB/hfi1: Add basic rdmavt capability flags for hfi1 Dennis Dalessandro
` (11 subsequent siblings)
12 siblings, 0 replies; 25+ messages in thread
From: Dennis Dalessandro @ 2015-12-14 17:27 UTC (permalink / raw)
To: devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-rdma-u79uwXL29TY76Z2rM5mHXA
Cc: gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r, dledford-H+wXaHxf7aLQT0dZR+AlfA
This patch begins to make use of rdmavt by registering with it and
providing access to the header files. This is just the beginning of
rdmavt support in hfi1.
Reviewed-by: Ira Weiny <ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Reviewed-by: Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
drivers/staging/rdma/hfi1/Kconfig | 2 +-
drivers/staging/rdma/hfi1/chip.c | 2 +-
drivers/staging/rdma/hfi1/diag.c | 2 +-
drivers/staging/rdma/hfi1/hfi.h | 1 +
drivers/staging/rdma/hfi1/init.c | 5 +++--
drivers/staging/rdma/hfi1/intr.c | 2 +-
drivers/staging/rdma/hfi1/mad.c | 5 +++--
drivers/staging/rdma/hfi1/qp.c | 4 ++--
drivers/staging/rdma/hfi1/sysfs.c | 18 +++++++++---------
drivers/staging/rdma/hfi1/verbs.c | 15 ++++++++++-----
drivers/staging/rdma/hfi1/verbs.h | 8 ++++++--
11 files changed, 38 insertions(+), 26 deletions(-)
diff --git a/drivers/staging/rdma/hfi1/Kconfig b/drivers/staging/rdma/hfi1/Kconfig
index fd25078..55048fe 100644
--- a/drivers/staging/rdma/hfi1/Kconfig
+++ b/drivers/staging/rdma/hfi1/Kconfig
@@ -1,6 +1,6 @@
config INFINIBAND_HFI1
tristate "Intel OPA Gen1 support"
- depends on X86_64
+ depends on X86_64 && INFINIBAND_RDMAVT
default m
---help---
This is a low-level driver for Intel OPA Gen1 adapter.
diff --git a/drivers/staging/rdma/hfi1/chip.c b/drivers/staging/rdma/hfi1/chip.c
index dc69159..f799b86 100644
--- a/drivers/staging/rdma/hfi1/chip.c
+++ b/drivers/staging/rdma/hfi1/chip.c
@@ -6631,7 +6631,7 @@ int set_link_state(struct hfi1_pportdata *ppd, u32 state)
sdma_all_running(dd);
/* Signal the IB layer that the port has went active */
- event.device = &dd->verbs_dev.ibdev;
+ event.device = &dd->verbs_dev.rdi.ibdev;
event.element.port_num = ppd->port;
event.event = IB_EVENT_PORT_ACTIVE;
}
diff --git a/drivers/staging/rdma/hfi1/diag.c b/drivers/staging/rdma/hfi1/diag.c
index 0aaad74..0cf324d 100644
--- a/drivers/staging/rdma/hfi1/diag.c
+++ b/drivers/staging/rdma/hfi1/diag.c
@@ -856,7 +856,7 @@ static ssize_t hfi1_snoop_write(struct file *fp, const char __user *data,
vl = sc4;
} else {
sl = (byte_two >> 4) & 0xf;
- ibp = to_iport(&dd->verbs_dev.ibdev, 1);
+ ibp = to_iport(&dd->verbs_dev.rdi.ibdev, 1);
sc5 = ibp->sl_to_sc[sl];
vl = sc_to_vlt(dd, sc5);
if (vl != sc4) {
diff --git a/drivers/staging/rdma/hfi1/hfi.h b/drivers/staging/rdma/hfi1/hfi.h
index 54ed6b3..c4991be 100644
--- a/drivers/staging/rdma/hfi1/hfi.h
+++ b/drivers/staging/rdma/hfi1/hfi.h
@@ -65,6 +65,7 @@
#include <linux/cdev.h>
#include <linux/delay.h>
#include <linux/kthread.h>
+#include <rdma/rdma_vt.h>
#include "chip_registers.h"
#include "common.h"
diff --git a/drivers/staging/rdma/hfi1/init.c b/drivers/staging/rdma/hfi1/init.c
index 1c8286f..1f64e4e 100644
--- a/drivers/staging/rdma/hfi1/init.c
+++ b/drivers/staging/rdma/hfi1/init.c
@@ -56,6 +56,7 @@
#include <linux/module.h>
#include <linux/printk.h>
#include <linux/hrtimer.h>
+#include <rdma/rdma_vt.h>
#include "hfi.h"
#include "device.h"
@@ -985,7 +986,7 @@ void hfi1_free_devdata(struct hfi1_devdata *dd)
rcu_barrier(); /* wait for rcu callbacks to complete */
free_percpu(dd->int_counter);
free_percpu(dd->rcv_limit);
- ib_dealloc_device(&dd->verbs_dev.ibdev);
+ ib_dealloc_device(&dd->verbs_dev.rdi.ibdev);
}
/*
@@ -1081,7 +1082,7 @@ struct hfi1_devdata *hfi1_alloc_devdata(struct pci_dev *pdev, size_t extra)
bail:
if (!list_empty(&dd->list))
list_del_init(&dd->list);
- ib_dealloc_device(&dd->verbs_dev.ibdev);
+ ib_dealloc_device(&dd->verbs_dev.rdi.ibdev);
return ERR_PTR(ret);
}
diff --git a/drivers/staging/rdma/hfi1/intr.c b/drivers/staging/rdma/hfi1/intr.c
index 426582b..1283f2d 100644
--- a/drivers/staging/rdma/hfi1/intr.c
+++ b/drivers/staging/rdma/hfi1/intr.c
@@ -98,7 +98,7 @@ static void signal_ib_event(struct hfi1_pportdata *ppd, enum ib_event_type ev)
*/
if (!(dd->flags & HFI1_INITTED))
return;
- event.device = &dd->verbs_dev.ibdev;
+ event.device = &dd->verbs_dev.rdi.ibdev;
event.element.port_num = ppd->port;
event.event = ev;
ib_dispatch_event(&event);
diff --git a/drivers/staging/rdma/hfi1/mad.c b/drivers/staging/rdma/hfi1/mad.c
index a122565..1c34396 100644
--- a/drivers/staging/rdma/hfi1/mad.c
+++ b/drivers/staging/rdma/hfi1/mad.c
@@ -1387,7 +1387,7 @@ static int set_pkeys(struct hfi1_devdata *dd, u8 port, u16 *pkeys)
(void)hfi1_set_ib_cfg(ppd, HFI1_IB_CFG_PKEYS, 0);
event.event = IB_EVENT_PKEY_CHANGE;
- event.device = &dd->verbs_dev.ibdev;
+ event.device = &dd->verbs_dev.rdi.ibdev;
event.element.port_num = port;
ib_dispatch_event(&event);
}
@@ -4209,7 +4209,8 @@ int hfi1_create_agents(struct hfi1_ibdev *dev)
for (p = 0; p < dd->num_pports; p++) {
ibp = &dd->pport[p].ibport_data;
- agent = ib_register_mad_agent(&dev->ibdev, p + 1, IB_QPT_SMI,
+ agent = ib_register_mad_agent(&dev->rdi.ibdev, p + 1,
+ IB_QPT_SMI,
NULL, 0, send_handler,
NULL, NULL, 0);
if (IS_ERR(agent)) {
diff --git a/drivers/staging/rdma/hfi1/qp.c b/drivers/staging/rdma/hfi1/qp.c
index ce03681..bb447b5 100644
--- a/drivers/staging/rdma/hfi1/qp.c
+++ b/drivers/staging/rdma/hfi1/qp.c
@@ -1570,7 +1570,7 @@ struct qp_iter *qp_iter_init(struct hfi1_ibdev *dev)
return NULL;
iter->dev = dev;
- iter->specials = dev->ibdev.phys_port_cnt * 2;
+ iter->specials = dev->rdi.ibdev.phys_port_cnt * 2;
if (qp_iter_next(iter)) {
kfree(iter);
return NULL;
@@ -1610,7 +1610,7 @@ int qp_iter_next(struct qp_iter *iter)
struct hfi1_ibport *ibp;
int pidx;
- pidx = n % dev->ibdev.phys_port_cnt;
+ pidx = n % dev->rdi.ibdev.phys_port_cnt;
ppd = &dd_from_dev(dev)->pport[pidx];
ibp = &ppd->ibport_data;
diff --git a/drivers/staging/rdma/hfi1/sysfs.c b/drivers/staging/rdma/hfi1/sysfs.c
index 1dd6727..d05b9f3 100644
--- a/drivers/staging/rdma/hfi1/sysfs.c
+++ b/drivers/staging/rdma/hfi1/sysfs.c
@@ -446,7 +446,7 @@ static ssize_t show_rev(struct device *device, struct device_attribute *attr,
char *buf)
{
struct hfi1_ibdev *dev =
- container_of(device, struct hfi1_ibdev, ibdev.dev);
+ container_of(device, struct hfi1_ibdev, rdi.ibdev.dev);
return sprintf(buf, "%x\n", dd_from_dev(dev)->minrev);
}
@@ -455,7 +455,7 @@ static ssize_t show_hfi(struct device *device, struct device_attribute *attr,
char *buf)
{
struct hfi1_ibdev *dev =
- container_of(device, struct hfi1_ibdev, ibdev.dev);
+ container_of(device, struct hfi1_ibdev, rdi.ibdev.dev);
struct hfi1_devdata *dd = dd_from_dev(dev);
int ret;
@@ -470,7 +470,7 @@ static ssize_t show_boardversion(struct device *device,
struct device_attribute *attr, char *buf)
{
struct hfi1_ibdev *dev =
- container_of(device, struct hfi1_ibdev, ibdev.dev);
+ container_of(device, struct hfi1_ibdev, rdi.ibdev.dev);
struct hfi1_devdata *dd = dd_from_dev(dev);
/* The string printed here is already newline-terminated. */
@@ -482,7 +482,7 @@ static ssize_t show_nctxts(struct device *device,
struct device_attribute *attr, char *buf)
{
struct hfi1_ibdev *dev =
- container_of(device, struct hfi1_ibdev, ibdev.dev);
+ container_of(device, struct hfi1_ibdev, rdi.ibdev.dev);
struct hfi1_devdata *dd = dd_from_dev(dev);
/*
@@ -500,7 +500,7 @@ static ssize_t show_nfreectxts(struct device *device,
struct device_attribute *attr, char *buf)
{
struct hfi1_ibdev *dev =
- container_of(device, struct hfi1_ibdev, ibdev.dev);
+ container_of(device, struct hfi1_ibdev, rdi.ibdev.dev);
struct hfi1_devdata *dd = dd_from_dev(dev);
/* Return the number of free user ports (contexts) available. */
@@ -511,7 +511,7 @@ static ssize_t show_serial(struct device *device,
struct device_attribute *attr, char *buf)
{
struct hfi1_ibdev *dev =
- container_of(device, struct hfi1_ibdev, ibdev.dev);
+ container_of(device, struct hfi1_ibdev, rdi.ibdev.dev);
struct hfi1_devdata *dd = dd_from_dev(dev);
return scnprintf(buf, PAGE_SIZE, "%s", dd->serial);
@@ -523,7 +523,7 @@ static ssize_t store_chip_reset(struct device *device,
size_t count)
{
struct hfi1_ibdev *dev =
- container_of(device, struct hfi1_ibdev, ibdev.dev);
+ container_of(device, struct hfi1_ibdev, rdi.ibdev.dev);
struct hfi1_devdata *dd = dd_from_dev(dev);
int ret;
@@ -552,7 +552,7 @@ static ssize_t show_tempsense(struct device *device,
struct device_attribute *attr, char *buf)
{
struct hfi1_ibdev *dev =
- container_of(device, struct hfi1_ibdev, ibdev.dev);
+ container_of(device, struct hfi1_ibdev, rdi.ibdev.dev);
struct hfi1_devdata *dd = dd_from_dev(dev);
struct hfi1_temp temp;
int ret;
@@ -700,7 +700,7 @@ bail:
*/
int hfi1_verbs_register_sysfs(struct hfi1_devdata *dd)
{
- struct ib_device *dev = &dd->verbs_dev.ibdev;
+ struct ib_device *dev = &dd->verbs_dev.rdi.ibdev;
int i, ret;
for (i = 0; i < ARRAY_SIZE(hfi1_attributes); ++i) {
diff --git a/drivers/staging/rdma/hfi1/verbs.c b/drivers/staging/rdma/hfi1/verbs.c
index ef0feaa..4292d52 100644
--- a/drivers/staging/rdma/hfi1/verbs.c
+++ b/drivers/staging/rdma/hfi1/verbs.c
@@ -1903,7 +1903,7 @@ static void verbs_txreq_kmem_cache_ctor(void *obj)
int hfi1_register_ib_device(struct hfi1_devdata *dd)
{
struct hfi1_ibdev *dev = &dd->verbs_dev;
- struct ib_device *ibdev = &dev->ibdev;
+ struct ib_device *ibdev = &dev->rdi.ibdev;
struct hfi1_pportdata *ppd = dd->pport;
unsigned i, lk_tab_size;
int ret;
@@ -2070,7 +2070,13 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
strncpy(ibdev->node_desc, init_utsname()->nodename,
sizeof(ibdev->node_desc));
- ret = ib_register_device(ibdev, hfi1_create_port_files);
+ /*
+ * Fill in rvt info object.
+ */
+ dd->verbs_dev.rdi.driver_f.port_callback = hfi1_create_port_files;
+ dd->verbs_dev.rdi.dparms.props.max_pd = hfi1_max_pds;
+
+ ret = rvt_register_device(&dd->verbs_dev.rdi);
if (ret)
goto err_reg;
@@ -2087,7 +2093,7 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
err_class:
hfi1_free_agents(dev);
err_agents:
- ib_unregister_device(ibdev);
+ rvt_unregister_device(&dd->verbs_dev.rdi);
err_reg:
err_verbs_txreq:
kmem_cache_destroy(dev->verbs_txreq_cache);
@@ -2103,13 +2109,12 @@ bail:
void hfi1_unregister_ib_device(struct hfi1_devdata *dd)
{
struct hfi1_ibdev *dev = &dd->verbs_dev;
- struct ib_device *ibdev = &dev->ibdev;
hfi1_verbs_unregister_sysfs(dd);
hfi1_free_agents(dev);
- ib_unregister_device(ibdev);
+ rvt_unregister_device(&dd->verbs_dev.rdi);
if (!list_empty(&dev->txwait))
dd_dev_err(dd, "txwait list not empty!\n");
diff --git a/drivers/staging/rdma/hfi1/verbs.h b/drivers/staging/rdma/hfi1/verbs.h
index 7e27531..a290ed3 100644
--- a/drivers/staging/rdma/hfi1/verbs.h
+++ b/drivers/staging/rdma/hfi1/verbs.h
@@ -62,6 +62,7 @@
#include <rdma/ib_pack.h>
#include <rdma/ib_user_verbs.h>
#include <rdma/ib_mad.h>
+#include <rdma/rdma_vt.h>
struct hfi1_ctxtdata;
struct hfi1_pportdata;
@@ -749,7 +750,7 @@ struct hfi1_ibport {
struct hfi1_qp_ibdev;
struct hfi1_ibdev {
- struct ib_device ibdev;
+ struct rvt_dev_info rdi; /* Must be first */
struct list_head pending_mmaps;
spinlock_t mmap_offset_lock; /* protect mmap_offset */
u32 mmap_offset;
@@ -843,7 +844,10 @@ static inline struct hfi1_qp *to_iqp(struct ib_qp *ibqp)
static inline struct hfi1_ibdev *to_idev(struct ib_device *ibdev)
{
- return container_of(ibdev, struct hfi1_ibdev, ibdev);
+ struct rvt_dev_info *rdi;
+
+ rdi = container_of(ibdev, struct rvt_dev_info, ibdev);
+ return container_of(rdi, struct hfi1_ibdev, rdi);
}
/*
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [RFC PATCH 02/15] IB/hfi1: Add basic rdmavt capability flags for hfi1
[not found] ` <20151214172509.11576.19722.stgit-K+u1se/DcYrLESAwzcoQNrvm/XP+8Wra@public.gmane.org>
2015-12-14 17:27 ` [RFC PATCH 01/15] IB/hfi1: Begin to use rdmavt for verbs Dennis Dalessandro
@ 2015-12-14 17:28 ` Dennis Dalessandro
2015-12-14 17:28 ` [RFC PATCH 03/15] IB/hfi1: Consolidate dma ops " Dennis Dalessandro
` (10 subsequent siblings)
12 siblings, 0 replies; 25+ messages in thread
From: Dennis Dalessandro @ 2015-12-14 17:28 UTC (permalink / raw)
To: devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-rdma-u79uwXL29TY76Z2rM5mHXA
Cc: gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r, dledford-H+wXaHxf7aLQT0dZR+AlfA
Most functionality is still being done in the driver, set flags so that
rdmavt will let hfi1 continue to handle mr, qp, and cq init.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
drivers/staging/rdma/hfi1/verbs.c | 3 +++
1 files changed, 3 insertions(+), 0 deletions(-)
diff --git a/drivers/staging/rdma/hfi1/verbs.c b/drivers/staging/rdma/hfi1/verbs.c
index 4292d52..c457e82 100644
--- a/drivers/staging/rdma/hfi1/verbs.c
+++ b/drivers/staging/rdma/hfi1/verbs.c
@@ -2075,6 +2075,9 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
*/
dd->verbs_dev.rdi.driver_f.port_callback = hfi1_create_port_files;
dd->verbs_dev.rdi.dparms.props.max_pd = hfi1_max_pds;
+ dd->verbs_dev.rdi.flags = (RVT_FLAG_MR_INIT_DRIVER |
+ RVT_FLAG_QP_INIT_DRIVER |
+ RVT_FLAG_CQ_INIT_DRIVER);
ret = rvt_register_device(&dd->verbs_dev.rdi);
if (ret)
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [RFC PATCH 03/15] IB/hfi1: Consolidate dma ops for hfi1
[not found] ` <20151214172509.11576.19722.stgit-K+u1se/DcYrLESAwzcoQNrvm/XP+8Wra@public.gmane.org>
2015-12-14 17:27 ` [RFC PATCH 01/15] IB/hfi1: Begin to use rdmavt for verbs Dennis Dalessandro
2015-12-14 17:28 ` [RFC PATCH 02/15] IB/hfi1: Add basic rdmavt capability flags for hfi1 Dennis Dalessandro
@ 2015-12-14 17:28 ` Dennis Dalessandro
2015-12-14 17:28 ` [RFC PATCH 04/15] IB/hfi1: Use rdmavt protection domain Dennis Dalessandro
` (9 subsequent siblings)
12 siblings, 0 replies; 25+ messages in thread
From: Dennis Dalessandro @ 2015-12-14 17:28 UTC (permalink / raw)
To: devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-rdma-u79uwXL29TY76Z2rM5mHXA
Cc: gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r, dledford-H+wXaHxf7aLQT0dZR+AlfA
Remove the dma.c file from hfi1 in favor of using that which is
present in rdmavt.
Reviewed-by: Ira Weiny <ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Reviewed-by: Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
drivers/staging/rdma/hfi1/Makefile | 2 +-
drivers/staging/rdma/hfi1/verbs.c | 2 +-
drivers/staging/rdma/hfi1/verbs.h | 2 --
3 files changed, 2 insertions(+), 4 deletions(-)
diff --git a/drivers/staging/rdma/hfi1/Makefile b/drivers/staging/rdma/hfi1/Makefile
index 2e5daa6..2126b8b 100644
--- a/drivers/staging/rdma/hfi1/Makefile
+++ b/drivers/staging/rdma/hfi1/Makefile
@@ -7,7 +7,7 @@
#
obj-$(CONFIG_INFINIBAND_HFI1) += hfi1.o
-hfi1-y := chip.o cq.o device.o diag.o dma.o driver.o eprom.o file_ops.o firmware.o \
+hfi1-y := chip.o cq.o device.o diag.o driver.o eprom.o file_ops.o firmware.o \
init.o intr.o keys.o mad.o mmap.o mr.o pcie.o pio.o pio_copy.o \
qp.o qsfp.o rc.o ruc.o sdma.o srq.o sysfs.o trace.o twsi.o \
uc.o ud.o user_pages.o user_sdma.o verbs_mcast.o verbs.o
diff --git a/drivers/staging/rdma/hfi1/verbs.c b/drivers/staging/rdma/hfi1/verbs.c
index c457e82..22e2742 100644
--- a/drivers/staging/rdma/hfi1/verbs.c
+++ b/drivers/staging/rdma/hfi1/verbs.c
@@ -2064,7 +2064,7 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
ibdev->detach_mcast = hfi1_multicast_detach;
ibdev->process_mad = hfi1_process_mad;
ibdev->mmap = hfi1_mmap;
- ibdev->dma_ops = &hfi1_dma_mapping_ops;
+ ibdev->dma_ops = NULL;
ibdev->get_port_immutable = port_immutable;
strncpy(ibdev->node_desc, init_utsname()->nodename,
diff --git a/drivers/staging/rdma/hfi1/verbs.h b/drivers/staging/rdma/hfi1/verbs.h
index a290ed3..4f23e00 100644
--- a/drivers/staging/rdma/hfi1/verbs.h
+++ b/drivers/staging/rdma/hfi1/verbs.h
@@ -1155,6 +1155,4 @@ extern unsigned int hfi1_max_srq_wrs;
extern const u32 ib_hfi1_rnr_table[];
-extern struct ib_dma_mapping_ops hfi1_dma_mapping_ops;
-
#endif /* HFI1_VERBS_H */
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [RFC PATCH 04/15] IB/hfi1: Use rdmavt protection domain
[not found] ` <20151214172509.11576.19722.stgit-K+u1se/DcYrLESAwzcoQNrvm/XP+8Wra@public.gmane.org>
` (2 preceding siblings ...)
2015-12-14 17:28 ` [RFC PATCH 03/15] IB/hfi1: Consolidate dma ops " Dennis Dalessandro
@ 2015-12-14 17:28 ` Dennis Dalessandro
2015-12-14 17:28 ` [RFC PATCH 05/15] IB/hfi1: Remove MR data structures from hfi1 Dennis Dalessandro
` (8 subsequent siblings)
12 siblings, 0 replies; 25+ messages in thread
From: Dennis Dalessandro @ 2015-12-14 17:28 UTC (permalink / raw)
To: devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-rdma-u79uwXL29TY76Z2rM5mHXA
Cc: gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r, dledford-H+wXaHxf7aLQT0dZR+AlfA
Remove protection domain from hfi1 and use rdmavt's version.
Reviewed-by: Ira Weiny <ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Reviewed-by: Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
drivers/staging/rdma/hfi1/keys.c | 4 +-
drivers/staging/rdma/hfi1/mr.c | 2 +
drivers/staging/rdma/hfi1/ruc.c | 4 +-
drivers/staging/rdma/hfi1/verbs.c | 67 +++----------------------------------
drivers/staging/rdma/hfi1/verbs.h | 15 +-------
5 files changed, 12 insertions(+), 80 deletions(-)
diff --git a/drivers/staging/rdma/hfi1/keys.c b/drivers/staging/rdma/hfi1/keys.c
index cb4e608..57a266f 100644
--- a/drivers/staging/rdma/hfi1/keys.c
+++ b/drivers/staging/rdma/hfi1/keys.c
@@ -176,7 +176,7 @@ out:
* Check the IB SGE for validity and initialize our internal version
* of it.
*/
-int hfi1_lkey_ok(struct hfi1_lkey_table *rkt, struct hfi1_pd *pd,
+int hfi1_lkey_ok(struct hfi1_lkey_table *rkt, struct rvt_pd *pd,
struct hfi1_sge *isge, struct ib_sge *sge, int acc)
{
struct hfi1_mregion *mr;
@@ -285,7 +285,7 @@ int hfi1_rkey_ok(struct hfi1_qp *qp, struct hfi1_sge *sge,
*/
rcu_read_lock();
if (rkey == 0) {
- struct hfi1_pd *pd = to_ipd(qp->ibqp.pd);
+ struct rvt_pd *pd = ibpd_to_rvtpd(qp->ibqp.pd);
struct hfi1_ibdev *dev = to_idev(pd->ibpd.device);
if (pd->user)
diff --git a/drivers/staging/rdma/hfi1/mr.c b/drivers/staging/rdma/hfi1/mr.c
index 568f185..02589b2 100644
--- a/drivers/staging/rdma/hfi1/mr.c
+++ b/drivers/staging/rdma/hfi1/mr.c
@@ -116,7 +116,7 @@ struct ib_mr *hfi1_get_dma_mr(struct ib_pd *pd, int acc)
struct ib_mr *ret;
int rval;
- if (to_ipd(pd)->user) {
+ if (ibpd_to_rvtpd(pd)->user) {
ret = ERR_PTR(-EPERM);
goto bail;
}
diff --git a/drivers/staging/rdma/hfi1/ruc.c b/drivers/staging/rdma/hfi1/ruc.c
index 317bf6f..eb7aea9 100644
--- a/drivers/staging/rdma/hfi1/ruc.c
+++ b/drivers/staging/rdma/hfi1/ruc.c
@@ -102,11 +102,11 @@ static int init_sge(struct hfi1_qp *qp, struct hfi1_rwqe *wqe)
int i, j, ret;
struct ib_wc wc;
struct hfi1_lkey_table *rkt;
- struct hfi1_pd *pd;
+ struct rvt_pd *pd;
struct hfi1_sge_state *ss;
rkt = &to_idev(qp->ibqp.device)->lk_table;
- pd = to_ipd(qp->ibqp.srq ? qp->ibqp.srq->pd : qp->ibqp.pd);
+ pd = ibpd_to_rvtpd(qp->ibqp.srq ? qp->ibqp.srq->pd : qp->ibqp.pd);
ss = &qp->r_sge;
ss->sg_list = qp->r_sg_list;
qp->r_len = 0;
diff --git a/drivers/staging/rdma/hfi1/verbs.c b/drivers/staging/rdma/hfi1/verbs.c
index 22e2742..1390755 100644
--- a/drivers/staging/rdma/hfi1/verbs.c
+++ b/drivers/staging/rdma/hfi1/verbs.c
@@ -368,7 +368,7 @@ static int post_one_send(struct hfi1_qp *qp, struct ib_send_wr *wr)
int j;
int acc;
struct hfi1_lkey_table *rkt;
- struct hfi1_pd *pd;
+ struct rvt_pd *pd;
struct hfi1_devdata *dd = dd_from_ibdev(qp->ibqp.device);
struct hfi1_pportdata *ppd;
struct hfi1_ibport *ibp;
@@ -413,7 +413,7 @@ static int post_one_send(struct hfi1_qp *qp, struct ib_send_wr *wr)
return -ENOMEM;
rkt = &to_idev(qp->ibqp.device)->lk_table;
- pd = to_ipd(qp->ibqp.pd);
+ pd = ibpd_to_rvtpd(qp->ibqp.pd);
wqe = get_swqe_ptr(qp, qp->s_head);
@@ -1394,7 +1394,7 @@ static int query_device(struct ib_device *ibdev,
props->max_mr = dev->lk_table.max;
props->max_fmr = dev->lk_table.max;
props->max_map_per_fmr = 32767;
- props->max_pd = hfi1_max_pds;
+ props->max_pd = dev->rdi.dparms.props.max_pd;
props->max_qp_rd_atom = HFI1_MAX_RDMA_ATOMIC;
props->max_qp_init_rd_atom = 255;
/* props->max_res_rd_atom */
@@ -1592,61 +1592,6 @@ static int query_gid(struct ib_device *ibdev, u8 port,
return ret;
}
-static struct ib_pd *alloc_pd(struct ib_device *ibdev,
- struct ib_ucontext *context,
- struct ib_udata *udata)
-{
- struct hfi1_ibdev *dev = to_idev(ibdev);
- struct hfi1_pd *pd;
- struct ib_pd *ret;
-
- /*
- * This is actually totally arbitrary. Some correctness tests
- * assume there's a maximum number of PDs that can be allocated.
- * We don't actually have this limit, but we fail the test if
- * we allow allocations of more than we report for this value.
- */
-
- pd = kmalloc(sizeof(*pd), GFP_KERNEL);
- if (!pd) {
- ret = ERR_PTR(-ENOMEM);
- goto bail;
- }
-
- spin_lock(&dev->n_pds_lock);
- if (dev->n_pds_allocated == hfi1_max_pds) {
- spin_unlock(&dev->n_pds_lock);
- kfree(pd);
- ret = ERR_PTR(-ENOMEM);
- goto bail;
- }
-
- dev->n_pds_allocated++;
- spin_unlock(&dev->n_pds_lock);
-
- /* ib_alloc_pd() will initialize pd->ibpd. */
- pd->user = udata != NULL;
-
- ret = &pd->ibpd;
-
-bail:
- return ret;
-}
-
-static int dealloc_pd(struct ib_pd *ibpd)
-{
- struct hfi1_pd *pd = to_ipd(ibpd);
- struct hfi1_ibdev *dev = to_idev(ibpd->device);
-
- spin_lock(&dev->n_pds_lock);
- dev->n_pds_allocated--;
- spin_unlock(&dev->n_pds_lock);
-
- kfree(pd);
-
- return 0;
-}
-
/*
* convert ah port,sl to sc
*/
@@ -1920,7 +1865,7 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
init_ibport(ppd + i);
/* Only need to initialize non-zero fields. */
- spin_lock_init(&dev->n_pds_lock);
+
spin_lock_init(&dev->n_ahs_lock);
spin_lock_init(&dev->n_cqs_lock);
spin_lock_init(&dev->n_qps_lock);
@@ -2029,8 +1974,8 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
ibdev->query_gid = query_gid;
ibdev->alloc_ucontext = alloc_ucontext;
ibdev->dealloc_ucontext = dealloc_ucontext;
- ibdev->alloc_pd = alloc_pd;
- ibdev->dealloc_pd = dealloc_pd;
+ ibdev->alloc_pd = NULL;
+ ibdev->dealloc_pd = NULL;
ibdev->create_ah = create_ah;
ibdev->destroy_ah = destroy_ah;
ibdev->modify_ah = modify_ah;
diff --git a/drivers/staging/rdma/hfi1/verbs.h b/drivers/staging/rdma/hfi1/verbs.h
index 4f23e00..9043e5d 100644
--- a/drivers/staging/rdma/hfi1/verbs.h
+++ b/drivers/staging/rdma/hfi1/verbs.h
@@ -234,12 +234,6 @@ struct hfi1_mcast {
int n_attached;
};
-/* Protection domain */
-struct hfi1_pd {
- struct ib_pd ibpd;
- int user; /* non-zero if created from user space */
-};
-
/* Address Handle */
struct hfi1_ah {
struct ib_ah ibah;
@@ -776,8 +770,6 @@ struct hfi1_ibdev {
u64 n_kmem_wait;
u64 n_send_schedule;
- u32 n_pds_allocated; /* number of PDs allocated for device */
- spinlock_t n_pds_lock;
u32 n_ahs_allocated; /* number of AHs allocated for device */
spinlock_t n_ahs_lock;
u32 n_cqs_allocated; /* number of CQs allocated for device */
@@ -817,11 +809,6 @@ static inline struct hfi1_mr *to_imr(struct ib_mr *ibmr)
return container_of(ibmr, struct hfi1_mr, ibmr);
}
-static inline struct hfi1_pd *to_ipd(struct ib_pd *ibpd)
-{
- return container_of(ibpd, struct hfi1_pd, ibpd);
-}
-
static inline struct hfi1_ah *to_iah(struct ib_ah *ibah)
{
return container_of(ibah, struct hfi1_ah, ibah);
@@ -983,7 +970,7 @@ int hfi1_alloc_lkey(struct hfi1_mregion *mr, int dma_region);
void hfi1_free_lkey(struct hfi1_mregion *mr);
-int hfi1_lkey_ok(struct hfi1_lkey_table *rkt, struct hfi1_pd *pd,
+int hfi1_lkey_ok(struct hfi1_lkey_table *rkt, struct rvt_pd *pd,
struct hfi1_sge *isge, struct ib_sge *sge, int acc);
int hfi1_rkey_ok(struct hfi1_qp *qp, struct hfi1_sge *sge,
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [RFC PATCH 05/15] IB/hfi1: Remove MR data structures from hfi1
[not found] ` <20151214172509.11576.19722.stgit-K+u1se/DcYrLESAwzcoQNrvm/XP+8Wra@public.gmane.org>
` (3 preceding siblings ...)
2015-12-14 17:28 ` [RFC PATCH 04/15] IB/hfi1: Use rdmavt protection domain Dennis Dalessandro
@ 2015-12-14 17:28 ` Dennis Dalessandro
2015-12-14 17:28 ` [RFC PATCH 08/15] IB/hfi1: Use correct rdmavt header files after move Dennis Dalessandro
` (7 subsequent siblings)
12 siblings, 0 replies; 25+ messages in thread
From: Dennis Dalessandro @ 2015-12-14 17:28 UTC (permalink / raw)
To: devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-rdma-u79uwXL29TY76Z2rM5mHXA
Cc: gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r, dledford-H+wXaHxf7aLQT0dZR+AlfA
Remove MR data structures from hfi1 and use the version in rdmavt
Reviewed-by: Dean Luick <dean.luick-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Reviewed-by: Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
drivers/staging/rdma/hfi1/keys.c | 30 +++++++++---------
drivers/staging/rdma/hfi1/mr.c | 22 ++++++-------
drivers/staging/rdma/hfi1/ruc.c | 4 +-
drivers/staging/rdma/hfi1/sdma.h | 2 +
drivers/staging/rdma/hfi1/ud.c | 2 +
drivers/staging/rdma/hfi1/verbs.c | 16 +++++----
drivers/staging/rdma/hfi1/verbs.h | 63 ++++++-------------------------------
7 files changed, 48 insertions(+), 91 deletions(-)
diff --git a/drivers/staging/rdma/hfi1/keys.c b/drivers/staging/rdma/hfi1/keys.c
index 57a266f..ffaaa6f 100644
--- a/drivers/staging/rdma/hfi1/keys.c
+++ b/drivers/staging/rdma/hfi1/keys.c
@@ -63,21 +63,21 @@
*
*/
-int hfi1_alloc_lkey(struct hfi1_mregion *mr, int dma_region)
+int hfi1_alloc_lkey(struct rvt_mregion *mr, int dma_region)
{
unsigned long flags;
u32 r;
u32 n;
int ret = 0;
struct hfi1_ibdev *dev = to_idev(mr->pd->device);
- struct hfi1_lkey_table *rkt = &dev->lk_table;
+ struct rvt_lkey_table *rkt = &dev->lk_table;
hfi1_get_mr(mr);
spin_lock_irqsave(&rkt->lock, flags);
/* special case for dma_mr lkey == 0 */
if (dma_region) {
- struct hfi1_mregion *tmr;
+ struct rvt_mregion *tmr;
tmr = rcu_access_pointer(dev->dma_mr);
if (!tmr) {
@@ -133,13 +133,13 @@ bail:
* hfi1_free_lkey - free an lkey
* @mr: mr to free from tables
*/
-void hfi1_free_lkey(struct hfi1_mregion *mr)
+void hfi1_free_lkey(struct rvt_mregion *mr)
{
unsigned long flags;
u32 lkey = mr->lkey;
u32 r;
struct hfi1_ibdev *dev = to_idev(mr->pd->device);
- struct hfi1_lkey_table *rkt = &dev->lk_table;
+ struct rvt_lkey_table *rkt = &dev->lk_table;
int freed = 0;
spin_lock_irqsave(&rkt->lock, flags);
@@ -176,10 +176,10 @@ out:
* Check the IB SGE for validity and initialize our internal version
* of it.
*/
-int hfi1_lkey_ok(struct hfi1_lkey_table *rkt, struct rvt_pd *pd,
+int hfi1_lkey_ok(struct rvt_lkey_table *rkt, struct rvt_pd *pd,
struct hfi1_sge *isge, struct ib_sge *sge, int acc)
{
- struct hfi1_mregion *mr;
+ struct rvt_mregion *mr;
unsigned n, m;
size_t off;
@@ -231,15 +231,15 @@ int hfi1_lkey_ok(struct hfi1_lkey_table *rkt, struct rvt_pd *pd,
entries_spanned_by_off = off >> mr->page_shift;
off -= (entries_spanned_by_off << mr->page_shift);
- m = entries_spanned_by_off / HFI1_SEGSZ;
- n = entries_spanned_by_off % HFI1_SEGSZ;
+ m = entries_spanned_by_off / RVT_SEGSZ;
+ n = entries_spanned_by_off % RVT_SEGSZ;
} else {
m = 0;
n = 0;
while (off >= mr->map[m]->segs[n].length) {
off -= mr->map[m]->segs[n].length;
n++;
- if (n >= HFI1_SEGSZ) {
+ if (n >= RVT_SEGSZ) {
m++;
n = 0;
}
@@ -274,8 +274,8 @@ bail:
int hfi1_rkey_ok(struct hfi1_qp *qp, struct hfi1_sge *sge,
u32 len, u64 vaddr, u32 rkey, int acc)
{
- struct hfi1_lkey_table *rkt = &to_idev(qp->ibqp.device)->lk_table;
- struct hfi1_mregion *mr;
+ struct rvt_lkey_table *rkt = &to_idev(qp->ibqp.device)->lk_table;
+ struct rvt_mregion *mr;
unsigned n, m;
size_t off;
@@ -328,15 +328,15 @@ int hfi1_rkey_ok(struct hfi1_qp *qp, struct hfi1_sge *sge,
entries_spanned_by_off = off >> mr->page_shift;
off -= (entries_spanned_by_off << mr->page_shift);
- m = entries_spanned_by_off / HFI1_SEGSZ;
- n = entries_spanned_by_off % HFI1_SEGSZ;
+ m = entries_spanned_by_off / RVT_SEGSZ;
+ n = entries_spanned_by_off % RVT_SEGSZ;
} else {
m = 0;
n = 0;
while (off >= mr->map[m]->segs[n].length) {
off -= mr->map[m]->segs[n].length;
n++;
- if (n >= HFI1_SEGSZ) {
+ if (n >= RVT_SEGSZ) {
m++;
n = 0;
}
diff --git a/drivers/staging/rdma/hfi1/mr.c b/drivers/staging/rdma/hfi1/mr.c
index 02589b2..27f8081 100644
--- a/drivers/staging/rdma/hfi1/mr.c
+++ b/drivers/staging/rdma/hfi1/mr.c
@@ -56,7 +56,7 @@
/* Fast memory region */
struct hfi1_fmr {
struct ib_fmr ibfmr;
- struct hfi1_mregion mr; /* must be last */
+ struct rvt_mregion mr; /* must be last */
};
static inline struct hfi1_fmr *to_ifmr(struct ib_fmr *ibfmr)
@@ -64,13 +64,13 @@ static inline struct hfi1_fmr *to_ifmr(struct ib_fmr *ibfmr)
return container_of(ibfmr, struct hfi1_fmr, ibfmr);
}
-static int init_mregion(struct hfi1_mregion *mr, struct ib_pd *pd,
+static int init_mregion(struct rvt_mregion *mr, struct ib_pd *pd,
int count)
{
int m, i = 0;
int rval = 0;
- m = (count + HFI1_SEGSZ - 1) / HFI1_SEGSZ;
+ m = (count + RVT_SEGSZ - 1) / RVT_SEGSZ;
for (; i < m; i++) {
mr->map[i] = kzalloc(sizeof(*mr->map[0]), GFP_KERNEL);
if (!mr->map[i])
@@ -91,7 +91,7 @@ bail:
goto out;
}
-static void deinit_mregion(struct hfi1_mregion *mr)
+static void deinit_mregion(struct rvt_mregion *mr)
{
int i = mr->mapsz;
@@ -159,7 +159,7 @@ static struct hfi1_mr *alloc_mr(int count, struct ib_pd *pd)
int m;
/* Allocate struct plus pointers to first level page tables. */
- m = (count + HFI1_SEGSZ - 1) / HFI1_SEGSZ;
+ m = (count + RVT_SEGSZ - 1) / RVT_SEGSZ;
mr = kzalloc(sizeof(*mr) + m * sizeof(mr->mr.map[0]), GFP_KERNEL);
if (!mr)
goto bail;
@@ -221,7 +221,7 @@ struct ib_mr *hfi1_reg_phys_mr(struct ib_pd *pd,
mr->mr.map[m]->segs[n].length = buffer_list[i].size;
mr->mr.length += buffer_list[i].size;
n++;
- if (n == HFI1_SEGSZ) {
+ if (n == RVT_SEGSZ) {
m++;
n = 0;
}
@@ -294,7 +294,7 @@ struct ib_mr *hfi1_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
mr->mr.map[m]->segs[n].vaddr = vaddr;
mr->mr.map[m]->segs[n].length = umem->page_size;
n++;
- if (n == HFI1_SEGSZ) {
+ if (n == RVT_SEGSZ) {
m++;
n = 0;
}
@@ -382,7 +382,7 @@ struct ib_fmr *hfi1_alloc_fmr(struct ib_pd *pd, int mr_access_flags,
int rval = -ENOMEM;
/* Allocate struct plus pointers to first level page tables. */
- m = (fmr_attr->max_pages + HFI1_SEGSZ - 1) / HFI1_SEGSZ;
+ m = (fmr_attr->max_pages + RVT_SEGSZ - 1) / RVT_SEGSZ;
fmr = kzalloc(sizeof(*fmr) + m * sizeof(fmr->mr.map[0]), GFP_KERNEL);
if (!fmr)
goto bail;
@@ -434,7 +434,7 @@ int hfi1_map_phys_fmr(struct ib_fmr *ibfmr, u64 *page_list,
int list_len, u64 iova)
{
struct hfi1_fmr *fmr = to_ifmr(ibfmr);
- struct hfi1_lkey_table *rkt;
+ struct rvt_lkey_table *rkt;
unsigned long flags;
int m, n, i;
u32 ps;
@@ -459,7 +459,7 @@ int hfi1_map_phys_fmr(struct ib_fmr *ibfmr, u64 *page_list,
for (i = 0; i < list_len; i++) {
fmr->mr.map[m]->segs[n].vaddr = (void *) page_list[i];
fmr->mr.map[m]->segs[n].length = ps;
- if (++n == HFI1_SEGSZ) {
+ if (++n == RVT_SEGSZ) {
m++;
n = 0;
}
@@ -480,7 +480,7 @@ bail:
int hfi1_unmap_fmr(struct list_head *fmr_list)
{
struct hfi1_fmr *fmr;
- struct hfi1_lkey_table *rkt;
+ struct rvt_lkey_table *rkt;
unsigned long flags;
list_for_each_entry(fmr, fmr_list, ibfmr.list) {
diff --git a/drivers/staging/rdma/hfi1/ruc.c b/drivers/staging/rdma/hfi1/ruc.c
index eb7aea9..1c9c054 100644
--- a/drivers/staging/rdma/hfi1/ruc.c
+++ b/drivers/staging/rdma/hfi1/ruc.c
@@ -101,7 +101,7 @@ static int init_sge(struct hfi1_qp *qp, struct hfi1_rwqe *wqe)
{
int i, j, ret;
struct ib_wc wc;
- struct hfi1_lkey_table *rkt;
+ struct rvt_lkey_table *rkt;
struct rvt_pd *pd;
struct hfi1_sge_state *ss;
@@ -532,7 +532,7 @@ again:
if (--sqp->s_sge.num_sge)
*sge = *sqp->s_sge.sg_list++;
} else if (sge->length == 0 && sge->mr->lkey) {
- if (++sge->n >= HFI1_SEGSZ) {
+ if (++sge->n >= RVT_SEGSZ) {
if (++sge->m >= sge->mr->mapsz)
break;
sge->n = 0;
diff --git a/drivers/staging/rdma/hfi1/sdma.h b/drivers/staging/rdma/hfi1/sdma.h
index 85701ee..664b251 100644
--- a/drivers/staging/rdma/hfi1/sdma.h
+++ b/drivers/staging/rdma/hfi1/sdma.h
@@ -381,7 +381,7 @@ struct verbs_txreq {
struct sdma_txreq txreq;
struct hfi1_qp *qp;
struct hfi1_swqe *wqe;
- struct hfi1_mregion *mr;
+ struct rvt_mregion *mr;
struct hfi1_sge_state *ss;
struct sdma_engine *sde;
u16 hdr_dwords;
diff --git a/drivers/staging/rdma/hfi1/ud.c b/drivers/staging/rdma/hfi1/ud.c
index 54ff1f5..969335a 100644
--- a/drivers/staging/rdma/hfi1/ud.c
+++ b/drivers/staging/rdma/hfi1/ud.c
@@ -211,7 +211,7 @@ static void ud_loopback(struct hfi1_qp *sqp, struct hfi1_swqe *swqe)
if (--ssge.num_sge)
*sge = *ssge.sg_list++;
} else if (sge->length == 0 && sge->mr->lkey) {
- if (++sge->n >= HFI1_SEGSZ) {
+ if (++sge->n >= RVT_SEGSZ) {
if (++sge->m >= sge->mr->mapsz)
break;
sge->n = 0;
diff --git a/drivers/staging/rdma/hfi1/verbs.c b/drivers/staging/rdma/hfi1/verbs.c
index 1390755..9abf691 100644
--- a/drivers/staging/rdma/hfi1/verbs.c
+++ b/drivers/staging/rdma/hfi1/verbs.c
@@ -300,7 +300,7 @@ void hfi1_copy_sge(
if (--ss->num_sge)
*sge = *ss->sg_list++;
} else if (sge->length == 0 && sge->mr->lkey) {
- if (++sge->n >= HFI1_SEGSZ) {
+ if (++sge->n >= RVT_SEGSZ) {
if (++sge->m >= sge->mr->mapsz)
break;
sge->n = 0;
@@ -341,7 +341,7 @@ void hfi1_skip_sge(struct hfi1_sge_state *ss, u32 length, int release)
if (--ss->num_sge)
*sge = *ss->sg_list++;
} else if (sge->length == 0 && sge->mr->lkey) {
- if (++sge->n >= HFI1_SEGSZ) {
+ if (++sge->n >= RVT_SEGSZ) {
if (++sge->m >= sge->mr->mapsz)
break;
sge->n = 0;
@@ -367,7 +367,7 @@ static int post_one_send(struct hfi1_qp *qp, struct ib_send_wr *wr)
int i;
int j;
int acc;
- struct hfi1_lkey_table *rkt;
+ struct rvt_lkey_table *rkt;
struct rvt_pd *pd;
struct hfi1_devdata *dd = dd_from_ibdev(qp->ibqp.device);
struct hfi1_pportdata *ppd;
@@ -725,7 +725,7 @@ void update_sge(struct hfi1_sge_state *ss, u32 length)
if (--ss->num_sge)
*sge = *ss->sg_list++;
} else if (sge->length == 0 && sge->mr->lkey) {
- if (++sge->n >= HFI1_SEGSZ) {
+ if (++sge->n >= RVT_SEGSZ) {
if (++sge->m >= sge->mr->mapsz)
return;
sge->n = 0;
@@ -1883,13 +1883,13 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
spin_lock_init(&dev->lk_table.lock);
dev->lk_table.max = 1 << hfi1_lkey_table_size;
/* ensure generation is at least 4 bits (keys.c) */
- if (hfi1_lkey_table_size > MAX_LKEY_TABLE_BITS) {
+ if (hfi1_lkey_table_size > RVT_MAX_LKEY_TABLE_BITS) {
dd_dev_warn(dd, "lkey bits %u too large, reduced to %u\n",
- hfi1_lkey_table_size, MAX_LKEY_TABLE_BITS);
- hfi1_lkey_table_size = MAX_LKEY_TABLE_BITS;
+ hfi1_lkey_table_size, RVT_MAX_LKEY_TABLE_BITS);
+ hfi1_lkey_table_size = RVT_MAX_LKEY_TABLE_BITS;
}
lk_tab_size = dev->lk_table.max * sizeof(*dev->lk_table.table);
- dev->lk_table.table = (struct hfi1_mregion __rcu **)
+ dev->lk_table.table = (struct rvt_mregion __rcu **)
vmalloc(lk_tab_size);
if (dev->lk_table.table == NULL) {
ret = -ENOMEM;
diff --git a/drivers/staging/rdma/hfi1/verbs.h b/drivers/staging/rdma/hfi1/verbs.h
index 9043e5d..d80111f 100644
--- a/drivers/staging/rdma/hfi1/verbs.h
+++ b/drivers/staging/rdma/hfi1/verbs.h
@@ -285,44 +285,11 @@ struct hfi1_cq {
};
/*
- * A segment is a linear region of low physical memory.
- * Used by the verbs layer.
- */
-struct hfi1_seg {
- void *vaddr;
- size_t length;
-};
-
-/* The number of hfi1_segs that fit in a page. */
-#define HFI1_SEGSZ (PAGE_SIZE / sizeof(struct hfi1_seg))
-
-struct hfi1_segarray {
- struct hfi1_seg segs[HFI1_SEGSZ];
-};
-
-struct hfi1_mregion {
- struct ib_pd *pd; /* shares refcnt of ibmr.pd */
- u64 user_base; /* User's address for this region */
- u64 iova; /* IB start address of this region */
- size_t length;
- u32 lkey;
- u32 offset; /* offset (bytes) to start of region */
- int access_flags;
- u32 max_segs; /* number of hfi1_segs in all the arrays */
- u32 mapsz; /* size of the map array */
- u8 page_shift; /* 0 - non unform/non powerof2 sizes */
- u8 lkey_published; /* in global table */
- struct completion comp; /* complete when refcount goes to zero */
- atomic_t refcount;
- struct hfi1_segarray *map[0]; /* the segments */
-};
-
-/*
* These keep track of the copy progress within a memory region.
* Used by the verbs layer.
*/
struct hfi1_sge {
- struct hfi1_mregion *mr;
+ struct rvt_mregion *mr;
void *vaddr; /* kernel virtual address of segment */
u32 sge_length; /* length of the SGE */
u32 length; /* remaining length of the segment */
@@ -334,7 +301,7 @@ struct hfi1_sge {
struct hfi1_mr {
struct ib_mr ibmr;
struct ib_umem *umem;
- struct hfi1_mregion mr; /* must be last */
+ struct rvt_mregion mr; /* must be last */
};
/*
@@ -500,7 +467,7 @@ struct hfi1_qp {
u32 s_flags;
struct hfi1_swqe *s_wqe;
struct hfi1_sge_state s_sge; /* current send request data */
- struct hfi1_mregion *s_rdma_mr;
+ struct rvt_mregion *s_rdma_mr;
struct sdma_engine *s_sde; /* current sde */
u32 s_cur_size; /* size of send packet in bytes */
u32 s_len; /* total length of s_sge */
@@ -655,16 +622,6 @@ static inline struct hfi1_rwqe *get_rwqe_ptr(struct hfi1_rq *rq, unsigned n)
rq->max_sge * sizeof(struct ib_sge)) * n);
}
-#define MAX_LKEY_TABLE_BITS 23
-
-struct hfi1_lkey_table {
- spinlock_t lock; /* protect changes in this struct */
- u32 next; /* next unused index (speeds search) */
- u32 gen; /* generation count */
- u32 max; /* size of the table */
- struct hfi1_mregion __rcu **table;
-};
-
struct hfi1_opcode_stats {
u64 n_packets; /* number of packets */
u64 n_bytes; /* total number of bytes */
@@ -748,12 +705,12 @@ struct hfi1_ibdev {
struct list_head pending_mmaps;
spinlock_t mmap_offset_lock; /* protect mmap_offset */
u32 mmap_offset;
- struct hfi1_mregion __rcu *dma_mr;
+ struct rvt_mregion __rcu *dma_mr;
struct hfi1_qp_ibdev *qp_dev;
/* QP numbers are shared by all IB ports */
- struct hfi1_lkey_table lk_table;
+ struct rvt_lkey_table lk_table;
/* protect wait lists */
seqlock_t iowait_lock;
struct list_head txwait; /* list for wait verbs_txreq */
@@ -966,11 +923,11 @@ void hfi1_ud_rcv(struct hfi1_packet *packet);
int hfi1_lookup_pkey_idx(struct hfi1_ibport *ibp, u16 pkey);
-int hfi1_alloc_lkey(struct hfi1_mregion *mr, int dma_region);
+int hfi1_alloc_lkey(struct rvt_mregion *mr, int dma_region);
-void hfi1_free_lkey(struct hfi1_mregion *mr);
+void hfi1_free_lkey(struct rvt_mregion *mr);
-int hfi1_lkey_ok(struct hfi1_lkey_table *rkt, struct rvt_pd *pd,
+int hfi1_lkey_ok(struct rvt_lkey_table *rkt, struct rvt_pd *pd,
struct hfi1_sge *isge, struct ib_sge *sge, int acc);
int hfi1_rkey_ok(struct hfi1_qp *qp, struct hfi1_sge *sge,
@@ -1039,12 +996,12 @@ int hfi1_unmap_fmr(struct list_head *fmr_list);
int hfi1_dealloc_fmr(struct ib_fmr *ibfmr);
-static inline void hfi1_get_mr(struct hfi1_mregion *mr)
+static inline void hfi1_get_mr(struct rvt_mregion *mr)
{
atomic_inc(&mr->refcount);
}
-static inline void hfi1_put_mr(struct hfi1_mregion *mr)
+static inline void hfi1_put_mr(struct rvt_mregion *mr)
{
if (unlikely(atomic_dec_and_test(&mr->refcount)))
complete(&mr->comp);
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [RFC PATCH 06/15] IB/hfi1: Remove driver specific members from hfi1 qp type
2015-12-14 17:27 [RFC PATCH 00/15] staging/rdma/hfi1: Initial patches to add rdmavt support in HFI1 Dennis Dalessandro
@ 2015-12-14 17:28 ` Dennis Dalessandro
2015-12-14 17:28 ` [RFC PATCH 07/15] IB/hfi1: Add device specific info prints Dennis Dalessandro
` (2 subsequent siblings)
3 siblings, 0 replies; 25+ messages in thread
From: Dennis Dalessandro @ 2015-12-14 17:28 UTC (permalink / raw)
To: devel, linux-rdma; +Cc: gregkh, dledford
In preparation for moving the queue pair data structure to rdmavt the
members of the driver specific queue pairs which are not common need to be
pushed off to a private driver structure. This structure will be available
in the queue pair once moved to rdmavt as a void pointer. This patch while
not adding a lot of value in and of itself is a prerequisite to move the
queue pair out of the drivers and into rdmavt.
The driver specific, private queue pair data structure should condense as
more of the send side code moves to rdmavt.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jubin John <jubin.john@intel.com>
---
drivers/staging/rdma/hfi1/diag.c | 3 +
drivers/staging/rdma/hfi1/pio.c | 6 ++-
drivers/staging/rdma/hfi1/qp.c | 76 ++++++++++++++++++++++++-------------
drivers/staging/rdma/hfi1/qp.h | 9 +++-
drivers/staging/rdma/hfi1/rc.c | 7 ++-
drivers/staging/rdma/hfi1/ruc.c | 46 ++++++++++++----------
drivers/staging/rdma/hfi1/uc.c | 7 ++-
drivers/staging/rdma/hfi1/ud.c | 37 +++++++++---------
drivers/staging/rdma/hfi1/verbs.c | 54 ++++++++++++++++----------
drivers/staging/rdma/hfi1/verbs.h | 26 ++++++++++---
10 files changed, 166 insertions(+), 105 deletions(-)
diff --git a/drivers/staging/rdma/hfi1/diag.c b/drivers/staging/rdma/hfi1/diag.c
index 0cf324d..e172f2a 100644
--- a/drivers/staging/rdma/hfi1/diag.c
+++ b/drivers/staging/rdma/hfi1/diag.c
@@ -1634,7 +1634,8 @@ int snoop_send_dma_handler(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
int snoop_send_pio_handler(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
u64 pbc)
{
- struct ahg_ib_header *ahdr = qp->s_hdr;
+ struct hfi1_qp_priv *priv = qp->priv;
+ struct ahg_ib_header *ahdr = priv->s_hdr;
u32 hdrwords = qp->s_hdrwords;
struct hfi1_sge_state *ss = qp->s_cur_sge;
u32 len = qp->s_cur_size;
diff --git a/drivers/staging/rdma/hfi1/pio.c b/drivers/staging/rdma/hfi1/pio.c
index eab58c1..b5bce5e 100644
--- a/drivers/staging/rdma/hfi1/pio.c
+++ b/drivers/staging/rdma/hfi1/pio.c
@@ -1498,6 +1498,7 @@ static void sc_piobufavail(struct send_context *sc)
struct list_head *list;
struct hfi1_qp *qps[PIO_WAIT_BATCH_SIZE];
struct hfi1_qp *qp;
+ struct hfi1_qp_priv *priv;
unsigned long flags;
unsigned i, n = 0;
@@ -1517,8 +1518,9 @@ static void sc_piobufavail(struct send_context *sc)
if (n == ARRAY_SIZE(qps))
goto full;
wait = list_first_entry(list, struct iowait, list);
- qp = container_of(wait, struct hfi1_qp, s_iowait);
- list_del_init(&qp->s_iowait.list);
+ qp = iowait_to_qp(wait);
+ priv = qp->priv;
+ list_del_init(&priv->s_iowait.list);
/* refcount held until actual wake up */
qps[n++] = qp;
}
diff --git a/drivers/staging/rdma/hfi1/qp.c b/drivers/staging/rdma/hfi1/qp.c
index bb447b5..d49b1e9 100644
--- a/drivers/staging/rdma/hfi1/qp.c
+++ b/drivers/staging/rdma/hfi1/qp.c
@@ -349,11 +349,12 @@ bail:
*/
static void reset_qp(struct hfi1_qp *qp, enum ib_qp_type type)
{
+ struct hfi1_qp_priv *priv = qp->priv;
qp->remote_qpn = 0;
qp->qkey = 0;
qp->qp_access_flags = 0;
iowait_init(
- &qp->s_iowait,
+ &priv->s_iowait,
1,
hfi1_do_send,
iowait_sleep,
@@ -460,6 +461,7 @@ static void clear_mr_refs(struct hfi1_qp *qp, int clr_sends)
int hfi1_error_qp(struct hfi1_qp *qp, enum ib_wc_status err)
{
struct hfi1_ibdev *dev = to_idev(qp->ibqp.device);
+ struct hfi1_qp_priv *priv = qp->priv;
struct ib_wc wc;
int ret = 0;
@@ -477,9 +479,9 @@ int hfi1_error_qp(struct hfi1_qp *qp, enum ib_wc_status err)
qp->s_flags &= ~HFI1_S_ANY_WAIT_SEND;
write_seqlock(&dev->iowait_lock);
- if (!list_empty(&qp->s_iowait.list) && !(qp->s_flags & HFI1_S_BUSY)) {
+ if (!list_empty(&priv->s_iowait.list) && !(qp->s_flags & HFI1_S_BUSY)) {
qp->s_flags &= ~HFI1_S_ANY_WAIT_IO;
- list_del_init(&qp->s_iowait.list);
+ list_del_init(&priv->s_iowait.list);
if (atomic_dec_and_test(&qp->refcount))
wake_up(&qp->wait);
}
@@ -544,11 +546,13 @@ bail:
static void flush_tx_list(struct hfi1_qp *qp)
{
- while (!list_empty(&qp->s_iowait.tx_head)) {
+ struct hfi1_qp_priv *priv = qp->priv;
+
+ while (!list_empty(&priv->s_iowait.tx_head)) {
struct sdma_txreq *tx;
tx = list_first_entry(
- &qp->s_iowait.tx_head,
+ &priv->s_iowait.tx_head,
struct sdma_txreq,
list);
list_del_init(&tx->list);
@@ -559,12 +563,13 @@ static void flush_tx_list(struct hfi1_qp *qp)
static void flush_iowait(struct hfi1_qp *qp)
{
+ struct hfi1_qp_priv *priv = qp->priv;
struct hfi1_ibdev *dev = to_idev(qp->ibqp.device);
unsigned long flags;
write_seqlock_irqsave(&dev->iowait_lock, flags);
- if (!list_empty(&qp->s_iowait.list)) {
- list_del_init(&qp->s_iowait.list);
+ if (!list_empty(&priv->s_iowait.list)) {
+ list_del_init(&priv->s_iowait.list);
if (atomic_dec_and_test(&qp->refcount))
wake_up(&qp->wait);
}
@@ -612,6 +617,7 @@ int hfi1_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
{
struct hfi1_ibdev *dev = to_idev(ibqp->device);
struct hfi1_qp *qp = to_iqp(ibqp);
+ struct hfi1_qp_priv *priv = qp->priv;
enum ib_qp_state cur_state, new_state;
struct ib_event ev;
int lastwqe = 0;
@@ -738,9 +744,9 @@ int hfi1_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
spin_unlock(&qp->s_lock);
spin_unlock_irq(&qp->r_lock);
/* Stop the sending work queue and retry timer */
- cancel_work_sync(&qp->s_iowait.iowork);
+ cancel_work_sync(&priv->s_iowait.iowork);
del_timer_sync(&qp->s_timer);
- iowait_sdma_drain(&qp->s_iowait);
+ iowait_sdma_drain(&priv->s_iowait);
flush_tx_list(qp);
remove_qp(dev, qp);
wait_event(qp->wait, !atomic_read(&qp->refcount));
@@ -805,8 +811,8 @@ int hfi1_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
qp->remote_ah_attr = attr->ah_attr;
qp->s_srate = attr->ah_attr.static_rate;
qp->srate_mbps = ib_rate_to_mbps(qp->s_srate);
- qp->s_sc = ah_to_sc(ibqp->device, &qp->remote_ah_attr);
- qp->s_sde = qp_to_sdma_engine(qp, qp->s_sc);
+ priv->s_sc = ah_to_sc(ibqp->device, &qp->remote_ah_attr);
+ qp->s_sde = qp_to_sdma_engine(qp, priv->s_sc);
}
if (attr_mask & IB_QP_ALT_PATH) {
@@ -821,8 +827,8 @@ int hfi1_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
qp->port_num = qp->alt_ah_attr.port_num;
qp->s_pkey_index = qp->s_alt_pkey_index;
qp->s_flags |= HFI1_S_AHG_CLEAR;
- qp->s_sc = ah_to_sc(ibqp->device, &qp->remote_ah_attr);
- qp->s_sde = qp_to_sdma_engine(qp, qp->s_sc);
+ priv->s_sc = ah_to_sc(ibqp->device, &qp->remote_ah_attr);
+ qp->s_sde = qp_to_sdma_engine(qp, priv->s_sc);
}
}
@@ -1031,6 +1037,7 @@ struct ib_qp *hfi1_create_qp(struct ib_pd *ibpd,
struct ib_udata *udata)
{
struct hfi1_qp *qp;
+ struct hfi1_qp_priv *priv;
int err;
struct hfi1_swqe *swq = NULL;
struct hfi1_ibdev *dev;
@@ -1098,11 +1105,18 @@ struct ib_qp *hfi1_create_qp(struct ib_pd *ibpd,
goto bail_swq;
}
RCU_INIT_POINTER(qp->next, NULL);
- qp->s_hdr = kzalloc(sizeof(*qp->s_hdr), GFP_KERNEL);
- if (!qp->s_hdr) {
+ priv = kzalloc(sizeof(*priv), GFP_KERNEL);
+ if (!priv) {
+ ret = ERR_PTR(-ENOMEM);
+ goto bail_qp_priv;
+ }
+ priv->owner = qp;
+ priv->s_hdr = kzalloc(sizeof(*priv->s_hdr), GFP_KERNEL);
+ if (!priv->s_hdr) {
ret = ERR_PTR(-ENOMEM);
goto bail_qp;
}
+ qp->priv = priv;
qp->timeout_jiffies =
usecs_to_jiffies((4096UL * (1UL << qp->timeout)) /
1000UL);
@@ -1245,7 +1259,9 @@ bail_ip:
vfree(qp->r_rq.wq);
free_qpn(&dev->qp_dev->qpn_table, qp->ibqp.qp_num);
bail_qp:
- kfree(qp->s_hdr);
+ kfree(priv->s_hdr);
+ kfree(priv);
+bail_qp_priv:
kfree(qp);
bail_swq:
vfree(swq);
@@ -1266,6 +1282,7 @@ int hfi1_destroy_qp(struct ib_qp *ibqp)
{
struct hfi1_qp *qp = to_iqp(ibqp);
struct hfi1_ibdev *dev = to_idev(ibqp->device);
+ struct hfi1_qp_priv *priv = qp->priv;
/* Make sure HW and driver activity is stopped. */
spin_lock_irq(&qp->r_lock);
@@ -1276,9 +1293,9 @@ int hfi1_destroy_qp(struct ib_qp *ibqp)
qp->s_flags &= ~(HFI1_S_TIMER | HFI1_S_ANY_WAIT);
spin_unlock(&qp->s_lock);
spin_unlock_irq(&qp->r_lock);
- cancel_work_sync(&qp->s_iowait.iowork);
+ cancel_work_sync(&priv->s_iowait.iowork);
del_timer_sync(&qp->s_timer);
- iowait_sdma_drain(&qp->s_iowait);
+ iowait_sdma_drain(&priv->s_iowait);
flush_tx_list(qp);
remove_qp(dev, qp);
wait_event(qp->wait, !atomic_read(&qp->refcount));
@@ -1301,7 +1318,8 @@ int hfi1_destroy_qp(struct ib_qp *ibqp)
else
vfree(qp->r_rq.wq);
vfree(qp->s_wq);
- kfree(qp->s_hdr);
+ kfree(priv->s_hdr);
+ kfree(priv);
kfree(qp);
return 0;
}
@@ -1422,11 +1440,13 @@ static int iowait_sleep(
{
struct verbs_txreq *tx = container_of(stx, struct verbs_txreq, txreq);
struct hfi1_qp *qp;
+ struct hfi1_qp_priv *priv;
unsigned long flags;
int ret = 0;
struct hfi1_ibdev *dev;
qp = tx->qp;
+ priv = qp->priv;
spin_lock_irqsave(&qp->s_lock, flags);
if (ib_hfi1_state_ops[qp->state] & HFI1_PROCESS_RECV_OK) {
@@ -1442,13 +1462,13 @@ static int iowait_sleep(
write_seqlock(&dev->iowait_lock);
if (sdma_progress(sde, seq, stx))
goto eagain;
- if (list_empty(&qp->s_iowait.list)) {
+ if (list_empty(&priv->s_iowait.list)) {
struct hfi1_ibport *ibp =
to_iport(qp->ibqp.device, qp->port_num);
ibp->n_dmawait++;
qp->s_flags |= HFI1_S_WAIT_DMA_DESC;
- list_add_tail(&qp->s_iowait.list, &sde->dmawait);
+ list_add_tail(&priv->s_iowait.list, &sde->dmawait);
trace_hfi1_qpsleep(qp, HFI1_S_WAIT_DMA_DESC);
atomic_inc(&qp->refcount);
}
@@ -1470,7 +1490,7 @@ eagain:
static void iowait_wakeup(struct iowait *wait, int reason)
{
- struct hfi1_qp *qp = container_of(wait, struct hfi1_qp, s_iowait);
+ struct hfi1_qp *qp = iowait_to_qp(wait);
WARN_ON(reason != SDMA_AVAIL_REASON);
hfi1_qp_wakeup(qp, HFI1_S_WAIT_DMA_DESC);
@@ -1651,9 +1671,10 @@ void qp_iter_print(struct seq_file *s, struct qp_iter *iter)
{
struct hfi1_swqe *wqe;
struct hfi1_qp *qp = iter->qp;
+ struct hfi1_qp_priv *priv = qp->priv;
struct sdma_engine *sde;
- sde = qp_to_sdma_engine(qp, qp->s_sc);
+ sde = qp_to_sdma_engine(qp, priv->s_sc);
wqe = get_swqe_ptr(qp, qp->s_last);
seq_printf(s,
"N %d %s QP%u R %u %s %u %u %u f=%x %u %u %u %u %u PSN %x %x %x %x %x (%u %u %u %u %u %u) QP%u LID %x SL %u MTU %d %u %u %u SDE %p,%u\n",
@@ -1666,8 +1687,8 @@ void qp_iter_print(struct seq_file *s, struct qp_iter *iter)
wqe ? wqe->wr.opcode : 0,
qp->s_hdrwords,
qp->s_flags,
- atomic_read(&qp->s_iowait.sdma_busy),
- !list_empty(&qp->s_iowait.list),
+ atomic_read(&priv->s_iowait.sdma_busy),
+ !list_empty(&priv->s_iowait.list),
qp->timeout,
wqe ? wqe->ssn : 0,
qp->s_lsn,
@@ -1707,14 +1728,15 @@ void qp_comm_est(struct hfi1_qp *qp)
void hfi1_migrate_qp(struct hfi1_qp *qp)
{
struct ib_event ev;
+ struct hfi1_qp_priv *priv = qp->priv;
qp->s_mig_state = IB_MIG_MIGRATED;
qp->remote_ah_attr = qp->alt_ah_attr;
qp->port_num = qp->alt_ah_attr.port_num;
qp->s_pkey_index = qp->s_alt_pkey_index;
qp->s_flags |= HFI1_S_AHG_CLEAR;
- qp->s_sc = ah_to_sc(qp->ibqp.device, &qp->remote_ah_attr);
- qp->s_sde = qp_to_sdma_engine(qp, qp->s_sc);
+ priv->s_sc = ah_to_sc(qp->ibqp.device, &qp->remote_ah_attr);
+ qp->s_sde = qp_to_sdma_engine(qp, priv->s_sc);
ev.device = qp->ibqp.device;
ev.element.qp = &qp->ibqp;
diff --git a/drivers/staging/rdma/hfi1/qp.h b/drivers/staging/rdma/hfi1/qp.h
index 94fd748..1d34c22 100644
--- a/drivers/staging/rdma/hfi1/qp.h
+++ b/drivers/staging/rdma/hfi1/qp.h
@@ -123,9 +123,11 @@ static inline struct hfi1_qp *hfi1_lookup_qpn(struct hfi1_ibport *ibp,
*/
static inline void clear_ahg(struct hfi1_qp *qp)
{
- qp->s_hdr->ahgcount = 0;
+ struct hfi1_qp_priv *priv = qp->priv;
+
+ priv->s_hdr->ahgcount = 0;
qp->s_flags &= ~(HFI1_S_AHG_VALID | HFI1_S_AHG_CLEAR);
- if (qp->s_sde && qp->s_ahgidx >= 0)
+ if (qp->s_sde)
sdma_ahg_free(qp->s_sde, qp->s_ahgidx);
qp->s_ahgidx = -1;
}
@@ -257,12 +259,13 @@ void qp_comm_est(struct hfi1_qp *qp);
*/
static inline void _hfi1_schedule_send(struct hfi1_qp *qp)
{
+ struct hfi1_qp_priv *priv = qp->priv;
struct hfi1_ibport *ibp =
to_iport(qp->ibqp.device, qp->port_num);
struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
struct hfi1_devdata *dd = dd_from_ibdev(qp->ibqp.device);
- iowait_schedule(&qp->s_iowait, ppd->hfi1_wq,
+ iowait_schedule(&priv->s_iowait, ppd->hfi1_wq,
qp->s_sde ?
qp->s_sde->cpu :
cpumask_first(cpumask_of_node(dd->assigned_node_id)));
diff --git a/drivers/staging/rdma/hfi1/rc.c b/drivers/staging/rdma/hfi1/rc.c
index 6f4a155..d830a8f 100644
--- a/drivers/staging/rdma/hfi1/rc.c
+++ b/drivers/staging/rdma/hfi1/rc.c
@@ -259,6 +259,7 @@ bail:
*/
int hfi1_make_rc_req(struct hfi1_qp *qp)
{
+ struct hfi1_qp_priv *priv = qp->priv;
struct hfi1_ibdev *dev = to_idev(qp->ibqp.device);
struct hfi1_other_headers *ohdr;
struct hfi1_sge_state *ss;
@@ -275,9 +276,9 @@ int hfi1_make_rc_req(struct hfi1_qp *qp)
int middle = 0;
int delta;
- ohdr = &qp->s_hdr->ibh.u.oth;
+ ohdr = &priv->s_hdr->ibh.u.oth;
if (qp->remote_ah_attr.ah_flags & IB_AH_GRH)
- ohdr = &qp->s_hdr->ibh.u.l.oth;
+ ohdr = &priv->s_hdr->ibh.u.l.oth;
/*
* The lock is needed to synchronize between the sending tasklet,
@@ -297,7 +298,7 @@ int hfi1_make_rc_req(struct hfi1_qp *qp)
if (qp->s_last == qp->s_head)
goto bail;
/* If DMAs are in progress, we can't flush immediately. */
- if (atomic_read(&qp->s_iowait.sdma_busy)) {
+ if (atomic_read(&priv->s_iowait.sdma_busy)) {
qp->s_flags |= HFI1_S_WAIT_DMA;
goto bail;
}
diff --git a/drivers/staging/rdma/hfi1/ruc.c b/drivers/staging/rdma/hfi1/ruc.c
index 1c9c054..736b44d 100644
--- a/drivers/staging/rdma/hfi1/ruc.c
+++ b/drivers/staging/rdma/hfi1/ruc.c
@@ -690,6 +690,7 @@ u32 hfi1_make_grh(struct hfi1_ibport *ibp, struct ib_grh *hdr,
*/
static inline void build_ahg(struct hfi1_qp *qp, u32 npsn)
{
+ struct hfi1_qp_priv *priv = qp->priv;
if (unlikely(qp->s_flags & HFI1_S_AHG_CLEAR))
clear_ahg(qp);
if (!(qp->s_flags & HFI1_S_AHG_VALID)) {
@@ -698,19 +699,19 @@ static inline void build_ahg(struct hfi1_qp *qp, u32 npsn)
qp->s_ahgidx = sdma_ahg_alloc(qp->s_sde);
if (qp->s_ahgidx >= 0) {
qp->s_ahgpsn = npsn;
- qp->s_hdr->tx_flags |= SDMA_TXREQ_F_AHG_COPY;
+ priv->s_hdr->tx_flags |= SDMA_TXREQ_F_AHG_COPY;
/* save to protect a change in another thread */
- qp->s_hdr->sde = qp->s_sde;
- qp->s_hdr->ahgidx = qp->s_ahgidx;
+ priv->s_hdr->sde = qp->s_sde;
+ priv->s_hdr->ahgidx = qp->s_ahgidx;
qp->s_flags |= HFI1_S_AHG_VALID;
}
} else {
/* subsequent middle after valid */
if (qp->s_ahgidx >= 0) {
- qp->s_hdr->tx_flags |= SDMA_TXREQ_F_USE_AHG;
- qp->s_hdr->ahgidx = qp->s_ahgidx;
- qp->s_hdr->ahgcount++;
- qp->s_hdr->ahgdesc[0] =
+ priv->s_hdr->tx_flags |= SDMA_TXREQ_F_USE_AHG;
+ priv->s_hdr->ahgidx = qp->s_ahgidx;
+ priv->s_hdr->ahgcount++;
+ priv->s_hdr->ahgdesc[0] =
sdma_build_ahg_descriptor(
(__force u16)cpu_to_be16((u16)npsn),
BTH2_OFFSET,
@@ -718,8 +719,8 @@ static inline void build_ahg(struct hfi1_qp *qp, u32 npsn)
16);
if ((npsn & 0xffff0000) !=
(qp->s_ahgpsn & 0xffff0000)) {
- qp->s_hdr->ahgcount++;
- qp->s_hdr->ahgdesc[1] =
+ priv->s_hdr->ahgcount++;
+ priv->s_hdr->ahgdesc[1] =
sdma_build_ahg_descriptor(
(__force u16)cpu_to_be16(
(u16)(npsn >> 16)),
@@ -735,6 +736,7 @@ void hfi1_make_ruc_header(struct hfi1_qp *qp, struct hfi1_other_headers *ohdr,
u32 bth0, u32 bth2, int middle)
{
struct hfi1_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
+ struct hfi1_qp_priv *priv = qp->priv;
u16 lrh0;
u32 nwords;
u32 extra_bytes;
@@ -745,13 +747,13 @@ void hfi1_make_ruc_header(struct hfi1_qp *qp, struct hfi1_other_headers *ohdr,
nwords = (qp->s_cur_size + extra_bytes) >> 2;
lrh0 = HFI1_LRH_BTH;
if (unlikely(qp->remote_ah_attr.ah_flags & IB_AH_GRH)) {
- qp->s_hdrwords += hfi1_make_grh(ibp, &qp->s_hdr->ibh.u.l.grh,
- &qp->remote_ah_attr.grh,
- qp->s_hdrwords, nwords);
+ qp->s_hdrwords += hfi1_make_grh(ibp, &priv->s_hdr->ibh.u.l.grh,
+ &qp->remote_ah_attr.grh,
+ qp->s_hdrwords, nwords);
lrh0 = HFI1_LRH_GRH;
middle = 0;
}
- lrh0 |= (qp->s_sc & 0xf) << 12 | (qp->remote_ah_attr.sl & 0xf) << 4;
+ lrh0 |= (priv->s_sc & 0xf) << 12 | (qp->remote_ah_attr.sl & 0xf) << 4;
/*
* reset s_hdr/AHG fields
*
@@ -763,10 +765,10 @@ void hfi1_make_ruc_header(struct hfi1_qp *qp, struct hfi1_other_headers *ohdr,
* build_ahg() will modify as appropriate
* to use the AHG feature.
*/
- qp->s_hdr->tx_flags = 0;
- qp->s_hdr->ahgcount = 0;
- qp->s_hdr->ahgidx = 0;
- qp->s_hdr->sde = NULL;
+ priv->s_hdr->tx_flags = 0;
+ priv->s_hdr->ahgcount = 0;
+ priv->s_hdr->ahgidx = 0;
+ priv->s_hdr->sde = NULL;
if (qp->s_mig_state == IB_MIG_MIGRATED)
bth0 |= IB_BTH_MIG_REQ;
else
@@ -775,11 +777,11 @@ void hfi1_make_ruc_header(struct hfi1_qp *qp, struct hfi1_other_headers *ohdr,
build_ahg(qp, bth2);
else
qp->s_flags &= ~HFI1_S_AHG_VALID;
- qp->s_hdr->ibh.lrh[0] = cpu_to_be16(lrh0);
- qp->s_hdr->ibh.lrh[1] = cpu_to_be16(qp->remote_ah_attr.dlid);
- qp->s_hdr->ibh.lrh[2] =
+ priv->s_hdr->ibh.lrh[0] = cpu_to_be16(lrh0);
+ priv->s_hdr->ibh.lrh[1] = cpu_to_be16(qp->remote_ah_attr.dlid);
+ priv->s_hdr->ibh.lrh[2] =
cpu_to_be16(qp->s_hdrwords + nwords + SIZE_OF_CRC);
- qp->s_hdr->ibh.lrh[3] = cpu_to_be16(ppd_from_ibp(ibp)->lid |
+ priv->s_hdr->ibh.lrh[3] = cpu_to_be16(ppd_from_ibp(ibp)->lid |
qp->remote_ah_attr.src_path_bits);
bth0 |= hfi1_get_pkey(ibp, qp->s_pkey_index);
bth0 |= extra_bytes << 20;
@@ -808,7 +810,7 @@ void hfi1_make_ruc_header(struct hfi1_qp *qp, struct hfi1_other_headers *ohdr,
void hfi1_do_send(struct work_struct *work)
{
struct iowait *wait = container_of(work, struct iowait, iowork);
- struct hfi1_qp *qp = container_of(wait, struct hfi1_qp, s_iowait);
+ struct hfi1_qp *qp = iowait_to_qp(wait);
struct hfi1_pkt_state ps;
int (*make_req)(struct hfi1_qp *qp);
unsigned long flags;
diff --git a/drivers/staging/rdma/hfi1/uc.c b/drivers/staging/rdma/hfi1/uc.c
index 4f2a788..1908a28 100644
--- a/drivers/staging/rdma/hfi1/uc.c
+++ b/drivers/staging/rdma/hfi1/uc.c
@@ -63,6 +63,7 @@
*/
int hfi1_make_uc_req(struct hfi1_qp *qp)
{
+ struct hfi1_qp_priv *priv = qp->priv;
struct hfi1_other_headers *ohdr;
struct hfi1_swqe *wqe;
unsigned long flags;
@@ -82,7 +83,7 @@ int hfi1_make_uc_req(struct hfi1_qp *qp)
if (qp->s_last == qp->s_head)
goto bail;
/* If DMAs are in progress, we can't flush immediately. */
- if (atomic_read(&qp->s_iowait.sdma_busy)) {
+ if (atomic_read(&priv->s_iowait.sdma_busy)) {
qp->s_flags |= HFI1_S_WAIT_DMA;
goto bail;
}
@@ -92,9 +93,9 @@ int hfi1_make_uc_req(struct hfi1_qp *qp)
goto done;
}
- ohdr = &qp->s_hdr->ibh.u.oth;
+ ohdr = &priv->s_hdr->ibh.u.oth;
if (qp->remote_ah_attr.ah_flags & IB_AH_GRH)
- ohdr = &qp->s_hdr->ibh.u.l.oth;
+ ohdr = &priv->s_hdr->ibh.u.l.oth;
/* Get the next send request. */
wqe = get_swqe_ptr(qp, qp->s_cur);
diff --git a/drivers/staging/rdma/hfi1/ud.c b/drivers/staging/rdma/hfi1/ud.c
index 969335a..ba8a557 100644
--- a/drivers/staging/rdma/hfi1/ud.c
+++ b/drivers/staging/rdma/hfi1/ud.c
@@ -265,6 +265,7 @@ drop:
*/
int hfi1_make_ud_req(struct hfi1_qp *qp)
{
+ struct hfi1_qp_priv *priv = qp->priv;
struct hfi1_other_headers *ohdr;
struct ib_ah_attr *ah_attr;
struct hfi1_pportdata *ppd;
@@ -289,7 +290,7 @@ int hfi1_make_ud_req(struct hfi1_qp *qp)
if (qp->s_last == qp->s_head)
goto bail;
/* If DMAs are in progress, we can't flush immediately. */
- if (atomic_read(&qp->s_iowait.sdma_busy)) {
+ if (atomic_read(&priv->s_iowait.sdma_busy)) {
qp->s_flags |= HFI1_S_WAIT_DMA;
goto bail;
}
@@ -323,7 +324,7 @@ int hfi1_make_ud_req(struct hfi1_qp *qp)
* Instead of waiting, we could queue a
* zero length descriptor so we get a callback.
*/
- if (atomic_read(&qp->s_iowait.sdma_busy)) {
+ if (atomic_read(&priv->s_iowait.sdma_busy)) {
qp->s_flags |= HFI1_S_WAIT_DMA;
goto bail;
}
@@ -354,11 +355,11 @@ int hfi1_make_ud_req(struct hfi1_qp *qp)
if (ah_attr->ah_flags & IB_AH_GRH) {
/* Header size in 32-bit words. */
- qp->s_hdrwords += hfi1_make_grh(ibp, &qp->s_hdr->ibh.u.l.grh,
+ qp->s_hdrwords += hfi1_make_grh(ibp, &priv->s_hdr->ibh.u.l.grh,
&ah_attr->grh,
qp->s_hdrwords, nwords);
lrh0 = HFI1_LRH_GRH;
- ohdr = &qp->s_hdr->ibh.u.l.oth;
+ ohdr = &priv->s_hdr->ibh.u.l.oth;
/*
* Don't worry about sending to locally attached multicast
* QPs. It is unspecified by the spec. what happens.
@@ -366,7 +367,7 @@ int hfi1_make_ud_req(struct hfi1_qp *qp)
} else {
/* Header size in 32-bit words. */
lrh0 = HFI1_LRH_BTH;
- ohdr = &qp->s_hdr->ibh.u.oth;
+ ohdr = &priv->s_hdr->ibh.u.oth;
}
if (wqe->wr.opcode == IB_WR_SEND_WITH_IMM) {
qp->s_hdrwords++;
@@ -378,25 +379,25 @@ int hfi1_make_ud_req(struct hfi1_qp *qp)
lrh0 |= (ah_attr->sl & 0xf) << 4;
if (qp->ibqp.qp_type == IB_QPT_SMI) {
lrh0 |= 0xF000; /* Set VL (see ch. 13.5.3.1) */
- qp->s_sc = 0xf;
+ priv->s_sc = 0xf;
} else {
lrh0 |= (sc5 & 0xf) << 12;
- qp->s_sc = sc5;
+ priv->s_sc = sc5;
}
- qp->s_sde = qp_to_sdma_engine(qp, qp->s_sc);
- qp->s_hdr->ibh.lrh[0] = cpu_to_be16(lrh0);
- qp->s_hdr->ibh.lrh[1] = cpu_to_be16(ah_attr->dlid); /* DEST LID */
- qp->s_hdr->ibh.lrh[2] =
+ qp->s_sde = qp_to_sdma_engine(qp, priv->s_sc);
+ priv->s_hdr->ibh.lrh[0] = cpu_to_be16(lrh0);
+ priv->s_hdr->ibh.lrh[1] = cpu_to_be16(ah_attr->dlid); /* DEST LID */
+ priv->s_hdr->ibh.lrh[2] =
cpu_to_be16(qp->s_hdrwords + nwords + SIZE_OF_CRC);
if (ah_attr->dlid == be16_to_cpu(IB_LID_PERMISSIVE))
- qp->s_hdr->ibh.lrh[3] = IB_LID_PERMISSIVE;
+ priv->s_hdr->ibh.lrh[3] = IB_LID_PERMISSIVE;
else {
lid = ppd->lid;
if (lid) {
lid |= ah_attr->src_path_bits & ((1 << ppd->lmc) - 1);
- qp->s_hdr->ibh.lrh[3] = cpu_to_be16(lid);
+ priv->s_hdr->ibh.lrh[3] = cpu_to_be16(lid);
} else
- qp->s_hdr->ibh.lrh[3] = IB_LID_PERMISSIVE;
+ priv->s_hdr->ibh.lrh[3] = IB_LID_PERMISSIVE;
}
if (wqe->wr.send_flags & IB_SEND_SOLICITED)
bth0 |= IB_BTH_SOLICITED;
@@ -416,10 +417,10 @@ int hfi1_make_ud_req(struct hfi1_qp *qp)
qp->qkey : wqe->ud_wr.remote_qkey);
ohdr->u.ud.deth[1] = cpu_to_be32(qp->ibqp.qp_num);
/* disarm any ahg */
- qp->s_hdr->ahgcount = 0;
- qp->s_hdr->ahgidx = 0;
- qp->s_hdr->tx_flags = 0;
- qp->s_hdr->sde = NULL;
+ priv->s_hdr->ahgcount = 0;
+ priv->s_hdr->ahgidx = 0;
+ priv->s_hdr->tx_flags = 0;
+ priv->s_hdr->sde = NULL;
done:
ret = 1;
diff --git a/drivers/staging/rdma/hfi1/verbs.c b/drivers/staging/rdma/hfi1/verbs.c
index 9abf691..b810142 100644
--- a/drivers/staging/rdma/hfi1/verbs.c
+++ b/drivers/staging/rdma/hfi1/verbs.c
@@ -486,6 +486,7 @@ static int post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
struct ib_send_wr **bad_wr)
{
struct hfi1_qp *qp = to_iqp(ibqp);
+ struct hfi1_qp_priv *priv = qp->priv;
int err = 0;
int call_send;
unsigned long flags;
@@ -515,7 +516,7 @@ bail:
if (nreq && !call_send)
_hfi1_schedule_send(qp);
if (nreq && call_send)
- hfi1_do_send(&qp->s_iowait.iowork);
+ hfi1_do_send(&priv->s_iowait.iowork);
return err;
}
@@ -698,12 +699,14 @@ static void mem_timer(unsigned long data)
struct hfi1_qp *qp = NULL;
struct iowait *wait;
unsigned long flags;
+ struct hfi1_qp_priv *priv;
write_seqlock_irqsave(&dev->iowait_lock, flags);
if (!list_empty(list)) {
wait = list_first_entry(list, struct iowait, list);
- qp = container_of(wait, struct hfi1_qp, s_iowait);
- list_del_init(&qp->s_iowait.list);
+ qp = iowait_to_qp(wait);
+ priv = qp->priv;
+ list_del_init(&priv->s_iowait.list);
/* refcount held until actual wake up */
if (!list_empty(list))
mod_timer(&dev->mem_timer, jiffies + 1);
@@ -738,6 +741,7 @@ void update_sge(struct hfi1_sge_state *ss, u32 length)
static noinline struct verbs_txreq *__get_txreq(struct hfi1_ibdev *dev,
struct hfi1_qp *qp)
{
+ struct hfi1_qp_priv *priv = qp->priv;
struct verbs_txreq *tx;
unsigned long flags;
@@ -746,10 +750,10 @@ static noinline struct verbs_txreq *__get_txreq(struct hfi1_ibdev *dev,
spin_lock_irqsave(&qp->s_lock, flags);
write_seqlock(&dev->iowait_lock);
if (ib_hfi1_state_ops[qp->state] & HFI1_PROCESS_RECV_OK &&
- list_empty(&qp->s_iowait.list)) {
+ list_empty(&priv->s_iowait.list)) {
dev->n_txwait++;
qp->s_flags |= HFI1_S_WAIT_TX;
- list_add_tail(&qp->s_iowait.list, &dev->txwait);
+ list_add_tail(&priv->s_iowait.list, &dev->txwait);
trace_hfi1_qpsleep(qp, HFI1_S_WAIT_TX);
atomic_inc(&qp->refcount);
}
@@ -783,6 +787,7 @@ void hfi1_put_txreq(struct verbs_txreq *tx)
struct hfi1_qp *qp;
unsigned long flags;
unsigned int seq;
+ struct hfi1_qp_priv *priv;
qp = tx->qp;
dev = to_idev(qp->ibqp.device);
@@ -805,8 +810,9 @@ void hfi1_put_txreq(struct verbs_txreq *tx)
/* Wake up first QP wanting a free struct */
wait = list_first_entry(&dev->txwait, struct iowait,
list);
- qp = container_of(wait, struct hfi1_qp, s_iowait);
- list_del_init(&qp->s_iowait.list);
+ qp = iowait_to_qp(wait);
+ priv = qp->priv;
+ list_del_init(&priv->s_iowait.list);
/* refcount held until actual wake up */
write_sequnlock_irqrestore(&dev->iowait_lock, flags);
hfi1_qp_wakeup(qp, HFI1_S_WAIT_TX);
@@ -856,17 +862,18 @@ static void verbs_sdma_complete(
static int wait_kmem(struct hfi1_ibdev *dev, struct hfi1_qp *qp)
{
+ struct hfi1_qp_priv *priv = qp->priv;
unsigned long flags;
int ret = 0;
spin_lock_irqsave(&qp->s_lock, flags);
if (ib_hfi1_state_ops[qp->state] & HFI1_PROCESS_RECV_OK) {
write_seqlock(&dev->iowait_lock);
- if (list_empty(&qp->s_iowait.list)) {
+ if (list_empty(&priv->s_iowait.list)) {
if (list_empty(&dev->memwait))
mod_timer(&dev->mem_timer, jiffies + 1);
qp->s_flags |= HFI1_S_WAIT_KMEM;
- list_add_tail(&qp->s_iowait.list, &dev->memwait);
+ list_add_tail(&priv->s_iowait.list, &dev->memwait);
trace_hfi1_qpsleep(qp, HFI1_S_WAIT_KMEM);
atomic_inc(&qp->refcount);
}
@@ -1004,7 +1011,8 @@ bail_txadd:
int hfi1_verbs_send_dma(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
u64 pbc)
{
- struct ahg_ib_header *ahdr = qp->s_hdr;
+ struct hfi1_qp_priv *priv = qp->priv;
+ struct ahg_ib_header *ahdr = priv->s_hdr;
u32 hdrwords = qp->s_hdrwords;
struct hfi1_sge_state *ss = qp->s_cur_sge;
u32 len = qp->s_cur_size;
@@ -1014,17 +1022,17 @@ int hfi1_verbs_send_dma(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
struct verbs_txreq *tx;
struct sdma_txreq *stx;
u64 pbc_flags = 0;
- u8 sc5 = qp->s_sc;
+ u8 sc5 = priv->s_sc;
int ret;
- if (!list_empty(&qp->s_iowait.tx_head)) {
+ if (!list_empty(&priv->s_iowait.tx_head)) {
stx = list_first_entry(
- &qp->s_iowait.tx_head,
+ &priv->s_iowait.tx_head,
struct sdma_txreq,
list);
list_del_init(&stx->list);
tx = container_of(stx, struct verbs_txreq, txreq);
- ret = sdma_send_txreq(tx->sde, &qp->s_iowait, stx);
+ ret = sdma_send_txreq(tx->sde, &priv->s_iowait, stx);
if (unlikely(ret == -ECOMM))
goto bail_ecomm;
return ret;
@@ -1053,7 +1061,7 @@ int hfi1_verbs_send_dma(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
if (unlikely(ret))
goto bail_build;
trace_output_ibhdr(dd_from_ibdev(qp->ibqp.device), &ahdr->ibh);
- ret = sdma_send_txreq(tx->sde, &qp->s_iowait, &tx->txreq);
+ ret = sdma_send_txreq(tx->sde, &priv->s_iowait, &tx->txreq);
if (unlikely(ret == -ECOMM))
goto bail_ecomm;
return ret;
@@ -1075,6 +1083,7 @@ bail_tx:
*/
static int no_bufs_available(struct hfi1_qp *qp, struct send_context *sc)
{
+ struct hfi1_qp_priv *priv = qp->priv;
struct hfi1_devdata *dd = sc->dd;
struct hfi1_ibdev *dev = &dd->verbs_dev;
unsigned long flags;
@@ -1089,14 +1098,14 @@ static int no_bufs_available(struct hfi1_qp *qp, struct send_context *sc)
spin_lock_irqsave(&qp->s_lock, flags);
if (ib_hfi1_state_ops[qp->state] & HFI1_PROCESS_RECV_OK) {
write_seqlock(&dev->iowait_lock);
- if (list_empty(&qp->s_iowait.list)) {
+ if (list_empty(&priv->s_iowait.list)) {
struct hfi1_ibdev *dev = &dd->verbs_dev;
int was_empty;
dev->n_piowait++;
qp->s_flags |= HFI1_S_WAIT_PIO;
was_empty = list_empty(&sc->piowait);
- list_add_tail(&qp->s_iowait.list, &sc->piowait);
+ list_add_tail(&priv->s_iowait.list, &sc->piowait);
trace_hfi1_qpsleep(qp, HFI1_S_WAIT_PIO);
atomic_inc(&qp->refcount);
/* counting: only call wantpiobuf_intr if first user */
@@ -1126,7 +1135,8 @@ struct send_context *qp_to_send_context(struct hfi1_qp *qp, u8 sc5)
int hfi1_verbs_send_pio(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
u64 pbc)
{
- struct ahg_ib_header *ahdr = qp->s_hdr;
+ struct hfi1_qp_priv *priv = qp->priv;
+ struct ahg_ib_header *ahdr = priv->s_hdr;
u32 hdrwords = qp->s_hdrwords;
struct hfi1_sge_state *ss = qp->s_cur_sge;
u32 len = qp->s_cur_size;
@@ -1142,7 +1152,7 @@ int hfi1_verbs_send_pio(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
int wc_status = IB_WC_SUCCESS;
/* vl15 special case taken care of in ud.c */
- sc5 = qp->s_sc;
+ sc5 = priv->s_sc;
sc = qp_to_send_context(qp, sc5);
if (!sc)
@@ -1249,11 +1259,12 @@ static inline int egress_pkey_check(struct hfi1_pportdata *ppd,
struct hfi1_ib_header *hdr,
struct hfi1_qp *qp)
{
+ struct hfi1_qp_priv *priv = qp->priv;
struct hfi1_other_headers *ohdr;
struct hfi1_devdata *dd;
int i = 0;
u16 pkey;
- u8 lnh, sc5 = qp->s_sc;
+ u8 lnh, sc5 = priv->s_sc;
if (!(ppd->part_enforce & HFI1_PART_ENFORCE_OUT))
return 0;
@@ -1312,7 +1323,8 @@ bad:
int hfi1_verbs_send(struct hfi1_qp *qp, struct hfi1_pkt_state *ps)
{
struct hfi1_devdata *dd = dd_from_ibdev(qp->ibqp.device);
- struct ahg_ib_header *ahdr = qp->s_hdr;
+ struct hfi1_qp_priv *priv = qp->priv;
+ struct ahg_ib_header *ahdr = priv->s_hdr;
int ret;
int pio = 0;
unsigned long flags = 0;
diff --git a/drivers/staging/rdma/hfi1/verbs.h b/drivers/staging/rdma/hfi1/verbs.h
index d80111f..e35c918 100644
--- a/drivers/staging/rdma/hfi1/verbs.h
+++ b/drivers/staging/rdma/hfi1/verbs.h
@@ -387,6 +387,18 @@ struct hfi1_ack_entry {
};
/*
+ * hfi1 specific data structures that will be hidden from rvt after the queue
+ * pair is made common
+ */
+struct hfi1_qp;
+struct hfi1_qp_priv {
+ struct ahg_ib_header *s_hdr; /* next packet header to send */
+ u8 s_sc; /* SC[0..4] for next packet */
+ struct iowait s_iowait;
+ struct hfi1_qp *owner;
+};
+
+/*
* Variables prefixed with s_ are for the requester (sender).
* Variables prefixed with r_ are for the responder (receiver).
* Variables prefixed with ack_ are for responder replies.
@@ -396,15 +408,13 @@ struct hfi1_ack_entry {
*/
struct hfi1_qp {
struct ib_qp ibqp;
+ void *priv;
/* read mostly fields above and below */
struct ib_ah_attr remote_ah_attr;
struct ib_ah_attr alt_ah_attr;
struct hfi1_qp __rcu *next; /* link list for QPN hash table */
struct hfi1_swqe *s_wq; /* send work queue */
struct hfi1_mmap_info *ip;
- struct ahg_ib_header *s_hdr; /* next packet header to send */
- /* sc for UC/RC QPs - based on ah for UD */
- u8 s_sc;
unsigned long timeout_jiffies; /* computed from timeout */
enum ib_mtu path_mtu;
@@ -501,8 +511,6 @@ struct hfi1_qp {
struct hfi1_sge_state s_ack_rdma_sge;
struct timer_list s_timer;
- struct iowait s_iowait;
-
struct hfi1_sge r_sg_list[0] /* verified SGEs */
____cacheline_aligned_in_smp;
};
@@ -794,6 +802,14 @@ static inline struct hfi1_ibdev *to_idev(struct ib_device *ibdev)
return container_of(rdi, struct hfi1_ibdev, rdi);
}
+static inline struct hfi1_qp *iowait_to_qp(struct iowait *s_iowait)
+{
+ struct hfi1_qp_priv *priv;
+
+ priv = container_of(s_iowait, struct hfi1_qp_priv, s_iowait);
+ return priv->owner;
+}
+
/*
* Send if not busy or waiting for I/O and either
* a RC response is pending or we can process send work requests.
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [RFC PATCH 07/15] IB/hfi1: Add device specific info prints
2015-12-14 17:27 [RFC PATCH 00/15] staging/rdma/hfi1: Initial patches to add rdmavt support in HFI1 Dennis Dalessandro
2015-12-14 17:28 ` [RFC PATCH 06/15] IB/hfi1: Remove driver specific members from hfi1 qp type Dennis Dalessandro
@ 2015-12-14 17:28 ` Dennis Dalessandro
[not found] ` <20151214172509.11576.19722.stgit-K+u1se/DcYrLESAwzcoQNrvm/XP+8Wra@public.gmane.org>
2015-12-14 17:29 ` [RFC PATCH 15/15] IB/hfi1: Use rdmavt pkey verbs function Dennis Dalessandro
3 siblings, 0 replies; 25+ messages in thread
From: Dennis Dalessandro @ 2015-12-14 17:28 UTC (permalink / raw)
To: devel, linux-rdma; +Cc: gregkh, dledford
Implement get_card_name and get_pci_dev helper functions for rdmavt
for hfi1.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandr@intel.com>
---
drivers/staging/rdma/hfi1/driver.c | 16 ++++++++++++++++
drivers/staging/rdma/hfi1/hfi.h | 2 ++
drivers/staging/rdma/hfi1/verbs.c | 2 ++
3 files changed, 20 insertions(+), 0 deletions(-)
diff --git a/drivers/staging/rdma/hfi1/driver.c b/drivers/staging/rdma/hfi1/driver.c
index 4c52e78..4f0103b 100644
--- a/drivers/staging/rdma/hfi1/driver.c
+++ b/drivers/staging/rdma/hfi1/driver.c
@@ -162,6 +162,22 @@ const char *get_unit_name(int unit)
return iname;
}
+const char *get_card_name(struct rvt_dev_info *rdi)
+{
+ struct hfi1_ibdev *ibdev = container_of(rdi, struct hfi1_ibdev, rdi);
+ struct hfi1_devdata *dd = container_of(ibdev,
+ struct hfi1_devdata, verbs_dev);
+ return get_unit_name(dd->unit);
+}
+
+struct pci_dev *get_pci_dev(struct rvt_dev_info *rdi)
+{
+ struct hfi1_ibdev *ibdev = container_of(rdi, struct hfi1_ibdev, rdi);
+ struct hfi1_devdata *dd = container_of(ibdev,
+ struct hfi1_devdata, verbs_dev);
+ return dd->pcidev;
+}
+
/*
* Return count of units with at least one port ACTIVE.
*/
diff --git a/drivers/staging/rdma/hfi1/hfi.h b/drivers/staging/rdma/hfi1/hfi.h
index c4991be..5925deb 100644
--- a/drivers/staging/rdma/hfi1/hfi.h
+++ b/drivers/staging/rdma/hfi1/hfi.h
@@ -1604,6 +1604,8 @@ int get_platform_config_field(struct hfi1_devdata *dd,
dma_addr_t hfi1_map_page(struct pci_dev *, struct page *, unsigned long,
size_t, int);
const char *get_unit_name(int unit);
+const char *get_card_name(struct rvt_dev_info *rdi);
+struct pci_dev *get_pci_dev(struct rvt_dev_info *rdi);
/*
* Flush write combining store buffers (if present) and perform a write
diff --git a/drivers/staging/rdma/hfi1/verbs.c b/drivers/staging/rdma/hfi1/verbs.c
index b810142..1477d00 100644
--- a/drivers/staging/rdma/hfi1/verbs.c
+++ b/drivers/staging/rdma/hfi1/verbs.c
@@ -2031,6 +2031,8 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
* Fill in rvt info object.
*/
dd->verbs_dev.rdi.driver_f.port_callback = hfi1_create_port_files;
+ dd->verbs_dev.rdi.driver_f.get_card_name = get_card_name;
+ dd->verbs_dev.rdi.driver_f.get_pci_dev = get_pci_dev;
dd->verbs_dev.rdi.dparms.props.max_pd = hfi1_max_pds;
dd->verbs_dev.rdi.flags = (RVT_FLAG_MR_INIT_DRIVER |
RVT_FLAG_QP_INIT_DRIVER |
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [RFC PATCH 08/15] IB/hfi1: Use correct rdmavt header files after move.
[not found] ` <20151214172509.11576.19722.stgit-K+u1se/DcYrLESAwzcoQNrvm/XP+8Wra@public.gmane.org>
` (4 preceding siblings ...)
2015-12-14 17:28 ` [RFC PATCH 05/15] IB/hfi1: Remove MR data structures from hfi1 Dennis Dalessandro
@ 2015-12-14 17:28 ` Dennis Dalessandro
2015-12-14 17:28 ` [RFC PATCH 09/15] IB/hfi1: Use address handle in rdmavt and remove from hfi1 Dennis Dalessandro
` (6 subsequent siblings)
12 siblings, 0 replies; 25+ messages in thread
From: Dennis Dalessandro @ 2015-12-14 17:28 UTC (permalink / raw)
To: devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-rdma-u79uwXL29TY76Z2rM5mHXA
Cc: gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r, dledford-H+wXaHxf7aLQT0dZR+AlfA
Rdmavt split the header files to be based on ibta object. This patch
makes changes in hfi1 to account for the move.
The actual removal of HFI1 code continues in the following patch.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Jubin John <jubin.john-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
drivers/staging/rdma/hfi1/driver.c | 3 ++-
drivers/staging/rdma/hfi1/mad.c | 4 ++--
drivers/staging/rdma/hfi1/qp.c | 5 +++--
drivers/staging/rdma/hfi1/ud.c | 14 +++++++-------
drivers/staging/rdma/hfi1/verbs.c | 4 ++--
5 files changed, 16 insertions(+), 14 deletions(-)
diff --git a/drivers/staging/rdma/hfi1/driver.c b/drivers/staging/rdma/hfi1/driver.c
index 4f0103b..fb52d07 100644
--- a/drivers/staging/rdma/hfi1/driver.c
+++ b/drivers/staging/rdma/hfi1/driver.c
@@ -56,6 +56,7 @@
#include <linux/vmalloc.h>
#include <linux/module.h>
#include <linux/prefetch.h>
+#include <rdma/ib_verbs.h>
#include "hfi.h"
#include "trace.h"
@@ -316,7 +317,7 @@ static void rcv_hdrerr(struct hfi1_ctxtdata *rcd, struct hfi1_pportdata *ppd,
/* Get the destination QP number. */
qp_num = be32_to_cpu(ohdr->bth[1]) & HFI1_QPN_MASK;
- if (lid < HFI1_MULTICAST_LID_BASE) {
+ if (lid < be16_to_cpu(IB_MULTICAST_LID_BASE)) {
struct hfi1_qp *qp;
unsigned long flags;
diff --git a/drivers/staging/rdma/hfi1/mad.c b/drivers/staging/rdma/hfi1/mad.c
index 1c34396..0a3f291 100644
--- a/drivers/staging/rdma/hfi1/mad.c
+++ b/drivers/staging/rdma/hfi1/mad.c
@@ -1096,7 +1096,7 @@ static int __subn_set_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
/* Must be a valid unicast LID address. */
if ((lid == 0 && ls_old > IB_PORT_INIT) ||
- lid >= HFI1_MULTICAST_LID_BASE) {
+ lid >= be16_to_cpu(IB_MULTICAST_LID_BASE)) {
smp->status |= IB_SMP_INVALID_FIELD;
pr_warn("SubnSet(OPA_PortInfo) lid invalid 0x%x\n",
lid);
@@ -1129,7 +1129,7 @@ static int __subn_set_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
/* Must be a valid unicast LID address. */
if ((smlid == 0 && ls_old > IB_PORT_INIT) ||
- smlid >= HFI1_MULTICAST_LID_BASE) {
+ smlid >= be16_to_cpu(IB_MULTICAST_LID_BASE)) {
smp->status |= IB_SMP_INVALID_FIELD;
pr_warn("SubnSet(OPA_PortInfo) smlid invalid 0x%x\n", smlid);
} else if (smlid != ibp->sm_lid || msl != ibp->sm_sl) {
diff --git a/drivers/staging/rdma/hfi1/qp.c b/drivers/staging/rdma/hfi1/qp.c
index d49b1e9..7c356e4 100644
--- a/drivers/staging/rdma/hfi1/qp.c
+++ b/drivers/staging/rdma/hfi1/qp.c
@@ -640,7 +640,7 @@ int hfi1_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
if (attr_mask & IB_QP_AV) {
u8 sc;
- if (attr->ah_attr.dlid >= HFI1_MULTICAST_LID_BASE)
+ if (attr->ah_attr.dlid >= be16_to_cpu(IB_MULTICAST_LID_BASE))
goto inval;
if (hfi1_check_ah(qp->ibqp.device, &attr->ah_attr))
goto inval;
@@ -653,7 +653,8 @@ int hfi1_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
if (attr_mask & IB_QP_ALT_PATH) {
u8 sc;
- if (attr->alt_ah_attr.dlid >= HFI1_MULTICAST_LID_BASE)
+ if (attr->alt_ah_attr.dlid >=
+ be16_to_cpu(IB_MULTICAST_LID_BASE))
goto inval;
if (hfi1_check_ah(qp->ibqp.device, &attr->alt_ah_attr))
goto inval;
diff --git a/drivers/staging/rdma/hfi1/ud.c b/drivers/staging/rdma/hfi1/ud.c
index ba8a557..aad4e49 100644
--- a/drivers/staging/rdma/hfi1/ud.c
+++ b/drivers/staging/rdma/hfi1/ud.c
@@ -243,7 +243,7 @@ static void ud_loopback(struct hfi1_qp *sqp, struct hfi1_swqe *swqe)
wc.slid = ppd->lid | (ah_attr->src_path_bits & ((1 << ppd->lmc) - 1));
/* Check for loopback when the port lid is not set */
if (wc.slid == 0 && sqp->ibqp.qp_type == IB_QPT_GSI)
- wc.slid = HFI1_PERMISSIVE_LID;
+ wc.slid = be16_to_cpu(IB_LID_PERMISSIVE);
wc.sl = ah_attr->sl;
wc.dlid_path_bits = ah_attr->dlid & ((1 << ppd->lmc) - 1);
wc.port_num = qp->port_num;
@@ -311,11 +311,11 @@ int hfi1_make_ud_req(struct hfi1_qp *qp)
ibp = to_iport(qp->ibqp.device, qp->port_num);
ppd = ppd_from_ibp(ibp);
ah_attr = &to_iah(wqe->ud_wr.ah)->attr;
- if (ah_attr->dlid < HFI1_MULTICAST_LID_BASE ||
- ah_attr->dlid == HFI1_PERMISSIVE_LID) {
+ if (ah_attr->dlid < be16_to_cpu(IB_MULTICAST_LID_BASE) ||
+ ah_attr->dlid == be16_to_cpu(IB_LID_PERMISSIVE)) {
lid = ah_attr->dlid & ~((1 << ppd->lmc) - 1);
if (unlikely(!loopback && (lid == ppd->lid ||
- (lid == HFI1_PERMISSIVE_LID &&
+ (lid == be16_to_cpu(IB_LID_PERMISSIVE) &&
qp->ibqp.qp_type == IB_QPT_GSI)))) {
/*
* If DMAs are in progress, we can't generate
@@ -667,8 +667,8 @@ void hfi1_ud_rcv(struct hfi1_packet *packet)
qkey = be32_to_cpu(ohdr->u.ud.deth[0]);
src_qp = be32_to_cpu(ohdr->u.ud.deth[1]) & HFI1_QPN_MASK;
dlid = be16_to_cpu(hdr->lrh[1]);
- is_mcast = (dlid > HFI1_MULTICAST_LID_BASE) &&
- (dlid != HFI1_PERMISSIVE_LID);
+ is_mcast = (dlid > be16_to_cpu(IB_MULTICAST_LID_BASE)) &&
+ (dlid != be16_to_cpu(IB_LID_PERMISSIVE));
bth1 = be32_to_cpu(ohdr->bth[1]);
if (unlikely(bth1 & HFI1_BECN_SMASK)) {
/*
@@ -873,7 +873,7 @@ void hfi1_ud_rcv(struct hfi1_packet *packet)
/*
* Save the LMC lower bits if the destination LID is a unicast LID.
*/
- wc.dlid_path_bits = dlid >= HFI1_MULTICAST_LID_BASE ? 0 :
+ wc.dlid_path_bits = dlid >= be16_to_cpu(IB_MULTICAST_LID_BASE) ? 0 :
dlid & ((1 << ppd_from_ibp(ibp)->lmc) - 1);
wc.port_num = qp->port_num;
/* Signal completion event if the solicited bit is set. */
diff --git a/drivers/staging/rdma/hfi1/verbs.c b/drivers/staging/rdma/hfi1/verbs.c
index 1477d00..644950e 100644
--- a/drivers/staging/rdma/hfi1/verbs.c
+++ b/drivers/staging/rdma/hfi1/verbs.c
@@ -646,8 +646,8 @@ void hfi1_ib_rcv(struct hfi1_packet *packet)
/* Get the destination QP number. */
qp_num = be32_to_cpu(packet->ohdr->bth[1]) & HFI1_QPN_MASK;
lid = be16_to_cpu(hdr->lrh[1]);
- if (unlikely((lid >= HFI1_MULTICAST_LID_BASE) &&
- (lid != HFI1_PERMISSIVE_LID))) {
+ if (unlikely((lid >= be16_to_cpu(IB_MULTICAST_LID_BASE)) &&
+ (lid != be16_to_cpu(IB_LID_PERMISSIVE)))) {
struct hfi1_mcast *mcast;
struct hfi1_mcast_qp *p;
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [RFC PATCH 09/15] IB/hfi1: Use address handle in rdmavt and remove from hfi1
[not found] ` <20151214172509.11576.19722.stgit-K+u1se/DcYrLESAwzcoQNrvm/XP+8Wra@public.gmane.org>
` (5 preceding siblings ...)
2015-12-14 17:28 ` [RFC PATCH 08/15] IB/hfi1: Use correct rdmavt header files after move Dennis Dalessandro
@ 2015-12-14 17:28 ` Dennis Dalessandro
2015-12-14 17:29 ` [RFC PATCH 10/15] IB/hfi1: Implement hfi1 support for AH notification Dennis Dalessandro
` (5 subsequent siblings)
12 siblings, 0 replies; 25+ messages in thread
From: Dennis Dalessandro @ 2015-12-14 17:28 UTC (permalink / raw)
To: devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-rdma-u79uwXL29TY76Z2rM5mHXA
Cc: gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r, dledford-H+wXaHxf7aLQT0dZR+AlfA
Original patch from Kamal Heib <kamalh-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>, split
apart from original and modified to accomodate recent changes
in rdmavt.
Remove AH from hfi1 and use rdmavt version.
Signed-off-by: Kamal Heib <kamalh-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Jubin John <jubin.john-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
drivers/staging/rdma/hfi1/common.h | 2 -
drivers/staging/rdma/hfi1/mad.c | 2 -
drivers/staging/rdma/hfi1/qp.c | 6 +-
drivers/staging/rdma/hfi1/ruc.c | 2 -
drivers/staging/rdma/hfi1/ud.c | 4 +
drivers/staging/rdma/hfi1/verbs.c | 131 ++----------------------------------
drivers/staging/rdma/hfi1/verbs.h | 20 +----
7 files changed, 18 insertions(+), 149 deletions(-)
diff --git a/drivers/staging/rdma/hfi1/common.h b/drivers/staging/rdma/hfi1/common.h
index 5dd9272..b0c415a 100644
--- a/drivers/staging/rdma/hfi1/common.h
+++ b/drivers/staging/rdma/hfi1/common.h
@@ -341,7 +341,6 @@ struct hfi1_message_header {
#define FULL_MGMT_P_KEY 0xFFFF
#define DEFAULT_P_KEY LIM_MGMT_P_KEY
-#define HFI1_PERMISSIVE_LID 0xFFFF
#define HFI1_AETH_CREDIT_SHIFT 24
#define HFI1_AETH_CREDIT_MASK 0x1F
#define HFI1_AETH_CREDIT_INVAL 0x1F
@@ -353,7 +352,6 @@ struct hfi1_message_header {
#define HFI1_BECN_SHIFT 30
#define HFI1_BECN_MASK 1
#define HFI1_BECN_SMASK (1 << HFI1_BECN_SHIFT)
-#define HFI1_MULTICAST_LID_BASE 0xC000
static inline __u64 rhf_to_cpu(const __le32 *rbuf)
{
diff --git a/drivers/staging/rdma/hfi1/mad.c b/drivers/staging/rdma/hfi1/mad.c
index 0a3f291..8e9d1e7 100644
--- a/drivers/staging/rdma/hfi1/mad.c
+++ b/drivers/staging/rdma/hfi1/mad.c
@@ -137,7 +137,7 @@ static void send_trap(struct hfi1_ibport *ibp, void *data, unsigned len)
ret = PTR_ERR(ah);
else {
send_buf->ah = ah;
- ibp->sm_ah = to_iah(ah);
+ ibp->sm_ah = ibah_to_rvtah(ah);
ret = 0;
}
} else
diff --git a/drivers/staging/rdma/hfi1/qp.c b/drivers/staging/rdma/hfi1/qp.c
index 7c356e4..bbe6b4d 100644
--- a/drivers/staging/rdma/hfi1/qp.c
+++ b/drivers/staging/rdma/hfi1/qp.c
@@ -424,7 +424,7 @@ static void clear_mr_refs(struct hfi1_qp *qp, int clr_sends)
if (qp->ibqp.qp_type == IB_QPT_UD ||
qp->ibqp.qp_type == IB_QPT_SMI ||
qp->ibqp.qp_type == IB_QPT_GSI)
- atomic_dec(&to_iah(wqe->ud_wr.ah)->refcount);
+ atomic_dec(&ibah_to_rvtah(wqe->ud_wr.ah)->refcount);
if (++qp->s_last >= qp->s_size)
qp->s_last = 0;
}
@@ -642,7 +642,7 @@ int hfi1_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
if (attr->ah_attr.dlid >= be16_to_cpu(IB_MULTICAST_LID_BASE))
goto inval;
- if (hfi1_check_ah(qp->ibqp.device, &attr->ah_attr))
+ if (rvt_check_ah(qp->ibqp.device, &attr->ah_attr))
goto inval;
sc = ah_to_sc(ibqp->device, &attr->ah_attr);
if (!qp_to_sdma_engine(qp, sc) &&
@@ -656,7 +656,7 @@ int hfi1_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
if (attr->alt_ah_attr.dlid >=
be16_to_cpu(IB_MULTICAST_LID_BASE))
goto inval;
- if (hfi1_check_ah(qp->ibqp.device, &attr->alt_ah_attr))
+ if (rvt_check_ah(qp->ibqp.device, &attr->alt_ah_attr))
goto inval;
if (attr->alt_pkey_index >= hfi1_get_npkeys(dd))
goto inval;
diff --git a/drivers/staging/rdma/hfi1/ruc.c b/drivers/staging/rdma/hfi1/ruc.c
index 736b44d..4108c6a 100644
--- a/drivers/staging/rdma/hfi1/ruc.c
+++ b/drivers/staging/rdma/hfi1/ruc.c
@@ -891,7 +891,7 @@ void hfi1_send_complete(struct hfi1_qp *qp, struct hfi1_swqe *wqe,
if (qp->ibqp.qp_type == IB_QPT_UD ||
qp->ibqp.qp_type == IB_QPT_SMI ||
qp->ibqp.qp_type == IB_QPT_GSI)
- atomic_dec(&to_iah(wqe->ud_wr.ah)->refcount);
+ atomic_dec(&ibah_to_rvtah(wqe->ud_wr.ah)->refcount);
/* See ch. 11.2.4.1 and 10.7.3.1 */
if (!(qp->s_flags & HFI1_S_SIGNAL_REQ_WR) ||
diff --git a/drivers/staging/rdma/hfi1/ud.c b/drivers/staging/rdma/hfi1/ud.c
index aad4e49..24b6077 100644
--- a/drivers/staging/rdma/hfi1/ud.c
+++ b/drivers/staging/rdma/hfi1/ud.c
@@ -98,7 +98,7 @@ static void ud_loopback(struct hfi1_qp *sqp, struct hfi1_swqe *swqe)
goto drop;
}
- ah_attr = &to_iah(swqe->ud_wr.ah)->attr;
+ ah_attr = &ibah_to_rvtah(swqe->ud_wr.ah)->attr;
ppd = ppd_from_ibp(ibp);
if (qp->ibqp.qp_num > 1) {
@@ -310,7 +310,7 @@ int hfi1_make_ud_req(struct hfi1_qp *qp)
/* Construct the header. */
ibp = to_iport(qp->ibqp.device, qp->port_num);
ppd = ppd_from_ibp(ibp);
- ah_attr = &to_iah(wqe->ud_wr.ah)->attr;
+ ah_attr = &ibah_to_rvtah(wqe->ud_wr.ah)->attr;
if (ah_attr->dlid < be16_to_cpu(IB_MULTICAST_LID_BASE) ||
ah_attr->dlid == be16_to_cpu(IB_LID_PERMISSIVE)) {
lid = ah_attr->dlid & ~((1 << ppd->lmc) - 1);
diff --git a/drivers/staging/rdma/hfi1/verbs.c b/drivers/staging/rdma/hfi1/verbs.c
index 644950e..e007c52 100644
--- a/drivers/staging/rdma/hfi1/verbs.c
+++ b/drivers/staging/rdma/hfi1/verbs.c
@@ -455,9 +455,7 @@ static int post_one_send(struct hfi1_qp *qp, struct ib_send_wr *wr)
if (wqe->length > 0x80000000U)
goto bail_inval_free;
} else {
- struct hfi1_ah *ah = to_iah(ud_wr(wr)->ah);
-
- atomic_inc(&ah->refcount);
+ atomic_inc(&ibah_to_rvtah(ud_wr(wr)->ah)->refcount);
}
wqe->ssn = qp->s_ssn++;
qp->s_head = next;
@@ -1614,88 +1612,21 @@ u8 ah_to_sc(struct ib_device *ibdev, struct ib_ah_attr *ah)
return ibp->sl_to_sc[ah->sl];
}
-int hfi1_check_ah(struct ib_device *ibdev, struct ib_ah_attr *ah_attr)
+static int hfi1_check_ah(struct ib_device *ibdev, struct ib_ah_attr *ah_attr)
{
struct hfi1_ibport *ibp;
struct hfi1_pportdata *ppd;
struct hfi1_devdata *dd;
u8 sc5;
- /* A multicast address requires a GRH (see ch. 8.4.1). */
- if (ah_attr->dlid >= HFI1_MULTICAST_LID_BASE &&
- ah_attr->dlid != HFI1_PERMISSIVE_LID &&
- !(ah_attr->ah_flags & IB_AH_GRH))
- goto bail;
- if ((ah_attr->ah_flags & IB_AH_GRH) &&
- ah_attr->grh.sgid_index >= HFI1_GUIDS_PER_PORT)
- goto bail;
- if (ah_attr->dlid == 0)
- goto bail;
- if (ah_attr->port_num < 1 ||
- ah_attr->port_num > ibdev->phys_port_cnt)
- goto bail;
- if (ah_attr->static_rate != IB_RATE_PORT_CURRENT &&
- ib_rate_to_mbps(ah_attr->static_rate) < 0)
- goto bail;
- if (ah_attr->sl >= OPA_MAX_SLS)
- goto bail;
/* test the mapping for validity */
ibp = to_iport(ibdev, ah_attr->port_num);
ppd = ppd_from_ibp(ibp);
sc5 = ibp->sl_to_sc[ah_attr->sl];
dd = dd_from_ppd(ppd);
if (sc_to_vlt(dd, sc5) > num_vls && sc_to_vlt(dd, sc5) != 0xf)
- goto bail;
+ return -EINVAL;
return 0;
-bail:
- return -EINVAL;
-}
-
-/**
- * create_ah - create an address handle
- * @pd: the protection domain
- * @ah_attr: the attributes of the AH
- *
- * This may be called from interrupt context.
- */
-static struct ib_ah *create_ah(struct ib_pd *pd,
- struct ib_ah_attr *ah_attr)
-{
- struct hfi1_ah *ah;
- struct ib_ah *ret;
- struct hfi1_ibdev *dev = to_idev(pd->device);
- unsigned long flags;
-
- if (hfi1_check_ah(pd->device, ah_attr)) {
- ret = ERR_PTR(-EINVAL);
- goto bail;
- }
-
- ah = kmalloc(sizeof(*ah), GFP_ATOMIC);
- if (!ah) {
- ret = ERR_PTR(-ENOMEM);
- goto bail;
- }
-
- spin_lock_irqsave(&dev->n_ahs_lock, flags);
- if (dev->n_ahs_allocated == hfi1_max_ahs) {
- spin_unlock_irqrestore(&dev->n_ahs_lock, flags);
- kfree(ah);
- ret = ERR_PTR(-ENOMEM);
- goto bail;
- }
-
- dev->n_ahs_allocated++;
- spin_unlock_irqrestore(&dev->n_ahs_lock, flags);
-
- /* ib_create_ah() will initialize ah->ibah. */
- ah->attr = *ah_attr;
- atomic_set(&ah->refcount, 0);
-
- ret = &ah->ibah;
-
-bail:
- return ret;
}
struct ib_ah *hfi1_create_qp0_ah(struct hfi1_ibport *ibp, u16 dlid)
@@ -1716,51 +1647,6 @@ struct ib_ah *hfi1_create_qp0_ah(struct hfi1_ibport *ibp, u16 dlid)
}
/**
- * destroy_ah - destroy an address handle
- * @ibah: the AH to destroy
- *
- * This may be called from interrupt context.
- */
-static int destroy_ah(struct ib_ah *ibah)
-{
- struct hfi1_ibdev *dev = to_idev(ibah->device);
- struct hfi1_ah *ah = to_iah(ibah);
- unsigned long flags;
-
- if (atomic_read(&ah->refcount) != 0)
- return -EBUSY;
-
- spin_lock_irqsave(&dev->n_ahs_lock, flags);
- dev->n_ahs_allocated--;
- spin_unlock_irqrestore(&dev->n_ahs_lock, flags);
-
- kfree(ah);
-
- return 0;
-}
-
-static int modify_ah(struct ib_ah *ibah, struct ib_ah_attr *ah_attr)
-{
- struct hfi1_ah *ah = to_iah(ibah);
-
- if (hfi1_check_ah(ibah->device, ah_attr))
- return -EINVAL;
-
- ah->attr = *ah_attr;
-
- return 0;
-}
-
-static int query_ah(struct ib_ah *ibah, struct ib_ah_attr *ah_attr)
-{
- struct hfi1_ah *ah = to_iah(ibah);
-
- *ah_attr = ah->attr;
-
- return 0;
-}
-
-/**
* hfi1_get_npkeys - return the size of the PKEY table for context 0
* @dd: the hfi1_ib device
*/
@@ -1878,7 +1764,6 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
/* Only need to initialize non-zero fields. */
- spin_lock_init(&dev->n_ahs_lock);
spin_lock_init(&dev->n_cqs_lock);
spin_lock_init(&dev->n_qps_lock);
spin_lock_init(&dev->n_srqs_lock);
@@ -1988,10 +1873,10 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
ibdev->dealloc_ucontext = dealloc_ucontext;
ibdev->alloc_pd = NULL;
ibdev->dealloc_pd = NULL;
- ibdev->create_ah = create_ah;
- ibdev->destroy_ah = destroy_ah;
- ibdev->modify_ah = modify_ah;
- ibdev->query_ah = query_ah;
+ ibdev->create_ah = NULL;
+ ibdev->destroy_ah = NULL;
+ ibdev->modify_ah = NULL;
+ ibdev->query_ah = NULL;
ibdev->create_srq = hfi1_create_srq;
ibdev->modify_srq = hfi1_modify_srq;
ibdev->query_srq = hfi1_query_srq;
@@ -2033,6 +1918,8 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
dd->verbs_dev.rdi.driver_f.port_callback = hfi1_create_port_files;
dd->verbs_dev.rdi.driver_f.get_card_name = get_card_name;
dd->verbs_dev.rdi.driver_f.get_pci_dev = get_pci_dev;
+ dd->verbs_dev.rdi.driver_f.check_ah = hfi1_check_ah;
+ dd->verbs_dev.rdi.dparms.props.max_ah = hfi1_max_ahs;
dd->verbs_dev.rdi.dparms.props.max_pd = hfi1_max_pds;
dd->verbs_dev.rdi.flags = (RVT_FLAG_MR_INIT_DRIVER |
RVT_FLAG_QP_INIT_DRIVER |
diff --git a/drivers/staging/rdma/hfi1/verbs.h b/drivers/staging/rdma/hfi1/verbs.h
index e35c918..3f3815c 100644
--- a/drivers/staging/rdma/hfi1/verbs.h
+++ b/drivers/staging/rdma/hfi1/verbs.h
@@ -234,13 +234,6 @@ struct hfi1_mcast {
int n_attached;
};
-/* Address Handle */
-struct hfi1_ah {
- struct ib_ah ibah;
- struct ib_ah_attr attr;
- atomic_t refcount;
-};
-
/*
* This structure is used by hfi1_mmap() to validate an offset
* when an mmap() request is made. The vm_area_struct then uses
@@ -652,8 +645,8 @@ static inline void inc_opstats(
struct hfi1_ibport {
struct hfi1_qp __rcu *qp[2];
struct ib_mad_agent *send_agent; /* agent for SMI (traps) */
- struct hfi1_ah *sm_ah;
- struct hfi1_ah *smi_ah;
+ struct rvt_ah *sm_ah;
+ struct rvt_ah *smi_ah;
struct rb_root mcast_tree;
spinlock_t lock; /* protect changes in this struct */
@@ -735,8 +728,6 @@ struct hfi1_ibdev {
u64 n_kmem_wait;
u64 n_send_schedule;
- u32 n_ahs_allocated; /* number of AHs allocated for device */
- spinlock_t n_ahs_lock;
u32 n_cqs_allocated; /* number of CQs allocated for device */
spinlock_t n_cqs_lock;
u32 n_qps_allocated; /* number of QPs allocated for device */
@@ -774,11 +765,6 @@ static inline struct hfi1_mr *to_imr(struct ib_mr *ibmr)
return container_of(ibmr, struct hfi1_mr, ibmr);
}
-static inline struct hfi1_ah *to_iah(struct ib_ah *ibah)
-{
- return container_of(ibah, struct hfi1_ah, ibah);
-}
-
static inline struct hfi1_cq *to_icq(struct ib_cq *ibcq)
{
return container_of(ibcq, struct hfi1_cq, ibcq);
@@ -925,8 +911,6 @@ void hfi1_rc_hdrerr(
u8 ah_to_sc(struct ib_device *ibdev, struct ib_ah_attr *ah_attr);
-int hfi1_check_ah(struct ib_device *ibdev, struct ib_ah_attr *ah_attr);
-
struct ib_ah *hfi1_create_qp0_ah(struct hfi1_ibport *ibp, u16 dlid);
void hfi1_rc_rnr_retry(unsigned long arg);
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [RFC PATCH 10/15] IB/hfi1: Implement hfi1 support for AH notification
[not found] ` <20151214172509.11576.19722.stgit-K+u1se/DcYrLESAwzcoQNrvm/XP+8Wra@public.gmane.org>
` (6 preceding siblings ...)
2015-12-14 17:28 ` [RFC PATCH 09/15] IB/hfi1: Use address handle in rdmavt and remove from hfi1 Dennis Dalessandro
@ 2015-12-14 17:29 ` Dennis Dalessandro
2015-12-14 17:29 ` [RFC PATCH 11/15] IB/hfi1: Remove hfi1 MR and hfi1 specific qp type Dennis Dalessandro
` (4 subsequent siblings)
12 siblings, 0 replies; 25+ messages in thread
From: Dennis Dalessandro @ 2015-12-14 17:29 UTC (permalink / raw)
To: devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-rdma-u79uwXL29TY76Z2rM5mHXA
Cc: gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r, dledford-H+wXaHxf7aLQT0dZR+AlfA
For OPA devices additional work is required to create an AH.
This patch adds support to set the VL correctly.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
drivers/staging/rdma/hfi1/verbs.c | 24 ++++++++++++++++++++++++
1 files changed, 24 insertions(+), 0 deletions(-)
diff --git a/drivers/staging/rdma/hfi1/verbs.c b/drivers/staging/rdma/hfi1/verbs.c
index e007c52..6d69491 100644
--- a/drivers/staging/rdma/hfi1/verbs.c
+++ b/drivers/staging/rdma/hfi1/verbs.c
@@ -1629,6 +1629,29 @@ static int hfi1_check_ah(struct ib_device *ibdev, struct ib_ah_attr *ah_attr)
return 0;
}
+static void hfi1_notify_new_ah(struct ib_device *ibdev,
+ struct ib_ah_attr *ah_attr,
+ struct rvt_ah *ah)
+{
+ struct hfi1_ibport *ibp;
+ struct hfi1_pportdata *ppd;
+ struct hfi1_devdata *dd;
+ u8 sc5;
+
+ /*
+ * Do not trust reading anything from rvt_ah at this point as it is not
+ * done being setup. We can however modify things which we need to set.
+ */
+
+ ibp = to_iport(ibdev, ah_attr->port_num);
+ ppd = ppd_from_ibp(ibp);
+ sc5 = ibp->sl_to_sc[ah->attr.sl];
+ dd = dd_from_ppd(ppd);
+ ah->vl = sc_to_vlt(dd, sc5);
+ if (ah->vl < num_vls || ah->vl == 15)
+ ah->log_pmtu = ilog2(dd->vld[ah->vl].mtu);
+}
+
struct ib_ah *hfi1_create_qp0_ah(struct hfi1_ibport *ibp, u16 dlid)
{
struct ib_ah_attr attr;
@@ -1919,6 +1942,7 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
dd->verbs_dev.rdi.driver_f.get_card_name = get_card_name;
dd->verbs_dev.rdi.driver_f.get_pci_dev = get_pci_dev;
dd->verbs_dev.rdi.driver_f.check_ah = hfi1_check_ah;
+ dd->verbs_dev.rdi.driver_f.notify_new_ah = hfi1_notify_new_ah;
dd->verbs_dev.rdi.dparms.props.max_ah = hfi1_max_ahs;
dd->verbs_dev.rdi.dparms.props.max_pd = hfi1_max_pds;
dd->verbs_dev.rdi.flags = (RVT_FLAG_MR_INIT_DRIVER |
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [RFC PATCH 11/15] IB/hfi1: Remove hfi1 MR and hfi1 specific qp type
[not found] ` <20151214172509.11576.19722.stgit-K+u1se/DcYrLESAwzcoQNrvm/XP+8Wra@public.gmane.org>
` (7 preceding siblings ...)
2015-12-14 17:29 ` [RFC PATCH 10/15] IB/hfi1: Implement hfi1 support for AH notification Dennis Dalessandro
@ 2015-12-14 17:29 ` Dennis Dalessandro
2015-12-14 17:29 ` [RFC PATCH 12/15] IB/hfi1: Remove srq from hfi1 Dennis Dalessandro
` (3 subsequent siblings)
12 siblings, 0 replies; 25+ messages in thread
From: Dennis Dalessandro @ 2015-12-14 17:29 UTC (permalink / raw)
To: devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-rdma-u79uwXL29TY76Z2rM5mHXA
Cc: gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r, dledford-H+wXaHxf7aLQT0dZR+AlfA
This patch does the actual removal of the queue pair from the hfi1 driver
along with a number of dependent data structures. These were moved to rvt.
It also removes the MR functions to use those in rdmavt.
These two pieces can not reasonably be split apart becuase they depend on
each other.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Jubin John <jubin.john-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
drivers/staging/rdma/hfi1/Makefile | 2
drivers/staging/rdma/hfi1/cq.c | 2
drivers/staging/rdma/hfi1/diag.c | 8
drivers/staging/rdma/hfi1/driver.c | 10 -
drivers/staging/rdma/hfi1/hfi.h | 16 -
drivers/staging/rdma/hfi1/keys.c | 356 ---------------------
drivers/staging/rdma/hfi1/mmap.c | 24 +
drivers/staging/rdma/hfi1/mr.c | 522 -------------------------------
drivers/staging/rdma/hfi1/pio.c | 4
drivers/staging/rdma/hfi1/qp.c | 88 +++--
drivers/staging/rdma/hfi1/qp.h | 41 +-
drivers/staging/rdma/hfi1/rc.c | 116 +++----
drivers/staging/rdma/hfi1/ruc.c | 89 +++--
drivers/staging/rdma/hfi1/sdma.h | 6
drivers/staging/rdma/hfi1/srq.c | 28 +-
drivers/staging/rdma/hfi1/trace.h | 22 +
drivers/staging/rdma/hfi1/uc.c | 10 -
drivers/staging/rdma/hfi1/ud.c | 18 +
drivers/staging/rdma/hfi1/verbs.c | 143 +++-----
drivers/staging/rdma/hfi1/verbs.h | 372 +++-------------------
drivers/staging/rdma/hfi1/verbs_mcast.c | 8
21 files changed, 349 insertions(+), 1536 deletions(-)
delete mode 100644 drivers/staging/rdma/hfi1/keys.c
delete mode 100644 drivers/staging/rdma/hfi1/mr.c
diff --git a/drivers/staging/rdma/hfi1/Makefile b/drivers/staging/rdma/hfi1/Makefile
index 2126b8b..3ba64fe 100644
--- a/drivers/staging/rdma/hfi1/Makefile
+++ b/drivers/staging/rdma/hfi1/Makefile
@@ -8,7 +8,7 @@
obj-$(CONFIG_INFINIBAND_HFI1) += hfi1.o
hfi1-y := chip.o cq.o device.o diag.o driver.o eprom.o file_ops.o firmware.o \
- init.o intr.o keys.o mad.o mmap.o mr.o pcie.o pio.o pio_copy.o \
+ init.o intr.o mad.o mmap.o pcie.o pio.o pio_copy.o \
qp.o qsfp.o rc.o ruc.o sdma.o srq.o sysfs.o trace.o twsi.o \
uc.o ud.o user_pages.o user_sdma.o verbs_mcast.o verbs.o
hfi1-$(CONFIG_DEBUG_FS) += debugfs.o
diff --git a/drivers/staging/rdma/hfi1/cq.c b/drivers/staging/rdma/hfi1/cq.c
index 4f046ff..ffd0e7a 100644
--- a/drivers/staging/rdma/hfi1/cq.c
+++ b/drivers/staging/rdma/hfi1/cq.c
@@ -479,7 +479,7 @@ int hfi1_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata)
if (cq->ip) {
struct hfi1_ibdev *dev = to_idev(ibcq->device);
- struct hfi1_mmap_info *ip = cq->ip;
+ struct rvt_mmap_info *ip = cq->ip;
hfi1_update_mmap_info(dev, ip, sz, wc);
diff --git a/drivers/staging/rdma/hfi1/diag.c b/drivers/staging/rdma/hfi1/diag.c
index e172f2a..49822c1 100644
--- a/drivers/staging/rdma/hfi1/diag.c
+++ b/drivers/staging/rdma/hfi1/diag.c
@@ -1618,7 +1618,7 @@ int snoop_recv_handler(struct hfi1_packet *packet)
/*
* Handle snooping and capturing packets when sdma is being used.
*/
-int snoop_send_dma_handler(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
+int snoop_send_dma_handler(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
u64 pbc)
{
pr_alert("Snooping/Capture of Send DMA Packets Is Not Supported!\n");
@@ -1631,13 +1631,13 @@ int snoop_send_dma_handler(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
* bypass packets. The only way to send a bypass packet currently is to use the
* diagpkt interface. When that interface is enable snoop/capture is not.
*/
-int snoop_send_pio_handler(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
+int snoop_send_pio_handler(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
u64 pbc)
{
struct hfi1_qp_priv *priv = qp->priv;
struct ahg_ib_header *ahdr = priv->s_hdr;
u32 hdrwords = qp->s_hdrwords;
- struct hfi1_sge_state *ss = qp->s_cur_sge;
+ struct rvt_sge_state *ss = qp->s_cur_sge;
u32 len = qp->s_cur_size;
u32 dwords = (len + 3) >> 2;
u32 plen = hdrwords + dwords + 2; /* includes pbc */
@@ -1645,7 +1645,7 @@ int snoop_send_pio_handler(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
struct snoop_packet *s_packet = NULL;
u32 *hdr = (u32 *)&ahdr->ibh;
u32 length = 0;
- struct hfi1_sge_state temp_ss;
+ struct rvt_sge_state temp_ss;
void *data = NULL;
void *data_start = NULL;
int ret;
diff --git a/drivers/staging/rdma/hfi1/driver.c b/drivers/staging/rdma/hfi1/driver.c
index fb52d07..182e05f 100644
--- a/drivers/staging/rdma/hfi1/driver.c
+++ b/drivers/staging/rdma/hfi1/driver.c
@@ -318,7 +318,7 @@ static void rcv_hdrerr(struct hfi1_ctxtdata *rcd, struct hfi1_pportdata *ppd,
/* Get the destination QP number. */
qp_num = be32_to_cpu(ohdr->bth[1]) & HFI1_QPN_MASK;
if (lid < be16_to_cpu(IB_MULTICAST_LID_BASE)) {
- struct hfi1_qp *qp;
+ struct rvt_qp *qp;
unsigned long flags;
rcu_read_lock();
@@ -387,7 +387,7 @@ static void rcv_hdrerr(struct hfi1_ctxtdata *rcd, struct hfi1_pportdata *ppd,
* Only in pre-B0 h/w is the CNP_OPCODE handled
* via this code path (errata 291394).
*/
- struct hfi1_qp *qp = NULL;
+ struct rvt_qp *qp = NULL;
u32 lqpn, rqpn;
u16 rlid;
u8 svc_type, sl, sc5;
@@ -456,7 +456,7 @@ static void prescan_rxq(struct hfi1_packet *packet) {}
#else /* !CONFIG_PRESCAN_RXQ */
static int prescan_receive_queue;
-static void process_ecn(struct hfi1_qp *qp, struct hfi1_ib_header *hdr,
+static void process_ecn(struct rvt_qp *qp, struct hfi1_ib_header *hdr,
struct hfi1_other_headers *ohdr,
u64 rhf, u32 bth1, struct ib_grh *grh)
{
@@ -595,7 +595,7 @@ static void prescan_rxq(struct hfi1_packet *packet)
struct hfi1_ibport *ibp = &rcd->ppd->ibport_data;
__le32 *rhf_addr = (__le32 *) rcd->rcvhdrq + mdata.ps_head +
dd->rhf_offset;
- struct hfi1_qp *qp;
+ struct rvt_qp *qp;
struct hfi1_ib_header *hdr;
struct hfi1_other_headers *ohdr;
struct ib_grh *grh = NULL;
@@ -770,7 +770,7 @@ static inline void process_rcv_qp_work(struct hfi1_packet *packet)
{
struct hfi1_ctxtdata *rcd;
- struct hfi1_qp *qp, *nqp;
+ struct rvt_qp *qp, *nqp;
rcd = packet->rcd;
rcd->head = packet->rhqoff;
diff --git a/drivers/staging/rdma/hfi1/hfi.h b/drivers/staging/rdma/hfi1/hfi.h
index 5925deb..414d920 100644
--- a/drivers/staging/rdma/hfi1/hfi.h
+++ b/drivers/staging/rdma/hfi1/hfi.h
@@ -322,7 +322,7 @@ struct hfi1_packet {
void *hdr;
struct hfi1_ctxtdata *rcd;
__le32 *rhf_addr;
- struct hfi1_qp *qp;
+ struct rvt_qp *qp;
struct hfi1_other_headers *ohdr;
u64 rhf;
u32 maxcnt;
@@ -362,7 +362,7 @@ struct hfi1_snoop_data {
#define HFI1_PORT_SNOOP_MODE 1U
#define HFI1_PORT_CAPTURE_MODE 2U
-struct hfi1_sge_state;
+struct rvt_sge_state;
/*
* Get/Set IB link-level config parameters for f_get/set_ib_cfg()
@@ -1055,9 +1055,9 @@ struct hfi1_devdata {
* Handlers for outgoing data so that snoop/capture does not
* have to have its hooks in the send path
*/
- int (*process_pio_send)(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
+ int (*process_pio_send)(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
u64 pbc);
- int (*process_dma_send)(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
+ int (*process_dma_send)(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
u64 pbc);
void (*pio_inline_send)(struct hfi1_devdata *dd, struct pio_buf *pbuf,
u64 pbc, const void *from, size_t count);
@@ -1218,7 +1218,7 @@ static inline u32 egress_cycles(u32 len, u32 rate)
void set_link_ipg(struct hfi1_pportdata *ppd);
void process_becn(struct hfi1_pportdata *ppd, u8 sl, u16 rlid, u32 lqpn,
u32 rqpn, u8 svc_type);
-void return_cnp(struct hfi1_ibport *ibp, struct hfi1_qp *qp, u32 remote_qpn,
+void return_cnp(struct hfi1_ibport *ibp, struct rvt_qp *qp, u32 remote_qpn,
u32 pkey, u32 slid, u32 dlid, u8 sc5,
const struct ib_grh *old_grh);
@@ -1410,9 +1410,9 @@ void reset_link_credits(struct hfi1_devdata *dd);
void assign_remote_cm_au_table(struct hfi1_devdata *dd, u8 vcu);
int snoop_recv_handler(struct hfi1_packet *packet);
-int snoop_send_dma_handler(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
+int snoop_send_dma_handler(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
u64 pbc);
-int snoop_send_pio_handler(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
+int snoop_send_pio_handler(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
u64 pbc);
void snoop_inline_pio_send(struct hfi1_devdata *dd, struct pio_buf *pbuf,
u64 pbc, const void *from, size_t count);
@@ -1626,7 +1626,7 @@ int process_receive_invalid(struct hfi1_packet *packet);
extern rhf_rcv_function_ptr snoop_rhf_rcv_functions[8];
-void update_sge(struct hfi1_sge_state *ss, u32 length);
+void update_sge(struct rvt_sge_state *ss, u32 length);
/* global module parameter variables */
extern unsigned int hfi1_max_mtu;
diff --git a/drivers/staging/rdma/hfi1/keys.c b/drivers/staging/rdma/hfi1/keys.c
deleted file mode 100644
index ffaaa6f..0000000
--- a/drivers/staging/rdma/hfi1/keys.c
+++ /dev/null
@@ -1,356 +0,0 @@
-/*
- *
- * This file is provided under a dual BSD/GPLv2 license. When using or
- * redistributing this file, you may do so under either license.
- *
- * GPL LICENSE SUMMARY
- *
- * Copyright(c) 2015 Intel Corporation.
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of version 2 of the GNU General Public License as
- * published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- * General Public License for more details.
- *
- * BSD LICENSE
- *
- * Copyright(c) 2015 Intel Corporation.
- *
- * 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.
- * - Neither the name of Intel Corporation nor the names of its
- * contributors may be used to endorse or promote products derived
- * from this software without specific prior written permission.
- *
- * 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.
- *
- */
-
-#include "hfi.h"
-
-/**
- * hfi1_alloc_lkey - allocate an lkey
- * @mr: memory region that this lkey protects
- * @dma_region: 0->normal key, 1->restricted DMA key
- *
- * Returns 0 if successful, otherwise returns -errno.
- *
- * Increments mr reference count as required.
- *
- * Sets the lkey field mr for non-dma regions.
- *
- */
-
-int hfi1_alloc_lkey(struct rvt_mregion *mr, int dma_region)
-{
- unsigned long flags;
- u32 r;
- u32 n;
- int ret = 0;
- struct hfi1_ibdev *dev = to_idev(mr->pd->device);
- struct rvt_lkey_table *rkt = &dev->lk_table;
-
- hfi1_get_mr(mr);
- spin_lock_irqsave(&rkt->lock, flags);
-
- /* special case for dma_mr lkey == 0 */
- if (dma_region) {
- struct rvt_mregion *tmr;
-
- tmr = rcu_access_pointer(dev->dma_mr);
- if (!tmr) {
- rcu_assign_pointer(dev->dma_mr, mr);
- mr->lkey_published = 1;
- } else {
- hfi1_put_mr(mr);
- }
- goto success;
- }
-
- /* Find the next available LKEY */
- r = rkt->next;
- n = r;
- for (;;) {
- if (!rcu_access_pointer(rkt->table[r]))
- break;
- r = (r + 1) & (rkt->max - 1);
- if (r == n)
- goto bail;
- }
- rkt->next = (r + 1) & (rkt->max - 1);
- /*
- * Make sure lkey is never zero which is reserved to indicate an
- * unrestricted LKEY.
- */
- rkt->gen++;
- /*
- * bits are capped in verbs.c to ensure enough bits for
- * generation number
- */
- mr->lkey = (r << (32 - hfi1_lkey_table_size)) |
- ((((1 << (24 - hfi1_lkey_table_size)) - 1) & rkt->gen)
- << 8);
- if (mr->lkey == 0) {
- mr->lkey |= 1 << 8;
- rkt->gen++;
- }
- rcu_assign_pointer(rkt->table[r], mr);
- mr->lkey_published = 1;
-success:
- spin_unlock_irqrestore(&rkt->lock, flags);
-out:
- return ret;
-bail:
- hfi1_put_mr(mr);
- spin_unlock_irqrestore(&rkt->lock, flags);
- ret = -ENOMEM;
- goto out;
-}
-
-/**
- * hfi1_free_lkey - free an lkey
- * @mr: mr to free from tables
- */
-void hfi1_free_lkey(struct rvt_mregion *mr)
-{
- unsigned long flags;
- u32 lkey = mr->lkey;
- u32 r;
- struct hfi1_ibdev *dev = to_idev(mr->pd->device);
- struct rvt_lkey_table *rkt = &dev->lk_table;
- int freed = 0;
-
- spin_lock_irqsave(&rkt->lock, flags);
- if (!mr->lkey_published)
- goto out;
- if (lkey == 0)
- RCU_INIT_POINTER(dev->dma_mr, NULL);
- else {
- r = lkey >> (32 - hfi1_lkey_table_size);
- RCU_INIT_POINTER(rkt->table[r], NULL);
- }
- mr->lkey_published = 0;
- freed++;
-out:
- spin_unlock_irqrestore(&rkt->lock, flags);
- if (freed) {
- synchronize_rcu();
- hfi1_put_mr(mr);
- }
-}
-
-/**
- * hfi1_lkey_ok - check IB SGE for validity and initialize
- * @rkt: table containing lkey to check SGE against
- * @pd: protection domain
- * @isge: outgoing internal SGE
- * @sge: SGE to check
- * @acc: access flags
- *
- * Return 1 if valid and successful, otherwise returns 0.
- *
- * increments the reference count upon success
- *
- * Check the IB SGE for validity and initialize our internal version
- * of it.
- */
-int hfi1_lkey_ok(struct rvt_lkey_table *rkt, struct rvt_pd *pd,
- struct hfi1_sge *isge, struct ib_sge *sge, int acc)
-{
- struct rvt_mregion *mr;
- unsigned n, m;
- size_t off;
-
- /*
- * We use LKEY == zero for kernel virtual addresses
- * (see hfi1_get_dma_mr and dma.c).
- */
- rcu_read_lock();
- if (sge->lkey == 0) {
- struct hfi1_ibdev *dev = to_idev(pd->ibpd.device);
-
- if (pd->user)
- goto bail;
- mr = rcu_dereference(dev->dma_mr);
- if (!mr)
- goto bail;
- atomic_inc(&mr->refcount);
- rcu_read_unlock();
-
- isge->mr = mr;
- isge->vaddr = (void *) sge->addr;
- isge->length = sge->length;
- isge->sge_length = sge->length;
- isge->m = 0;
- isge->n = 0;
- goto ok;
- }
- mr = rcu_dereference(
- rkt->table[(sge->lkey >> (32 - hfi1_lkey_table_size))]);
- if (unlikely(!mr || mr->lkey != sge->lkey || mr->pd != &pd->ibpd))
- goto bail;
-
- off = sge->addr - mr->user_base;
- if (unlikely(sge->addr < mr->user_base ||
- off + sge->length > mr->length ||
- (mr->access_flags & acc) != acc))
- goto bail;
- atomic_inc(&mr->refcount);
- rcu_read_unlock();
-
- off += mr->offset;
- if (mr->page_shift) {
- /*
- page sizes are uniform power of 2 so no loop is necessary
- entries_spanned_by_off is the number of times the loop below
- would have executed.
- */
- size_t entries_spanned_by_off;
-
- entries_spanned_by_off = off >> mr->page_shift;
- off -= (entries_spanned_by_off << mr->page_shift);
- m = entries_spanned_by_off / RVT_SEGSZ;
- n = entries_spanned_by_off % RVT_SEGSZ;
- } else {
- m = 0;
- n = 0;
- while (off >= mr->map[m]->segs[n].length) {
- off -= mr->map[m]->segs[n].length;
- n++;
- if (n >= RVT_SEGSZ) {
- m++;
- n = 0;
- }
- }
- }
- isge->mr = mr;
- isge->vaddr = mr->map[m]->segs[n].vaddr + off;
- isge->length = mr->map[m]->segs[n].length - off;
- isge->sge_length = sge->length;
- isge->m = m;
- isge->n = n;
-ok:
- return 1;
-bail:
- rcu_read_unlock();
- return 0;
-}
-
-/**
- * hfi1_rkey_ok - check the IB virtual address, length, and RKEY
- * @qp: qp for validation
- * @sge: SGE state
- * @len: length of data
- * @vaddr: virtual address to place data
- * @rkey: rkey to check
- * @acc: access flags
- *
- * Return 1 if successful, otherwise 0.
- *
- * increments the reference count upon success
- */
-int hfi1_rkey_ok(struct hfi1_qp *qp, struct hfi1_sge *sge,
- u32 len, u64 vaddr, u32 rkey, int acc)
-{
- struct rvt_lkey_table *rkt = &to_idev(qp->ibqp.device)->lk_table;
- struct rvt_mregion *mr;
- unsigned n, m;
- size_t off;
-
- /*
- * We use RKEY == zero for kernel virtual addresses
- * (see hfi1_get_dma_mr and dma.c).
- */
- rcu_read_lock();
- if (rkey == 0) {
- struct rvt_pd *pd = ibpd_to_rvtpd(qp->ibqp.pd);
- struct hfi1_ibdev *dev = to_idev(pd->ibpd.device);
-
- if (pd->user)
- goto bail;
- mr = rcu_dereference(dev->dma_mr);
- if (!mr)
- goto bail;
- atomic_inc(&mr->refcount);
- rcu_read_unlock();
-
- sge->mr = mr;
- sge->vaddr = (void *) vaddr;
- sge->length = len;
- sge->sge_length = len;
- sge->m = 0;
- sge->n = 0;
- goto ok;
- }
-
- mr = rcu_dereference(
- rkt->table[(rkey >> (32 - hfi1_lkey_table_size))]);
- if (unlikely(!mr || mr->lkey != rkey || qp->ibqp.pd != mr->pd))
- goto bail;
-
- off = vaddr - mr->iova;
- if (unlikely(vaddr < mr->iova || off + len > mr->length ||
- (mr->access_flags & acc) == 0))
- goto bail;
- atomic_inc(&mr->refcount);
- rcu_read_unlock();
-
- off += mr->offset;
- if (mr->page_shift) {
- /*
- page sizes are uniform power of 2 so no loop is necessary
- entries_spanned_by_off is the number of times the loop below
- would have executed.
- */
- size_t entries_spanned_by_off;
-
- entries_spanned_by_off = off >> mr->page_shift;
- off -= (entries_spanned_by_off << mr->page_shift);
- m = entries_spanned_by_off / RVT_SEGSZ;
- n = entries_spanned_by_off % RVT_SEGSZ;
- } else {
- m = 0;
- n = 0;
- while (off >= mr->map[m]->segs[n].length) {
- off -= mr->map[m]->segs[n].length;
- n++;
- if (n >= RVT_SEGSZ) {
- m++;
- n = 0;
- }
- }
- }
- sge->mr = mr;
- sge->vaddr = mr->map[m]->segs[n].vaddr + off;
- sge->length = mr->map[m]->segs[n].length - off;
- sge->sge_length = len;
- sge->m = m;
- sge->n = n;
-ok:
- return 1;
-bail:
- rcu_read_unlock();
- return 0;
-}
diff --git a/drivers/staging/rdma/hfi1/mmap.c b/drivers/staging/rdma/hfi1/mmap.c
index 5173b1c..4ce6be6 100644
--- a/drivers/staging/rdma/hfi1/mmap.c
+++ b/drivers/staging/rdma/hfi1/mmap.c
@@ -59,12 +59,12 @@
/**
* hfi1_release_mmap_info - free mmap info structure
- * @ref: a pointer to the kref within struct hfi1_mmap_info
+ * @ref: a pointer to the kref within struct rvt_mmap_info
*/
void hfi1_release_mmap_info(struct kref *ref)
{
- struct hfi1_mmap_info *ip =
- container_of(ref, struct hfi1_mmap_info, ref);
+ struct rvt_mmap_info *ip =
+ container_of(ref, struct rvt_mmap_info, ref);
struct hfi1_ibdev *dev = to_idev(ip->context->device);
spin_lock_irq(&dev->pending_lock);
@@ -81,14 +81,14 @@ void hfi1_release_mmap_info(struct kref *ref)
*/
static void hfi1_vma_open(struct vm_area_struct *vma)
{
- struct hfi1_mmap_info *ip = vma->vm_private_data;
+ struct rvt_mmap_info *ip = vma->vm_private_data;
kref_get(&ip->ref);
}
static void hfi1_vma_close(struct vm_area_struct *vma)
{
- struct hfi1_mmap_info *ip = vma->vm_private_data;
+ struct rvt_mmap_info *ip = vma->vm_private_data;
kref_put(&ip->ref, hfi1_release_mmap_info);
}
@@ -109,7 +109,7 @@ int hfi1_mmap(struct ib_ucontext *context, struct vm_area_struct *vma)
struct hfi1_ibdev *dev = to_idev(context->device);
unsigned long offset = vma->vm_pgoff << PAGE_SHIFT;
unsigned long size = vma->vm_end - vma->vm_start;
- struct hfi1_mmap_info *ip, *pp;
+ struct rvt_mmap_info *ip, *pp;
int ret = -EINVAL;
/*
@@ -146,11 +146,11 @@ done:
/*
* Allocate information for hfi1_mmap
*/
-struct hfi1_mmap_info *hfi1_create_mmap_info(struct hfi1_ibdev *dev,
- u32 size,
- struct ib_ucontext *context,
- void *obj) {
- struct hfi1_mmap_info *ip;
+struct rvt_mmap_info *hfi1_create_mmap_info(struct hfi1_ibdev *dev,
+ u32 size,
+ struct ib_ucontext *context,
+ void *obj) {
+ struct rvt_mmap_info *ip;
ip = kmalloc(sizeof(*ip), GFP_KERNEL);
if (!ip)
@@ -175,7 +175,7 @@ bail:
return ip;
}
-void hfi1_update_mmap_info(struct hfi1_ibdev *dev, struct hfi1_mmap_info *ip,
+void hfi1_update_mmap_info(struct hfi1_ibdev *dev, struct rvt_mmap_info *ip,
u32 size, void *obj)
{
size = PAGE_ALIGN(size);
diff --git a/drivers/staging/rdma/hfi1/mr.c b/drivers/staging/rdma/hfi1/mr.c
deleted file mode 100644
index 27f8081..0000000
--- a/drivers/staging/rdma/hfi1/mr.c
+++ /dev/null
@@ -1,522 +0,0 @@
-/*
- *
- * This file is provided under a dual BSD/GPLv2 license. When using or
- * redistributing this file, you may do so under either license.
- *
- * GPL LICENSE SUMMARY
- *
- * Copyright(c) 2015 Intel Corporation.
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of version 2 of the GNU General Public License as
- * published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- * General Public License for more details.
- *
- * BSD LICENSE
- *
- * Copyright(c) 2015 Intel Corporation.
- *
- * 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.
- * - Neither the name of Intel Corporation nor the names of its
- * contributors may be used to endorse or promote products derived
- * from this software without specific prior written permission.
- *
- * 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.
- *
- */
-
-#include <rdma/ib_umem.h>
-#include <rdma/ib_smi.h>
-
-#include "hfi.h"
-
-/* Fast memory region */
-struct hfi1_fmr {
- struct ib_fmr ibfmr;
- struct rvt_mregion mr; /* must be last */
-};
-
-static inline struct hfi1_fmr *to_ifmr(struct ib_fmr *ibfmr)
-{
- return container_of(ibfmr, struct hfi1_fmr, ibfmr);
-}
-
-static int init_mregion(struct rvt_mregion *mr, struct ib_pd *pd,
- int count)
-{
- int m, i = 0;
- int rval = 0;
-
- m = (count + RVT_SEGSZ - 1) / RVT_SEGSZ;
- for (; i < m; i++) {
- mr->map[i] = kzalloc(sizeof(*mr->map[0]), GFP_KERNEL);
- if (!mr->map[i])
- goto bail;
- }
- mr->mapsz = m;
- init_completion(&mr->comp);
- /* count returning the ptr to user */
- atomic_set(&mr->refcount, 1);
- mr->pd = pd;
- mr->max_segs = count;
-out:
- return rval;
-bail:
- while (i)
- kfree(mr->map[--i]);
- rval = -ENOMEM;
- goto out;
-}
-
-static void deinit_mregion(struct rvt_mregion *mr)
-{
- int i = mr->mapsz;
-
- mr->mapsz = 0;
- while (i)
- kfree(mr->map[--i]);
-}
-
-
-/**
- * hfi1_get_dma_mr - get a DMA memory region
- * @pd: protection domain for this memory region
- * @acc: access flags
- *
- * Returns the memory region on success, otherwise returns an errno.
- * Note that all DMA addresses should be created via the
- * struct ib_dma_mapping_ops functions (see dma.c).
- */
-struct ib_mr *hfi1_get_dma_mr(struct ib_pd *pd, int acc)
-{
- struct hfi1_mr *mr = NULL;
- struct ib_mr *ret;
- int rval;
-
- if (ibpd_to_rvtpd(pd)->user) {
- ret = ERR_PTR(-EPERM);
- goto bail;
- }
-
- mr = kzalloc(sizeof(*mr), GFP_KERNEL);
- if (!mr) {
- ret = ERR_PTR(-ENOMEM);
- goto bail;
- }
-
- rval = init_mregion(&mr->mr, pd, 0);
- if (rval) {
- ret = ERR_PTR(rval);
- goto bail;
- }
-
-
- rval = hfi1_alloc_lkey(&mr->mr, 1);
- if (rval) {
- ret = ERR_PTR(rval);
- goto bail_mregion;
- }
-
- mr->mr.access_flags = acc;
- ret = &mr->ibmr;
-done:
- return ret;
-
-bail_mregion:
- deinit_mregion(&mr->mr);
-bail:
- kfree(mr);
- goto done;
-}
-
-static struct hfi1_mr *alloc_mr(int count, struct ib_pd *pd)
-{
- struct hfi1_mr *mr;
- int rval = -ENOMEM;
- int m;
-
- /* Allocate struct plus pointers to first level page tables. */
- m = (count + RVT_SEGSZ - 1) / RVT_SEGSZ;
- mr = kzalloc(sizeof(*mr) + m * sizeof(mr->mr.map[0]), GFP_KERNEL);
- if (!mr)
- goto bail;
-
- rval = init_mregion(&mr->mr, pd, count);
- if (rval)
- goto bail;
- /*
- * ib_reg_phys_mr() will initialize mr->ibmr except for
- * lkey and rkey.
- */
- rval = hfi1_alloc_lkey(&mr->mr, 0);
- if (rval)
- goto bail_mregion;
- mr->ibmr.lkey = mr->mr.lkey;
- mr->ibmr.rkey = mr->mr.lkey;
-done:
- return mr;
-
-bail_mregion:
- deinit_mregion(&mr->mr);
-bail:
- kfree(mr);
- mr = ERR_PTR(rval);
- goto done;
-}
-
-/**
- * hfi1_reg_phys_mr - register a physical memory region
- * @pd: protection domain for this memory region
- * @buffer_list: pointer to the list of physical buffers to register
- * @num_phys_buf: the number of physical buffers to register
- * @iova_start: the starting address passed over IB which maps to this MR
- *
- * Returns the memory region on success, otherwise returns an errno.
- */
-struct ib_mr *hfi1_reg_phys_mr(struct ib_pd *pd,
- struct ib_phys_buf *buffer_list,
- int num_phys_buf, int acc, u64 *iova_start)
-{
- struct hfi1_mr *mr;
- int n, m, i;
- struct ib_mr *ret;
-
- mr = alloc_mr(num_phys_buf, pd);
- if (IS_ERR(mr)) {
- ret = (struct ib_mr *)mr;
- goto bail;
- }
-
- mr->mr.user_base = *iova_start;
- mr->mr.iova = *iova_start;
- mr->mr.access_flags = acc;
-
- m = 0;
- n = 0;
- for (i = 0; i < num_phys_buf; i++) {
- mr->mr.map[m]->segs[n].vaddr = (void *) buffer_list[i].addr;
- mr->mr.map[m]->segs[n].length = buffer_list[i].size;
- mr->mr.length += buffer_list[i].size;
- n++;
- if (n == RVT_SEGSZ) {
- m++;
- n = 0;
- }
- }
-
- ret = &mr->ibmr;
-
-bail:
- return ret;
-}
-
-/**
- * hfi1_reg_user_mr - register a userspace memory region
- * @pd: protection domain for this memory region
- * @start: starting userspace address
- * @length: length of region to register
- * @mr_access_flags: access flags for this memory region
- * @udata: unused by the driver
- *
- * Returns the memory region on success, otherwise returns an errno.
- */
-struct ib_mr *hfi1_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
- u64 virt_addr, int mr_access_flags,
- struct ib_udata *udata)
-{
- struct hfi1_mr *mr;
- struct ib_umem *umem;
- struct scatterlist *sg;
- int n, m, entry;
- struct ib_mr *ret;
-
- if (length == 0) {
- ret = ERR_PTR(-EINVAL);
- goto bail;
- }
-
- umem = ib_umem_get(pd->uobject->context, start, length,
- mr_access_flags, 0);
- if (IS_ERR(umem))
- return (void *) umem;
-
- n = umem->nmap;
-
- mr = alloc_mr(n, pd);
- if (IS_ERR(mr)) {
- ret = (struct ib_mr *)mr;
- ib_umem_release(umem);
- goto bail;
- }
-
- mr->mr.user_base = start;
- mr->mr.iova = virt_addr;
- mr->mr.length = length;
- mr->mr.offset = ib_umem_offset(umem);
- mr->mr.access_flags = mr_access_flags;
- mr->umem = umem;
-
- if (is_power_of_2(umem->page_size))
- mr->mr.page_shift = ilog2(umem->page_size);
- m = 0;
- n = 0;
- for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) {
- void *vaddr;
-
- vaddr = page_address(sg_page(sg));
- if (!vaddr) {
- ret = ERR_PTR(-EINVAL);
- goto bail;
- }
- mr->mr.map[m]->segs[n].vaddr = vaddr;
- mr->mr.map[m]->segs[n].length = umem->page_size;
- n++;
- if (n == RVT_SEGSZ) {
- m++;
- n = 0;
- }
- }
- ret = &mr->ibmr;
-
-bail:
- return ret;
-}
-
-/**
- * hfi1_dereg_mr - unregister and free a memory region
- * @ibmr: the memory region to free
- *
- * Returns 0 on success.
- *
- * Note that this is called to free MRs created by hfi1_get_dma_mr()
- * or hfi1_reg_user_mr().
- */
-int hfi1_dereg_mr(struct ib_mr *ibmr)
-{
- struct hfi1_mr *mr = to_imr(ibmr);
- int ret = 0;
- unsigned long timeout;
-
- hfi1_free_lkey(&mr->mr);
-
- hfi1_put_mr(&mr->mr); /* will set completion if last */
- timeout = wait_for_completion_timeout(&mr->mr.comp,
- 5 * HZ);
- if (!timeout) {
- dd_dev_err(
- dd_from_ibdev(mr->mr.pd->device),
- "hfi1_dereg_mr timeout mr %p pd %p refcount %u\n",
- mr, mr->mr.pd, atomic_read(&mr->mr.refcount));
- hfi1_get_mr(&mr->mr);
- ret = -EBUSY;
- goto out;
- }
- deinit_mregion(&mr->mr);
- if (mr->umem)
- ib_umem_release(mr->umem);
- kfree(mr);
-out:
- return ret;
-}
-
-/*
- * Allocate a memory region usable with the
- * IB_WR_REG_MR send work request.
- *
- * Return the memory region on success, otherwise return an errno.
- * FIXME: IB_WR_REG_MR is not supported
- */
-struct ib_mr *hfi1_alloc_mr(struct ib_pd *pd,
- enum ib_mr_type mr_type,
- u32 max_num_sg)
-{
- struct hfi1_mr *mr;
-
- if (mr_type != IB_MR_TYPE_MEM_REG)
- return ERR_PTR(-EINVAL);
-
- mr = alloc_mr(max_num_sg, pd);
- if (IS_ERR(mr))
- return (struct ib_mr *)mr;
-
- return &mr->ibmr;
-}
-
-/**
- * hfi1_alloc_fmr - allocate a fast memory region
- * @pd: the protection domain for this memory region
- * @mr_access_flags: access flags for this memory region
- * @fmr_attr: fast memory region attributes
- *
- * Returns the memory region on success, otherwise returns an errno.
- */
-struct ib_fmr *hfi1_alloc_fmr(struct ib_pd *pd, int mr_access_flags,
- struct ib_fmr_attr *fmr_attr)
-{
- struct hfi1_fmr *fmr;
- int m;
- struct ib_fmr *ret;
- int rval = -ENOMEM;
-
- /* Allocate struct plus pointers to first level page tables. */
- m = (fmr_attr->max_pages + RVT_SEGSZ - 1) / RVT_SEGSZ;
- fmr = kzalloc(sizeof(*fmr) + m * sizeof(fmr->mr.map[0]), GFP_KERNEL);
- if (!fmr)
- goto bail;
-
- rval = init_mregion(&fmr->mr, pd, fmr_attr->max_pages);
- if (rval)
- goto bail;
-
- /*
- * ib_alloc_fmr() will initialize fmr->ibfmr except for lkey &
- * rkey.
- */
- rval = hfi1_alloc_lkey(&fmr->mr, 0);
- if (rval)
- goto bail_mregion;
- fmr->ibfmr.rkey = fmr->mr.lkey;
- fmr->ibfmr.lkey = fmr->mr.lkey;
- /*
- * Resources are allocated but no valid mapping (RKEY can't be
- * used).
- */
- fmr->mr.access_flags = mr_access_flags;
- fmr->mr.max_segs = fmr_attr->max_pages;
- fmr->mr.page_shift = fmr_attr->page_shift;
-
- ret = &fmr->ibfmr;
-done:
- return ret;
-
-bail_mregion:
- deinit_mregion(&fmr->mr);
-bail:
- kfree(fmr);
- ret = ERR_PTR(rval);
- goto done;
-}
-
-/**
- * hfi1_map_phys_fmr - set up a fast memory region
- * @ibmfr: the fast memory region to set up
- * @page_list: the list of pages to associate with the fast memory region
- * @list_len: the number of pages to associate with the fast memory region
- * @iova: the virtual address of the start of the fast memory region
- *
- * This may be called from interrupt context.
- */
-
-int hfi1_map_phys_fmr(struct ib_fmr *ibfmr, u64 *page_list,
- int list_len, u64 iova)
-{
- struct hfi1_fmr *fmr = to_ifmr(ibfmr);
- struct rvt_lkey_table *rkt;
- unsigned long flags;
- int m, n, i;
- u32 ps;
- int ret;
-
- i = atomic_read(&fmr->mr.refcount);
- if (i > 2)
- return -EBUSY;
-
- if (list_len > fmr->mr.max_segs) {
- ret = -EINVAL;
- goto bail;
- }
- rkt = &to_idev(ibfmr->device)->lk_table;
- spin_lock_irqsave(&rkt->lock, flags);
- fmr->mr.user_base = iova;
- fmr->mr.iova = iova;
- ps = 1 << fmr->mr.page_shift;
- fmr->mr.length = list_len * ps;
- m = 0;
- n = 0;
- for (i = 0; i < list_len; i++) {
- fmr->mr.map[m]->segs[n].vaddr = (void *) page_list[i];
- fmr->mr.map[m]->segs[n].length = ps;
- if (++n == RVT_SEGSZ) {
- m++;
- n = 0;
- }
- }
- spin_unlock_irqrestore(&rkt->lock, flags);
- ret = 0;
-
-bail:
- return ret;
-}
-
-/**
- * hfi1_unmap_fmr - unmap fast memory regions
- * @fmr_list: the list of fast memory regions to unmap
- *
- * Returns 0 on success.
- */
-int hfi1_unmap_fmr(struct list_head *fmr_list)
-{
- struct hfi1_fmr *fmr;
- struct rvt_lkey_table *rkt;
- unsigned long flags;
-
- list_for_each_entry(fmr, fmr_list, ibfmr.list) {
- rkt = &to_idev(fmr->ibfmr.device)->lk_table;
- spin_lock_irqsave(&rkt->lock, flags);
- fmr->mr.user_base = 0;
- fmr->mr.iova = 0;
- fmr->mr.length = 0;
- spin_unlock_irqrestore(&rkt->lock, flags);
- }
- return 0;
-}
-
-/**
- * hfi1_dealloc_fmr - deallocate a fast memory region
- * @ibfmr: the fast memory region to deallocate
- *
- * Returns 0 on success.
- */
-int hfi1_dealloc_fmr(struct ib_fmr *ibfmr)
-{
- struct hfi1_fmr *fmr = to_ifmr(ibfmr);
- int ret = 0;
- unsigned long timeout;
-
- hfi1_free_lkey(&fmr->mr);
- hfi1_put_mr(&fmr->mr); /* will set completion if last */
- timeout = wait_for_completion_timeout(&fmr->mr.comp,
- 5 * HZ);
- if (!timeout) {
- hfi1_get_mr(&fmr->mr);
- ret = -EBUSY;
- goto out;
- }
- deinit_mregion(&fmr->mr);
- kfree(fmr);
-out:
- return ret;
-}
diff --git a/drivers/staging/rdma/hfi1/pio.c b/drivers/staging/rdma/hfi1/pio.c
index b5bce5e..1852c4b 100644
--- a/drivers/staging/rdma/hfi1/pio.c
+++ b/drivers/staging/rdma/hfi1/pio.c
@@ -1496,8 +1496,8 @@ static void sc_piobufavail(struct send_context *sc)
struct hfi1_devdata *dd = sc->dd;
struct hfi1_ibdev *dev = &dd->verbs_dev;
struct list_head *list;
- struct hfi1_qp *qps[PIO_WAIT_BATCH_SIZE];
- struct hfi1_qp *qp;
+ struct rvt_qp *qps[PIO_WAIT_BATCH_SIZE];
+ struct rvt_qp *qp;
struct hfi1_qp_priv *priv;
unsigned long flags;
unsigned i, n = 0;
diff --git a/drivers/staging/rdma/hfi1/qp.c b/drivers/staging/rdma/hfi1/qp.c
index bbe6b4d..b82855f 100644
--- a/drivers/staging/rdma/hfi1/qp.c
+++ b/drivers/staging/rdma/hfi1/qp.c
@@ -67,7 +67,7 @@ static unsigned int hfi1_qp_table_size = 256;
module_param_named(qp_table_size, hfi1_qp_table_size, uint, S_IRUGO);
MODULE_PARM_DESC(qp_table_size, "QP table size");
-static void flush_tx_list(struct hfi1_qp *qp);
+static void flush_tx_list(struct rvt_qp *qp);
static int iowait_sleep(
struct sdma_engine *sde,
struct iowait *wait,
@@ -229,7 +229,7 @@ static void free_qpn(struct hfi1_qpn_table *qpt, u32 qpn)
* Put the QP into the hash table.
* The hash table holds a reference to the QP.
*/
-static void insert_qp(struct hfi1_ibdev *dev, struct hfi1_qp *qp)
+static void insert_qp(struct hfi1_ibdev *dev, struct rvt_qp *qp)
{
struct hfi1_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
unsigned long flags;
@@ -254,7 +254,7 @@ static void insert_qp(struct hfi1_ibdev *dev, struct hfi1_qp *qp)
* Remove the QP from the table so it can't be found asynchronously by
* the receive interrupt routine.
*/
-static void remove_qp(struct hfi1_ibdev *dev, struct hfi1_qp *qp)
+static void remove_qp(struct hfi1_ibdev *dev, struct rvt_qp *qp)
{
struct hfi1_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
u32 n = qpn_hash(dev->qp_dev, qp->ibqp.qp_num);
@@ -270,8 +270,8 @@ static void remove_qp(struct hfi1_ibdev *dev, struct hfi1_qp *qp)
lockdep_is_held(&dev->qp_dev->qpt_lock)) == qp) {
RCU_INIT_POINTER(ibp->qp[1], NULL);
} else {
- struct hfi1_qp *q;
- struct hfi1_qp __rcu **qpp;
+ struct rvt_qp *q;
+ struct rvt_qp __rcu **qpp;
removed = 0;
qpp = &dev->qp_dev->qp_table[n];
@@ -308,7 +308,7 @@ static unsigned free_all_qps(struct hfi1_devdata *dd)
{
struct hfi1_ibdev *dev = &dd->verbs_dev;
unsigned long flags;
- struct hfi1_qp *qp;
+ struct rvt_qp *qp;
unsigned n, qp_inuse = 0;
for (n = 0; n < dd->num_pports; n++) {
@@ -347,7 +347,7 @@ bail:
* @qp: the QP to reset
* @type: the QP type
*/
-static void reset_qp(struct hfi1_qp *qp, enum ib_qp_type type)
+static void reset_qp(struct rvt_qp *qp, enum ib_qp_type type)
{
struct hfi1_qp_priv *priv = qp->priv;
qp->remote_qpn = 0;
@@ -402,7 +402,7 @@ static void reset_qp(struct hfi1_qp *qp, enum ib_qp_type type)
qp->r_sge.num_sge = 0;
}
-static void clear_mr_refs(struct hfi1_qp *qp, int clr_sends)
+static void clear_mr_refs(struct rvt_qp *qp, int clr_sends)
{
unsigned n;
@@ -413,13 +413,13 @@ static void clear_mr_refs(struct hfi1_qp *qp, int clr_sends)
if (clr_sends) {
while (qp->s_last != qp->s_head) {
- struct hfi1_swqe *wqe = get_swqe_ptr(qp, qp->s_last);
+ struct rvt_swqe *wqe = get_swqe_ptr(qp, qp->s_last);
unsigned i;
for (i = 0; i < wqe->wr.num_sge; i++) {
- struct hfi1_sge *sge = &wqe->sg_list[i];
+ struct rvt_sge *sge = &wqe->sg_list[i];
- hfi1_put_mr(sge->mr);
+ rvt_put_mr(sge->mr);
}
if (qp->ibqp.qp_type == IB_QPT_UD ||
qp->ibqp.qp_type == IB_QPT_SMI ||
@@ -429,7 +429,7 @@ static void clear_mr_refs(struct hfi1_qp *qp, int clr_sends)
qp->s_last = 0;
}
if (qp->s_rdma_mr) {
- hfi1_put_mr(qp->s_rdma_mr);
+ rvt_put_mr(qp->s_rdma_mr);
qp->s_rdma_mr = NULL;
}
}
@@ -438,11 +438,11 @@ static void clear_mr_refs(struct hfi1_qp *qp, int clr_sends)
return;
for (n = 0; n < ARRAY_SIZE(qp->s_ack_queue); n++) {
- struct hfi1_ack_entry *e = &qp->s_ack_queue[n];
+ struct rvt_ack_entry *e = &qp->s_ack_queue[n];
if (e->opcode == IB_OPCODE_RC_RDMA_READ_REQUEST &&
e->rdma_sge.mr) {
- hfi1_put_mr(e->rdma_sge.mr);
+ rvt_put_mr(e->rdma_sge.mr);
e->rdma_sge.mr = NULL;
}
}
@@ -458,7 +458,7 @@ static void clear_mr_refs(struct hfi1_qp *qp, int clr_sends)
* The QP r_lock and s_lock should be held and interrupts disabled.
* If we are already in error state, just return.
*/
-int hfi1_error_qp(struct hfi1_qp *qp, enum ib_wc_status err)
+int hfi1_error_qp(struct rvt_qp *qp, enum ib_wc_status err)
{
struct hfi1_ibdev *dev = to_idev(qp->ibqp.device);
struct hfi1_qp_priv *priv = qp->priv;
@@ -490,7 +490,7 @@ int hfi1_error_qp(struct hfi1_qp *qp, enum ib_wc_status err)
if (!(qp->s_flags & HFI1_S_BUSY)) {
qp->s_hdrwords = 0;
if (qp->s_rdma_mr) {
- hfi1_put_mr(qp->s_rdma_mr);
+ rvt_put_mr(qp->s_rdma_mr);
qp->s_rdma_mr = NULL;
}
flush_tx_list(qp);
@@ -514,7 +514,7 @@ int hfi1_error_qp(struct hfi1_qp *qp, enum ib_wc_status err)
wc.status = IB_WC_WR_FLUSH_ERR;
if (qp->r_rq.wq) {
- struct hfi1_rwq *wq;
+ struct rvt_rwq *wq;
u32 head;
u32 tail;
@@ -544,7 +544,7 @@ bail:
return ret;
}
-static void flush_tx_list(struct hfi1_qp *qp)
+static void flush_tx_list(struct rvt_qp *qp)
{
struct hfi1_qp_priv *priv = qp->priv;
@@ -561,7 +561,7 @@ static void flush_tx_list(struct hfi1_qp *qp)
}
}
-static void flush_iowait(struct hfi1_qp *qp)
+static void flush_iowait(struct rvt_qp *qp)
{
struct hfi1_qp_priv *priv = qp->priv;
struct hfi1_ibdev *dev = to_idev(qp->ibqp.device);
@@ -616,7 +616,7 @@ int hfi1_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
int attr_mask, struct ib_udata *udata)
{
struct hfi1_ibdev *dev = to_idev(ibqp->device);
- struct hfi1_qp *qp = to_iqp(ibqp);
+ struct rvt_qp *qp = to_iqp(ibqp);
struct hfi1_qp_priv *priv = qp->priv;
enum ib_qp_state cur_state, new_state;
struct ib_event ev;
@@ -915,7 +915,7 @@ bail:
int hfi1_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
int attr_mask, struct ib_qp_init_attr *init_attr)
{
- struct hfi1_qp *qp = to_iqp(ibqp);
+ struct rvt_qp *qp = to_iqp(ibqp);
attr->qp_state = qp->state;
attr->cur_qp_state = attr->qp_state;
@@ -968,7 +968,7 @@ int hfi1_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
*
* Returns the AETH.
*/
-__be32 hfi1_compute_aeth(struct hfi1_qp *qp)
+__be32 hfi1_compute_aeth(struct rvt_qp *qp)
{
u32 aeth = qp->r_msn & HFI1_MSN_MASK;
@@ -981,7 +981,7 @@ __be32 hfi1_compute_aeth(struct hfi1_qp *qp)
} else {
u32 min, max, x;
u32 credits;
- struct hfi1_rwq *wq = qp->r_rq.wq;
+ struct rvt_rwq *wq = qp->r_rq.wq;
u32 head;
u32 tail;
@@ -1037,10 +1037,10 @@ struct ib_qp *hfi1_create_qp(struct ib_pd *ibpd,
struct ib_qp_init_attr *init_attr,
struct ib_udata *udata)
{
- struct hfi1_qp *qp;
+ struct rvt_qp *qp;
struct hfi1_qp_priv *priv;
int err;
- struct hfi1_swqe *swq = NULL;
+ struct rvt_swqe *swq = NULL;
struct hfi1_ibdev *dev;
struct hfi1_devdata *dd;
size_t sz;
@@ -1081,9 +1081,9 @@ struct ib_qp *hfi1_create_qp(struct ib_pd *ibpd,
case IB_QPT_UC:
case IB_QPT_RC:
case IB_QPT_UD:
- sz = sizeof(struct hfi1_sge) *
+ sz = sizeof(struct rvt_sge) *
init_attr->cap.max_send_sge +
- sizeof(struct hfi1_swqe);
+ sizeof(struct rvt_swqe);
swq = vmalloc((init_attr->cap.max_send_wr + 1) * sz);
if (swq == NULL) {
ret = ERR_PTR(-ENOMEM);
@@ -1127,8 +1127,8 @@ struct ib_qp *hfi1_create_qp(struct ib_pd *ibpd,
qp->r_rq.size = init_attr->cap.max_recv_wr + 1;
qp->r_rq.max_sge = init_attr->cap.max_recv_sge;
sz = (sizeof(struct ib_sge) * qp->r_rq.max_sge) +
- sizeof(struct hfi1_rwqe);
- qp->r_rq.wq = vmalloc_user(sizeof(struct hfi1_rwq) +
+ sizeof(struct rvt_rwqe);
+ qp->r_rq.wq = vmalloc_user(sizeof(struct rvt_rwq) +
qp->r_rq.size * sz);
if (!qp->r_rq.wq) {
ret = ERR_PTR(-ENOMEM);
@@ -1192,7 +1192,7 @@ struct ib_qp *hfi1_create_qp(struct ib_pd *ibpd,
goto bail_ip;
}
} else {
- u32 s = sizeof(struct hfi1_rwq) + qp->r_rq.size * sz;
+ u32 s = sizeof(struct rvt_rwq) + qp->r_rq.size * sz;
qp->ip = hfi1_create_mmap_info(dev, s,
ibpd->uobject->context,
@@ -1281,7 +1281,7 @@ bail:
*/
int hfi1_destroy_qp(struct ib_qp *ibqp)
{
- struct hfi1_qp *qp = to_iqp(ibqp);
+ struct rvt_qp *qp = to_iqp(ibqp);
struct hfi1_ibdev *dev = to_idev(ibqp->device);
struct hfi1_qp_priv *priv = qp->priv;
@@ -1387,7 +1387,7 @@ static void free_qpn_table(struct hfi1_qpn_table *qpt)
*
* The QP s_lock should be held.
*/
-void hfi1_get_credit(struct hfi1_qp *qp, u32 aeth)
+void hfi1_get_credit(struct rvt_qp *qp, u32 aeth)
{
u32 credit = (aeth >> HFI1_AETH_CREDIT_SHIFT) & HFI1_AETH_CREDIT_MASK;
@@ -1417,7 +1417,7 @@ void hfi1_get_credit(struct hfi1_qp *qp, u32 aeth)
}
}
-void hfi1_qp_wakeup(struct hfi1_qp *qp, u32 flag)
+void hfi1_qp_wakeup(struct rvt_qp *qp, u32 flag)
{
unsigned long flags;
@@ -1440,7 +1440,7 @@ static int iowait_sleep(
unsigned seq)
{
struct verbs_txreq *tx = container_of(stx, struct verbs_txreq, txreq);
- struct hfi1_qp *qp;
+ struct rvt_qp *qp;
struct hfi1_qp_priv *priv;
unsigned long flags;
int ret = 0;
@@ -1491,7 +1491,7 @@ eagain:
static void iowait_wakeup(struct iowait *wait, int reason)
{
- struct hfi1_qp *qp = iowait_to_qp(wait);
+ struct rvt_qp *qp = iowait_to_qp(wait);
WARN_ON(reason != SDMA_AVAIL_REASON);
hfi1_qp_wakeup(qp, HFI1_S_WAIT_DMA_DESC);
@@ -1558,7 +1558,7 @@ void hfi1_qp_exit(struct hfi1_ibdev *dev)
* Return:
* A send engine for the qp or NULL for SMI type qp.
*/
-struct sdma_engine *qp_to_sdma_engine(struct hfi1_qp *qp, u8 sc5)
+struct sdma_engine *qp_to_sdma_engine(struct rvt_qp *qp, u8 sc5)
{
struct hfi1_devdata *dd = dd_from_ibdev(qp->ibqp.device);
struct sdma_engine *sde;
@@ -1577,7 +1577,7 @@ struct sdma_engine *qp_to_sdma_engine(struct hfi1_qp *qp, u8 sc5)
struct qp_iter {
struct hfi1_ibdev *dev;
- struct hfi1_qp *qp;
+ struct rvt_qp *qp;
int specials;
int n;
};
@@ -1605,8 +1605,8 @@ int qp_iter_next(struct qp_iter *iter)
struct hfi1_ibdev *dev = iter->dev;
int n = iter->n;
int ret = 1;
- struct hfi1_qp *pqp = iter->qp;
- struct hfi1_qp *qp;
+ struct rvt_qp *pqp = iter->qp;
+ struct rvt_qp *qp;
/*
* The approach is to consider the special qps
@@ -1659,7 +1659,7 @@ static const char * const qp_type_str[] = {
"SMI", "GSI", "RC", "UC", "UD",
};
-static int qp_idle(struct hfi1_qp *qp)
+static int qp_idle(struct rvt_qp *qp)
{
return
qp->s_last == qp->s_acked &&
@@ -1670,8 +1670,8 @@ static int qp_idle(struct hfi1_qp *qp)
void qp_iter_print(struct seq_file *s, struct qp_iter *iter)
{
- struct hfi1_swqe *wqe;
- struct hfi1_qp *qp = iter->qp;
+ struct rvt_swqe *wqe;
+ struct rvt_qp *qp = iter->qp;
struct hfi1_qp_priv *priv = qp->priv;
struct sdma_engine *sde;
@@ -1709,7 +1709,7 @@ void qp_iter_print(struct seq_file *s, struct qp_iter *iter)
sde ? sde->this_idx : 0);
}
-void qp_comm_est(struct hfi1_qp *qp)
+void qp_comm_est(struct rvt_qp *qp)
{
qp->r_flags |= HFI1_R_COMM_EST;
if (qp->ibqp.event_handler) {
@@ -1726,7 +1726,7 @@ void qp_comm_est(struct hfi1_qp *qp)
* Switch to alternate path.
* The QP s_lock should be held and interrupts disabled.
*/
-void hfi1_migrate_qp(struct hfi1_qp *qp)
+void hfi1_migrate_qp(struct rvt_qp *qp)
{
struct ib_event ev;
struct hfi1_qp_priv *priv = qp->priv;
diff --git a/drivers/staging/rdma/hfi1/qp.h b/drivers/staging/rdma/hfi1/qp.h
index 1d34c22..fff61f9 100644
--- a/drivers/staging/rdma/hfi1/qp.h
+++ b/drivers/staging/rdma/hfi1/qp.h
@@ -80,7 +80,7 @@ struct hfi1_qpn_table {
struct hfi1_qp_ibdev {
u32 qp_table_size;
u32 qp_table_bits;
- struct hfi1_qp __rcu **qp_table;
+ struct rvt_qp __rcu **qp_table;
spinlock_t qpt_lock;
struct hfi1_qpn_table qpn_table;
};
@@ -98,10 +98,10 @@ static inline u32 qpn_hash(struct hfi1_qp_ibdev *dev, u32 qpn)
* The caller must hold the rcu_read_lock(), and keep the lock until
* the returned qp is no longer in use.
*/
-static inline struct hfi1_qp *hfi1_lookup_qpn(struct hfi1_ibport *ibp,
- u32 qpn) __must_hold(RCU)
+static inline struct rvt_qp *hfi1_lookup_qpn(struct hfi1_ibport *ibp,
+ u32 qpn) __must_hold(RCU)
{
- struct hfi1_qp *qp = NULL;
+ struct rvt_qp *qp = NULL;
if (unlikely(qpn <= 1)) {
qp = rcu_dereference(ibp->qp[qpn]);
@@ -118,21 +118,6 @@ static inline struct hfi1_qp *hfi1_lookup_qpn(struct hfi1_ibport *ibp,
}
/**
- * clear_ahg - reset ahg status in qp
- * @qp - qp pointer
- */
-static inline void clear_ahg(struct hfi1_qp *qp)
-{
- struct hfi1_qp_priv *priv = qp->priv;
-
- priv->s_hdr->ahgcount = 0;
- qp->s_flags &= ~(HFI1_S_AHG_VALID | HFI1_S_AHG_CLEAR);
- if (qp->s_sde)
- sdma_ahg_free(qp->s_sde, qp->s_ahgidx);
- qp->s_ahgidx = -1;
-}
-
-/**
* hfi1_error_qp - put a QP into the error state
* @qp: the QP to put into the error state
* @err: the receive completion error to signal if a RWQE is active
@@ -142,7 +127,7 @@ static inline void clear_ahg(struct hfi1_qp *qp)
* The QP r_lock and s_lock should be held and interrupts disabled.
* If we are already in error state, just return.
*/
-int hfi1_error_qp(struct hfi1_qp *qp, enum ib_wc_status err);
+int hfi1_error_qp(struct rvt_qp *qp, enum ib_wc_status err);
/**
* hfi1_modify_qp - modify the attributes of a queue pair
@@ -165,7 +150,7 @@ int hfi1_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
*
* Returns the AETH.
*/
-__be32 hfi1_compute_aeth(struct hfi1_qp *qp);
+__be32 hfi1_compute_aeth(struct rvt_qp *qp);
/**
* hfi1_create_qp - create a queue pair for a device
@@ -198,7 +183,7 @@ int hfi1_destroy_qp(struct ib_qp *ibqp);
*
* The QP s_lock should be held.
*/
-void hfi1_get_credit(struct hfi1_qp *qp, u32 aeth);
+void hfi1_get_credit(struct rvt_qp *qp, u32 aeth);
/**
* hfi1_qp_init - allocate QP tables
@@ -217,9 +202,9 @@ void hfi1_qp_exit(struct hfi1_ibdev *dev);
* @qp: the QP
* @flag: flag the qp on which the qp is stalled
*/
-void hfi1_qp_wakeup(struct hfi1_qp *qp, u32 flag);
+void hfi1_qp_wakeup(struct rvt_qp *qp, u32 flag);
-struct sdma_engine *qp_to_sdma_engine(struct hfi1_qp *qp, u8 sc5);
+struct sdma_engine *qp_to_sdma_engine(struct rvt_qp *qp, u8 sc5);
struct qp_iter;
@@ -246,7 +231,7 @@ void qp_iter_print(struct seq_file *s, struct qp_iter *iter);
* qp_comm_est - handle trap with QP established
* @qp: the QP
*/
-void qp_comm_est(struct hfi1_qp *qp);
+void qp_comm_est(struct rvt_qp *qp);
/**
* _hfi1_schedule_send - schedule progress
@@ -257,7 +242,7 @@ void qp_comm_est(struct hfi1_qp *qp);
* It is only used in the post send, which doesn't hold
* the s_lock.
*/
-static inline void _hfi1_schedule_send(struct hfi1_qp *qp)
+static inline void _hfi1_schedule_send(struct rvt_qp *qp)
{
struct hfi1_qp_priv *priv = qp->priv;
struct hfi1_ibport *ibp =
@@ -278,12 +263,12 @@ static inline void _hfi1_schedule_send(struct hfi1_qp *qp)
* This schedules qp progress and caller should hold
* the s_lock.
*/
-static inline void hfi1_schedule_send(struct hfi1_qp *qp)
+static inline void hfi1_schedule_send(struct rvt_qp *qp)
{
if (hfi1_send_ok(qp))
_hfi1_schedule_send(qp);
}
-void hfi1_migrate_qp(struct hfi1_qp *qp);
+void hfi1_migrate_qp(struct rvt_qp *qp);
#endif /* _QP_H */
diff --git a/drivers/staging/rdma/hfi1/rc.c b/drivers/staging/rdma/hfi1/rc.c
index d830a8f..752643d 100644
--- a/drivers/staging/rdma/hfi1/rc.c
+++ b/drivers/staging/rdma/hfi1/rc.c
@@ -60,7 +60,7 @@
static void rc_timeout(unsigned long arg);
-static u32 restart_sge(struct hfi1_sge_state *ss, struct hfi1_swqe *wqe,
+static u32 restart_sge(struct rvt_sge_state *ss, struct rvt_swqe *wqe,
u32 psn, u32 pmtu)
{
u32 len;
@@ -74,7 +74,7 @@ static u32 restart_sge(struct hfi1_sge_state *ss, struct hfi1_swqe *wqe,
return wqe->length - len;
}
-static void start_timer(struct hfi1_qp *qp)
+static void start_timer(struct rvt_qp *qp)
{
qp->s_flags |= HFI1_S_TIMER;
qp->s_timer.function = rc_timeout;
@@ -94,10 +94,10 @@ static void start_timer(struct hfi1_qp *qp)
* Note that we are in the responder's side of the QP context.
* Note the QP s_lock must be held.
*/
-static int make_rc_ack(struct hfi1_ibdev *dev, struct hfi1_qp *qp,
+static int make_rc_ack(struct hfi1_ibdev *dev, struct rvt_qp *qp,
struct hfi1_other_headers *ohdr, u32 pmtu)
{
- struct hfi1_ack_entry *e;
+ struct rvt_ack_entry *e;
u32 hwords;
u32 len;
u32 bth0;
@@ -116,7 +116,7 @@ static int make_rc_ack(struct hfi1_ibdev *dev, struct hfi1_qp *qp,
case OP(RDMA_READ_RESPONSE_ONLY):
e = &qp->s_ack_queue[qp->s_tail_ack_queue];
if (e->rdma_sge.mr) {
- hfi1_put_mr(e->rdma_sge.mr);
+ rvt_put_mr(e->rdma_sge.mr);
e->rdma_sge.mr = NULL;
}
/* FALLTHROUGH */
@@ -154,7 +154,7 @@ static int make_rc_ack(struct hfi1_ibdev *dev, struct hfi1_qp *qp,
/* Copy SGE state in case we need to resend */
qp->s_rdma_mr = e->rdma_sge.mr;
if (qp->s_rdma_mr)
- hfi1_get_mr(qp->s_rdma_mr);
+ rvt_get_mr(qp->s_rdma_mr);
qp->s_ack_rdma_sge.sge = e->rdma_sge;
qp->s_ack_rdma_sge.num_sge = 1;
qp->s_cur_sge = &qp->s_ack_rdma_sge;
@@ -193,7 +193,7 @@ static int make_rc_ack(struct hfi1_ibdev *dev, struct hfi1_qp *qp,
qp->s_cur_sge = &qp->s_ack_rdma_sge;
qp->s_rdma_mr = qp->s_ack_rdma_sge.sge.mr;
if (qp->s_rdma_mr)
- hfi1_get_mr(qp->s_rdma_mr);
+ rvt_get_mr(qp->s_rdma_mr);
len = qp->s_ack_rdma_sge.sge.sge_length;
if (len > pmtu) {
len = pmtu;
@@ -257,13 +257,13 @@ bail:
*
* Return 1 if constructed; otherwise, return 0.
*/
-int hfi1_make_rc_req(struct hfi1_qp *qp)
+int hfi1_make_rc_req(struct rvt_qp *qp)
{
struct hfi1_qp_priv *priv = qp->priv;
struct hfi1_ibdev *dev = to_idev(qp->ibqp.device);
struct hfi1_other_headers *ohdr;
- struct hfi1_sge_state *ss;
- struct hfi1_swqe *wqe;
+ struct rvt_sge_state *ss;
+ struct rvt_swqe *wqe;
/* header size in 32-bit words LRH+BTH = (8+12)/4. */
u32 hwords = 5;
u32 len;
@@ -683,7 +683,7 @@ unlock:
* Note that RDMA reads and atomics are handled in the
* send side QP state and tasklet.
*/
-void hfi1_send_rc_ack(struct hfi1_ctxtdata *rcd, struct hfi1_qp *qp,
+void hfi1_send_rc_ack(struct hfi1_ctxtdata *rcd, struct rvt_qp *qp,
int is_fecn)
{
struct hfi1_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
@@ -794,10 +794,10 @@ queue_ack:
* for the given QP.
* Called at interrupt level with the QP s_lock held.
*/
-static void reset_psn(struct hfi1_qp *qp, u32 psn)
+static void reset_psn(struct rvt_qp *qp, u32 psn)
{
u32 n = qp->s_acked;
- struct hfi1_swqe *wqe = get_swqe_ptr(qp, n);
+ struct rvt_swqe *wqe = get_swqe_ptr(qp, n);
u32 opcode;
qp->s_cur = n;
@@ -880,9 +880,9 @@ done:
* Back up requester to resend the last un-ACKed request.
* The QP r_lock and s_lock should be held and interrupts disabled.
*/
-static void restart_rc(struct hfi1_qp *qp, u32 psn, int wait)
+static void restart_rc(struct rvt_qp *qp, u32 psn, int wait)
{
- struct hfi1_swqe *wqe = get_swqe_ptr(qp, qp->s_acked);
+ struct rvt_swqe *wqe = get_swqe_ptr(qp, qp->s_acked);
struct hfi1_ibport *ibp;
if (qp->s_retry == 0) {
@@ -917,7 +917,7 @@ static void restart_rc(struct hfi1_qp *qp, u32 psn, int wait)
*/
static void rc_timeout(unsigned long arg)
{
- struct hfi1_qp *qp = (struct hfi1_qp *)arg;
+ struct rvt_qp *qp = (struct rvt_qp *)arg;
struct hfi1_ibport *ibp;
unsigned long flags;
@@ -941,7 +941,7 @@ static void rc_timeout(unsigned long arg)
*/
void hfi1_rc_rnr_retry(unsigned long arg)
{
- struct hfi1_qp *qp = (struct hfi1_qp *)arg;
+ struct rvt_qp *qp = (struct rvt_qp *)arg;
unsigned long flags;
spin_lock_irqsave(&qp->s_lock, flags);
@@ -957,9 +957,9 @@ void hfi1_rc_rnr_retry(unsigned long arg)
* Set qp->s_sending_psn to the next PSN after the given one.
* This would be psn+1 except when RDMA reads are present.
*/
-static void reset_sending_psn(struct hfi1_qp *qp, u32 psn)
+static void reset_sending_psn(struct rvt_qp *qp, u32 psn)
{
- struct hfi1_swqe *wqe;
+ struct rvt_swqe *wqe;
u32 n = qp->s_last;
/* Find the work request corresponding to the given PSN. */
@@ -982,10 +982,10 @@ static void reset_sending_psn(struct hfi1_qp *qp, u32 psn)
/*
* This should be called with the QP s_lock held and interrupts disabled.
*/
-void hfi1_rc_send_complete(struct hfi1_qp *qp, struct hfi1_ib_header *hdr)
+void hfi1_rc_send_complete(struct rvt_qp *qp, struct hfi1_ib_header *hdr)
{
struct hfi1_other_headers *ohdr;
- struct hfi1_swqe *wqe;
+ struct rvt_swqe *wqe;
struct ib_wc wc;
unsigned i;
u32 opcode;
@@ -1027,9 +1027,9 @@ void hfi1_rc_send_complete(struct hfi1_qp *qp, struct hfi1_ib_header *hdr)
cmp_psn(qp->s_sending_psn, qp->s_sending_hpsn) <= 0)
break;
for (i = 0; i < wqe->wr.num_sge; i++) {
- struct hfi1_sge *sge = &wqe->sg_list[i];
+ struct rvt_sge *sge = &wqe->sg_list[i];
- hfi1_put_mr(sge->mr);
+ rvt_put_mr(sge->mr);
}
/* Post a send completion queue entry if requested. */
if (!(qp->s_flags & HFI1_S_SIGNAL_REQ_WR) ||
@@ -1059,7 +1059,7 @@ void hfi1_rc_send_complete(struct hfi1_qp *qp, struct hfi1_ib_header *hdr)
}
}
-static inline void update_last_psn(struct hfi1_qp *qp, u32 psn)
+static inline void update_last_psn(struct rvt_qp *qp, u32 psn)
{
qp->s_last_psn = psn;
}
@@ -1069,9 +1069,9 @@ static inline void update_last_psn(struct hfi1_qp *qp, u32 psn)
* This is similar to hfi1_send_complete but has to check to be sure
* that the SGEs are not being referenced if the SWQE is being resent.
*/
-static struct hfi1_swqe *do_rc_completion(struct hfi1_qp *qp,
- struct hfi1_swqe *wqe,
- struct hfi1_ibport *ibp)
+static struct rvt_swqe *do_rc_completion(struct rvt_qp *qp,
+ struct rvt_swqe *wqe,
+ struct hfi1_ibport *ibp)
{
struct ib_wc wc;
unsigned i;
@@ -1084,9 +1084,9 @@ static struct hfi1_swqe *do_rc_completion(struct hfi1_qp *qp,
if (cmp_psn(wqe->lpsn, qp->s_sending_psn) < 0 ||
cmp_psn(qp->s_sending_psn, qp->s_sending_hpsn) > 0) {
for (i = 0; i < wqe->wr.num_sge; i++) {
- struct hfi1_sge *sge = &wqe->sg_list[i];
+ struct rvt_sge *sge = &wqe->sg_list[i];
- hfi1_put_mr(sge->mr);
+ rvt_put_mr(sge->mr);
}
/* Post a send completion queue entry if requested. */
if (!(qp->s_flags & HFI1_S_SIGNAL_REQ_WR) ||
@@ -1158,12 +1158,12 @@ static struct hfi1_swqe *do_rc_completion(struct hfi1_qp *qp,
* May be called at interrupt level, with the QP s_lock held.
* Returns 1 if OK, 0 if current operation should be aborted (NAK).
*/
-static int do_rc_ack(struct hfi1_qp *qp, u32 aeth, u32 psn, int opcode,
+static int do_rc_ack(struct rvt_qp *qp, u32 aeth, u32 psn, int opcode,
u64 val, struct hfi1_ctxtdata *rcd)
{
struct hfi1_ibport *ibp;
enum ib_wc_status status;
- struct hfi1_swqe *wqe;
+ struct rvt_swqe *wqe;
int ret = 0;
u32 ack_psn;
int diff;
@@ -1381,10 +1381,10 @@ bail:
* We have seen an out of sequence RDMA read middle or last packet.
* This ACKs SENDs and RDMA writes up to the first RDMA read or atomic SWQE.
*/
-static void rdma_seq_err(struct hfi1_qp *qp, struct hfi1_ibport *ibp, u32 psn,
+static void rdma_seq_err(struct rvt_qp *qp, struct hfi1_ibport *ibp, u32 psn,
struct hfi1_ctxtdata *rcd)
{
- struct hfi1_swqe *wqe;
+ struct rvt_swqe *wqe;
/* Remove QP from retry timer */
if (qp->s_flags & (HFI1_S_TIMER | HFI1_S_WAIT_RNR)) {
@@ -1430,11 +1430,11 @@ static void rdma_seq_err(struct hfi1_qp *qp, struct hfi1_ibport *ibp, u32 psn,
*/
static void rc_rcv_resp(struct hfi1_ibport *ibp,
struct hfi1_other_headers *ohdr,
- void *data, u32 tlen, struct hfi1_qp *qp,
+ void *data, u32 tlen, struct rvt_qp *qp,
u32 opcode, u32 psn, u32 hdrsize, u32 pmtu,
struct hfi1_ctxtdata *rcd)
{
- struct hfi1_swqe *wqe;
+ struct rvt_swqe *wqe;
enum ib_wc_status status;
unsigned long flags;
int diff;
@@ -1610,7 +1610,7 @@ bail:
}
static inline void rc_defered_ack(struct hfi1_ctxtdata *rcd,
- struct hfi1_qp *qp)
+ struct rvt_qp *qp)
{
if (list_empty(&qp->rspwait)) {
qp->r_flags |= HFI1_R_RSP_DEFERED_ACK;
@@ -1619,7 +1619,7 @@ static inline void rc_defered_ack(struct hfi1_ctxtdata *rcd,
}
}
-static inline void rc_cancel_ack(struct hfi1_qp *qp)
+static inline void rc_cancel_ack(struct rvt_qp *qp)
{
qp->r_adefered = 0;
if (list_empty(&qp->rspwait))
@@ -1646,11 +1646,11 @@ static inline void rc_cancel_ack(struct hfi1_qp *qp)
* schedule a response to be sent.
*/
static noinline int rc_rcv_error(struct hfi1_other_headers *ohdr, void *data,
- struct hfi1_qp *qp, u32 opcode, u32 psn, int diff,
+ struct rvt_qp *qp, u32 opcode, u32 psn, int diff,
struct hfi1_ctxtdata *rcd)
{
struct hfi1_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
- struct hfi1_ack_entry *e;
+ struct rvt_ack_entry *e;
unsigned long flags;
u8 i, prev;
int old_req;
@@ -1748,7 +1748,7 @@ static noinline int rc_rcv_error(struct hfi1_other_headers *ohdr, void *data,
if (unlikely(offset + len != e->rdma_sge.sge_length))
goto unlock_done;
if (e->rdma_sge.mr) {
- hfi1_put_mr(e->rdma_sge.mr);
+ rvt_put_mr(e->rdma_sge.mr);
e->rdma_sge.mr = NULL;
}
if (len != 0) {
@@ -1756,8 +1756,8 @@ static noinline int rc_rcv_error(struct hfi1_other_headers *ohdr, void *data,
u64 vaddr = be64_to_cpu(reth->vaddr);
int ok;
- ok = hfi1_rkey_ok(qp, &e->rdma_sge, len, vaddr, rkey,
- IB_ACCESS_REMOTE_READ);
+ ok = rvt_rkey_ok(qp, &e->rdma_sge, len, vaddr, rkey,
+ IB_ACCESS_REMOTE_READ);
if (unlikely(!ok))
goto unlock_done;
} else {
@@ -1824,7 +1824,7 @@ send_ack:
return 0;
}
-void hfi1_rc_error(struct hfi1_qp *qp, enum ib_wc_status err)
+void hfi1_rc_error(struct rvt_qp *qp, enum ib_wc_status err)
{
unsigned long flags;
int lastwqe;
@@ -1843,7 +1843,7 @@ void hfi1_rc_error(struct hfi1_qp *qp, enum ib_wc_status err)
}
}
-static inline void update_ack_queue(struct hfi1_qp *qp, unsigned n)
+static inline void update_ack_queue(struct rvt_qp *qp, unsigned n)
{
unsigned next;
@@ -1958,7 +1958,7 @@ void hfi1_rc_rcv(struct hfi1_packet *packet)
u32 rcv_flags = packet->rcv_flags;
void *data = packet->ebuf;
u32 tlen = packet->tlen;
- struct hfi1_qp *qp = packet->qp;
+ struct rvt_qp *qp = packet->qp;
struct hfi1_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
struct hfi1_other_headers *ohdr = packet->ohdr;
@@ -2175,8 +2175,8 @@ send_last:
int ok;
/* Check rkey & NAK */
- ok = hfi1_rkey_ok(qp, &qp->r_sge.sge, qp->r_len, vaddr,
- rkey, IB_ACCESS_REMOTE_WRITE);
+ ok = rvt_rkey_ok(qp, &qp->r_sge.sge, qp->r_len, vaddr,
+ rkey, IB_ACCESS_REMOTE_WRITE);
if (unlikely(!ok))
goto nack_acc;
qp->r_sge.num_sge = 1;
@@ -2201,7 +2201,7 @@ send_last:
goto send_last;
case OP(RDMA_READ_REQUEST): {
- struct hfi1_ack_entry *e;
+ struct rvt_ack_entry *e;
u32 len;
u8 next;
@@ -2219,7 +2219,7 @@ send_last:
}
e = &qp->s_ack_queue[qp->r_head_ack_queue];
if (e->opcode == OP(RDMA_READ_REQUEST) && e->rdma_sge.mr) {
- hfi1_put_mr(e->rdma_sge.mr);
+ rvt_put_mr(e->rdma_sge.mr);
e->rdma_sge.mr = NULL;
}
reth = &ohdr->u.rc.reth;
@@ -2230,8 +2230,8 @@ send_last:
int ok;
/* Check rkey & NAK */
- ok = hfi1_rkey_ok(qp, &e->rdma_sge, len, vaddr,
- rkey, IB_ACCESS_REMOTE_READ);
+ ok = rvt_rkey_ok(qp, &e->rdma_sge, len, vaddr,
+ rkey, IB_ACCESS_REMOTE_READ);
if (unlikely(!ok))
goto nack_acc_unlck;
/*
@@ -2274,7 +2274,7 @@ send_last:
case OP(COMPARE_SWAP):
case OP(FETCH_ADD): {
struct ib_atomic_eth *ateth;
- struct hfi1_ack_entry *e;
+ struct rvt_ack_entry *e;
u64 vaddr;
atomic64_t *maddr;
u64 sdata;
@@ -2294,7 +2294,7 @@ send_last:
}
e = &qp->s_ack_queue[qp->r_head_ack_queue];
if (e->opcode == OP(RDMA_READ_REQUEST) && e->rdma_sge.mr) {
- hfi1_put_mr(e->rdma_sge.mr);
+ rvt_put_mr(e->rdma_sge.mr);
e->rdma_sge.mr = NULL;
}
ateth = &ohdr->u.atomic_eth;
@@ -2304,9 +2304,9 @@ send_last:
goto nack_inv_unlck;
rkey = be32_to_cpu(ateth->rkey);
/* Check rkey & NAK */
- if (unlikely(!hfi1_rkey_ok(qp, &qp->r_sge.sge, sizeof(u64),
- vaddr, rkey,
- IB_ACCESS_REMOTE_ATOMIC)))
+ if (unlikely(!rvt_rkey_ok(qp, &qp->r_sge.sge, sizeof(u64),
+ vaddr, rkey,
+ IB_ACCESS_REMOTE_ATOMIC)))
goto nack_acc_unlck;
/* Perform atomic OP and save result. */
maddr = (atomic64_t *) qp->r_sge.sge.vaddr;
@@ -2316,7 +2316,7 @@ send_last:
(u64) cmpxchg((u64 *) qp->r_sge.sge.vaddr,
be64_to_cpu(ateth->compare_data),
sdata);
- hfi1_put_mr(qp->r_sge.sge.mr);
+ rvt_put_mr(qp->r_sge.sge.mr);
qp->r_sge.num_sge = 0;
e->opcode = opcode;
e->sent = 0;
@@ -2404,7 +2404,7 @@ void hfi1_rc_hdrerr(
struct hfi1_ctxtdata *rcd,
struct hfi1_ib_header *hdr,
u32 rcv_flags,
- struct hfi1_qp *qp)
+ struct rvt_qp *qp)
{
int has_grh = rcv_flags & HFI1_HAS_GRH;
struct hfi1_other_headers *ohdr;
diff --git a/drivers/staging/rdma/hfi1/ruc.c b/drivers/staging/rdma/hfi1/ruc.c
index 4108c6a..558dadb 100644
--- a/drivers/staging/rdma/hfi1/ruc.c
+++ b/drivers/staging/rdma/hfi1/ruc.c
@@ -97,15 +97,15 @@ const u32 ib_hfi1_rnr_table[32] = {
* Validate a RWQE and fill in the SGE state.
* Return 1 if OK.
*/
-static int init_sge(struct hfi1_qp *qp, struct hfi1_rwqe *wqe)
+static int init_sge(struct rvt_qp *qp, struct rvt_rwqe *wqe)
{
int i, j, ret;
struct ib_wc wc;
struct rvt_lkey_table *rkt;
struct rvt_pd *pd;
- struct hfi1_sge_state *ss;
+ struct rvt_sge_state *ss;
- rkt = &to_idev(qp->ibqp.device)->lk_table;
+ rkt = &to_idev(qp->ibqp.device)->rdi.lk_table;
pd = ibpd_to_rvtpd(qp->ibqp.srq ? qp->ibqp.srq->pd : qp->ibqp.pd);
ss = &qp->r_sge;
ss->sg_list = qp->r_sg_list;
@@ -114,8 +114,8 @@ static int init_sge(struct hfi1_qp *qp, struct hfi1_rwqe *wqe)
if (wqe->sg_list[i].length == 0)
continue;
/* Check LKEY */
- if (!hfi1_lkey_ok(rkt, pd, j ? &ss->sg_list[j - 1] : &ss->sge,
- &wqe->sg_list[i], IB_ACCESS_LOCAL_WRITE))
+ if (!rvt_lkey_ok(rkt, pd, j ? &ss->sg_list[j - 1] : &ss->sge,
+ &wqe->sg_list[i], IB_ACCESS_LOCAL_WRITE))
goto bad_lkey;
qp->r_len += wqe->sg_list[i].length;
j++;
@@ -127,9 +127,9 @@ static int init_sge(struct hfi1_qp *qp, struct hfi1_rwqe *wqe)
bad_lkey:
while (j) {
- struct hfi1_sge *sge = --j ? &ss->sg_list[j - 1] : &ss->sge;
+ struct rvt_sge *sge = --j ? &ss->sg_list[j - 1] : &ss->sge;
- hfi1_put_mr(sge->mr);
+ rvt_put_mr(sge->mr);
}
ss->num_sge = 0;
memset(&wc, 0, sizeof(wc));
@@ -154,13 +154,13 @@ bail:
*
* Can be called from interrupt level.
*/
-int hfi1_get_rwqe(struct hfi1_qp *qp, int wr_id_only)
+int hfi1_get_rwqe(struct rvt_qp *qp, int wr_id_only)
{
unsigned long flags;
- struct hfi1_rq *rq;
- struct hfi1_rwq *wq;
+ struct rvt_rq *rq;
+ struct rvt_rwq *wq;
struct hfi1_srq *srq;
- struct hfi1_rwqe *wqe;
+ struct rvt_rwqe *wqe;
void (*handler)(struct ib_event *, void *);
u32 tail;
int ret;
@@ -265,7 +265,7 @@ static int gid_ok(union ib_gid *gid, __be64 gid_prefix, __be64 id)
* The s_lock will be acquired around the hfi1_migrate_qp() call.
*/
int hfi1_ruc_check_hdr(struct hfi1_ibport *ibp, struct hfi1_ib_header *hdr,
- int has_grh, struct hfi1_qp *qp, u32 bth0)
+ int has_grh, struct rvt_qp *qp, u32 bth0)
{
__be64 guid;
unsigned long flags;
@@ -353,12 +353,12 @@ err:
* receive interrupts since this is a connected protocol and all packets
* will pass through here.
*/
-static void ruc_loopback(struct hfi1_qp *sqp)
+static void ruc_loopback(struct rvt_qp *sqp)
{
struct hfi1_ibport *ibp = to_iport(sqp->ibqp.device, sqp->port_num);
- struct hfi1_qp *qp;
- struct hfi1_swqe *wqe;
- struct hfi1_sge *sge;
+ struct rvt_qp *qp;
+ struct rvt_swqe *wqe;
+ struct rvt_sge *sge;
unsigned long flags;
struct ib_wc wc;
u64 sdata;
@@ -459,11 +459,10 @@ again:
if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_WRITE)))
goto inv_err;
if (wqe->length == 0)
- break;
- if (unlikely(!hfi1_rkey_ok(qp, &qp->r_sge.sge, wqe->length,
- wqe->rdma_wr.remote_addr,
- wqe->rdma_wr.rkey,
- IB_ACCESS_REMOTE_WRITE)))
+ if (unlikely(!rvt_rkey_ok(qp, &qp->r_sge.sge, wqe->length,
+ wqe->rdma_wr.remote_addr,
+ wqe->rdma_wr.rkey,
+ IB_ACCESS_REMOTE_WRITE)))
goto acc_err;
qp->r_sge.sg_list = NULL;
qp->r_sge.num_sge = 1;
@@ -473,10 +472,10 @@ again:
case IB_WR_RDMA_READ:
if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_READ)))
goto inv_err;
- if (unlikely(!hfi1_rkey_ok(qp, &sqp->s_sge.sge, wqe->length,
- wqe->rdma_wr.remote_addr,
- wqe->rdma_wr.rkey,
- IB_ACCESS_REMOTE_READ)))
+ if (unlikely(!rvt_rkey_ok(qp, &sqp->s_sge.sge, wqe->length,
+ wqe->rdma_wr.remote_addr,
+ wqe->rdma_wr.rkey,
+ IB_ACCESS_REMOTE_READ)))
goto acc_err;
release = 0;
sqp->s_sge.sg_list = NULL;
@@ -491,10 +490,10 @@ again:
case IB_WR_ATOMIC_FETCH_AND_ADD:
if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_ATOMIC)))
goto inv_err;
- if (unlikely(!hfi1_rkey_ok(qp, &qp->r_sge.sge, sizeof(u64),
- wqe->atomic_wr.remote_addr,
- wqe->atomic_wr.rkey,
- IB_ACCESS_REMOTE_ATOMIC)))
+ if (unlikely(!rvt_rkey_ok(qp, &qp->r_sge.sge, sizeof(u64),
+ wqe->atomic_wr.remote_addr,
+ wqe->atomic_wr.rkey,
+ IB_ACCESS_REMOTE_ATOMIC)))
goto acc_err;
/* Perform atomic OP and save result. */
maddr = (atomic64_t *) qp->r_sge.sge.vaddr;
@@ -504,7 +503,7 @@ again:
(u64) atomic64_add_return(sdata, maddr) - sdata :
(u64) cmpxchg((u64 *) qp->r_sge.sge.vaddr,
sdata, wqe->atomic_wr.swap);
- hfi1_put_mr(qp->r_sge.sge.mr);
+ rvt_put_mr(qp->r_sge.sge.mr);
qp->r_sge.num_sge = 0;
goto send_comp;
@@ -528,7 +527,7 @@ again:
sge->sge_length -= len;
if (sge->sge_length == 0) {
if (!release)
- hfi1_put_mr(sge->mr);
+ rvt_put_mr(sge->mr);
if (--sqp->s_sge.num_sge)
*sge = *sqp->s_sge.sg_list++;
} else if (sge->length == 0 && sge->mr->lkey) {
@@ -675,6 +674,20 @@ u32 hfi1_make_grh(struct hfi1_ibport *ibp, struct ib_grh *hdr,
return sizeof(struct ib_grh) / sizeof(u32);
}
+/*
+ * free_ahg - clear ahg from QP
+ */
+void clear_ahg(struct rvt_qp *qp)
+{
+ struct hfi1_qp_priv *priv = qp->priv;
+
+ priv->s_hdr->ahgcount = 0;
+ qp->s_flags &= ~(HFI1_S_AHG_VALID | HFI1_S_AHG_CLEAR);
+ if (qp->s_sde)
+ sdma_ahg_free(qp->s_sde, qp->s_ahgidx);
+ qp->s_ahgidx = -1;
+}
+
#define BTH2_OFFSET (offsetof(struct hfi1_pio_header, hdr.u.oth.bth[2]) / 4)
/**
@@ -688,7 +701,7 @@ u32 hfi1_make_grh(struct hfi1_ibport *ibp, struct ib_grh *hdr,
* Subsequent middles use the copied entry, editing the
* PSN with 1 or 2 edits.
*/
-static inline void build_ahg(struct hfi1_qp *qp, u32 npsn)
+static inline void build_ahg(struct rvt_qp *qp, u32 npsn)
{
struct hfi1_qp_priv *priv = qp->priv;
if (unlikely(qp->s_flags & HFI1_S_AHG_CLEAR))
@@ -732,7 +745,7 @@ static inline void build_ahg(struct hfi1_qp *qp, u32 npsn)
}
}
-void hfi1_make_ruc_header(struct hfi1_qp *qp, struct hfi1_other_headers *ohdr,
+void hfi1_make_ruc_header(struct rvt_qp *qp, struct hfi1_other_headers *ohdr,
u32 bth0, u32 bth2, int middle)
{
struct hfi1_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
@@ -810,9 +823,9 @@ void hfi1_make_ruc_header(struct hfi1_qp *qp, struct hfi1_other_headers *ohdr,
void hfi1_do_send(struct work_struct *work)
{
struct iowait *wait = container_of(work, struct iowait, iowork);
- struct hfi1_qp *qp = iowait_to_qp(wait);
+ struct rvt_qp *qp = iowait_to_qp(wait);
struct hfi1_pkt_state ps;
- int (*make_req)(struct hfi1_qp *qp);
+ int (*make_req)(struct rvt_qp *qp);
unsigned long flags;
unsigned long timeout;
@@ -874,7 +887,7 @@ void hfi1_do_send(struct work_struct *work)
/*
* This should be called with s_lock held.
*/
-void hfi1_send_complete(struct hfi1_qp *qp, struct hfi1_swqe *wqe,
+void hfi1_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe,
enum ib_wc_status status)
{
u32 old_last, last;
@@ -884,9 +897,9 @@ void hfi1_send_complete(struct hfi1_qp *qp, struct hfi1_swqe *wqe,
return;
for (i = 0; i < wqe->wr.num_sge; i++) {
- struct hfi1_sge *sge = &wqe->sg_list[i];
+ struct rvt_sge *sge = &wqe->sg_list[i];
- hfi1_put_mr(sge->mr);
+ rvt_put_mr(sge->mr);
}
if (qp->ibqp.qp_type == IB_QPT_UD ||
qp->ibqp.qp_type == IB_QPT_SMI ||
diff --git a/drivers/staging/rdma/hfi1/sdma.h b/drivers/staging/rdma/hfi1/sdma.h
index 664b251..e7ebd16 100644
--- a/drivers/staging/rdma/hfi1/sdma.h
+++ b/drivers/staging/rdma/hfi1/sdma.h
@@ -379,10 +379,10 @@ struct sdma_txreq {
struct verbs_txreq {
struct hfi1_pio_header phdr;
struct sdma_txreq txreq;
- struct hfi1_qp *qp;
- struct hfi1_swqe *wqe;
+ struct rvt_qp *qp;
+ struct rvt_swqe *wqe;
struct rvt_mregion *mr;
- struct hfi1_sge_state *ss;
+ struct rvt_sge_state *ss;
struct sdma_engine *sde;
u16 hdr_dwords;
u16 hdr_inx;
diff --git a/drivers/staging/rdma/hfi1/srq.c b/drivers/staging/rdma/hfi1/srq.c
index 67786d4..932bd96 100644
--- a/drivers/staging/rdma/hfi1/srq.c
+++ b/drivers/staging/rdma/hfi1/srq.c
@@ -66,12 +66,12 @@ int hfi1_post_srq_receive(struct ib_srq *ibsrq, struct ib_recv_wr *wr,
struct ib_recv_wr **bad_wr)
{
struct hfi1_srq *srq = to_isrq(ibsrq);
- struct hfi1_rwq *wq;
+ struct rvt_rwq *wq;
unsigned long flags;
int ret;
for (; wr; wr = wr->next) {
- struct hfi1_rwqe *wqe;
+ struct rvt_rwqe *wqe;
u32 next;
int i;
@@ -149,8 +149,8 @@ struct ib_srq *hfi1_create_srq(struct ib_pd *ibpd,
srq->rq.size = srq_init_attr->attr.max_wr + 1;
srq->rq.max_sge = srq_init_attr->attr.max_sge;
sz = sizeof(struct ib_sge) * srq->rq.max_sge +
- sizeof(struct hfi1_rwqe);
- srq->rq.wq = vmalloc_user(sizeof(struct hfi1_rwq) + srq->rq.size * sz);
+ sizeof(struct rvt_rwqe);
+ srq->rq.wq = vmalloc_user(sizeof(struct rvt_rwq) + srq->rq.size * sz);
if (!srq->rq.wq) {
ret = ERR_PTR(-ENOMEM);
goto bail_srq;
@@ -162,7 +162,7 @@ struct ib_srq *hfi1_create_srq(struct ib_pd *ibpd,
*/
if (udata && udata->outlen >= sizeof(__u64)) {
int err;
- u32 s = sizeof(struct hfi1_rwq) + srq->rq.size * sz;
+ u32 s = sizeof(struct rvt_rwq) + srq->rq.size * sz;
srq->ip =
hfi1_create_mmap_info(dev, s, ibpd->uobject->context,
@@ -230,12 +230,12 @@ int hfi1_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
struct ib_udata *udata)
{
struct hfi1_srq *srq = to_isrq(ibsrq);
- struct hfi1_rwq *wq;
+ struct rvt_rwq *wq;
int ret = 0;
if (attr_mask & IB_SRQ_MAX_WR) {
- struct hfi1_rwq *owq;
- struct hfi1_rwqe *p;
+ struct rvt_rwq *owq;
+ struct rvt_rwqe *p;
u32 sz, size, n, head, tail;
/* Check that the requested sizes are below the limits. */
@@ -246,10 +246,10 @@ int hfi1_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
goto bail;
}
- sz = sizeof(struct hfi1_rwqe) +
+ sz = sizeof(struct rvt_rwqe) +
srq->rq.max_sge * sizeof(struct ib_sge);
size = attr->max_wr + 1;
- wq = vmalloc_user(sizeof(struct hfi1_rwq) + size * sz);
+ wq = vmalloc_user(sizeof(struct rvt_rwq) + size * sz);
if (!wq) {
ret = -ENOMEM;
goto bail;
@@ -296,7 +296,7 @@ int hfi1_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
n = 0;
p = wq->wq;
while (tail != head) {
- struct hfi1_rwqe *wqe;
+ struct rvt_rwqe *wqe;
int i;
wqe = get_rwqe_ptr(&srq->rq, tail);
@@ -305,7 +305,7 @@ int hfi1_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
for (i = 0; i < wqe->num_sge; i++)
p->sg_list[i] = wqe->sg_list[i];
n++;
- p = (struct hfi1_rwqe *)((char *)p + sz);
+ p = (struct rvt_rwqe *)((char *)p + sz);
if (++tail >= srq->rq.size)
tail = 0;
}
@@ -320,9 +320,9 @@ int hfi1_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
vfree(owq);
if (srq->ip) {
- struct hfi1_mmap_info *ip = srq->ip;
+ struct rvt_mmap_info *ip = srq->ip;
struct hfi1_ibdev *dev = to_idev(srq->ibsrq.device);
- u32 s = sizeof(struct hfi1_rwq) + size * sz;
+ u32 s = sizeof(struct rvt_rwq) + size * sz;
hfi1_update_mmap_info(dev, ip, s, wq);
diff --git a/drivers/staging/rdma/hfi1/trace.h b/drivers/staging/rdma/hfi1/trace.h
index 5743029..570ad4c 100644
--- a/drivers/staging/rdma/hfi1/trace.h
+++ b/drivers/staging/rdma/hfi1/trace.h
@@ -294,7 +294,7 @@ TRACE_EVENT(hfi1_wantpiointr,
);
DECLARE_EVENT_CLASS(hfi1_qpsleepwakeup_template,
- TP_PROTO(struct hfi1_qp *qp, u32 flags),
+ TP_PROTO(struct rvt_qp *qp, u32 flags),
TP_ARGS(qp, flags),
TP_STRUCT__entry(
DD_DEV_ENTRY(dd_from_ibdev(qp->ibqp.device))
@@ -318,17 +318,17 @@ DECLARE_EVENT_CLASS(hfi1_qpsleepwakeup_template,
);
DEFINE_EVENT(hfi1_qpsleepwakeup_template, hfi1_qpwakeup,
- TP_PROTO(struct hfi1_qp *qp, u32 flags),
+ TP_PROTO(struct rvt_qp *qp, u32 flags),
TP_ARGS(qp, flags));
DEFINE_EVENT(hfi1_qpsleepwakeup_template, hfi1_qpsleep,
- TP_PROTO(struct hfi1_qp *qp, u32 flags),
+ TP_PROTO(struct rvt_qp *qp, u32 flags),
TP_ARGS(qp, flags));
#undef TRACE_SYSTEM
#define TRACE_SYSTEM hfi1_qphash
DECLARE_EVENT_CLASS(hfi1_qphash_template,
- TP_PROTO(struct hfi1_qp *qp, u32 bucket),
+ TP_PROTO(struct rvt_qp *qp, u32 bucket),
TP_ARGS(qp, bucket),
TP_STRUCT__entry(
DD_DEV_ENTRY(dd_from_ibdev(qp->ibqp.device))
@@ -349,11 +349,11 @@ DECLARE_EVENT_CLASS(hfi1_qphash_template,
);
DEFINE_EVENT(hfi1_qphash_template, hfi1_qpinsert,
- TP_PROTO(struct hfi1_qp *qp, u32 bucket),
+ TP_PROTO(struct rvt_qp *qp, u32 bucket),
TP_ARGS(qp, bucket));
DEFINE_EVENT(hfi1_qphash_template, hfi1_qpremove,
- TP_PROTO(struct hfi1_qp *qp, u32 bucket),
+ TP_PROTO(struct rvt_qp *qp, u32 bucket),
TP_ARGS(qp, bucket));
#undef TRACE_SYSTEM
@@ -1253,7 +1253,7 @@ TRACE_EVENT(hfi1_sdma_state,
#define TRACE_SYSTEM hfi1_rc
DECLARE_EVENT_CLASS(hfi1_rc_template,
- TP_PROTO(struct hfi1_qp *qp, u32 psn),
+ TP_PROTO(struct rvt_qp *qp, u32 psn),
TP_ARGS(qp, psn),
TP_STRUCT__entry(
DD_DEV_ENTRY(dd_from_ibdev(qp->ibqp.device))
@@ -1292,22 +1292,22 @@ DECLARE_EVENT_CLASS(hfi1_rc_template,
);
DEFINE_EVENT(hfi1_rc_template, hfi1_rc_sendcomplete,
- TP_PROTO(struct hfi1_qp *qp, u32 psn),
+ TP_PROTO(struct rvt_qp *qp, u32 psn),
TP_ARGS(qp, psn)
);
DEFINE_EVENT(hfi1_rc_template, hfi1_rc_ack,
- TP_PROTO(struct hfi1_qp *qp, u32 psn),
+ TP_PROTO(struct rvt_qp *qp, u32 psn),
TP_ARGS(qp, psn)
);
DEFINE_EVENT(hfi1_rc_template, hfi1_rc_timeout,
- TP_PROTO(struct hfi1_qp *qp, u32 psn),
+ TP_PROTO(struct rvt_qp *qp, u32 psn),
TP_ARGS(qp, psn)
);
DEFINE_EVENT(hfi1_rc_template, hfi1_rc_rcv_error,
- TP_PROTO(struct hfi1_qp *qp, u32 psn),
+ TP_PROTO(struct rvt_qp *qp, u32 psn),
TP_ARGS(qp, psn)
);
diff --git a/drivers/staging/rdma/hfi1/uc.c b/drivers/staging/rdma/hfi1/uc.c
index 1908a28..6686331 100644
--- a/drivers/staging/rdma/hfi1/uc.c
+++ b/drivers/staging/rdma/hfi1/uc.c
@@ -61,11 +61,11 @@
*
* Return 1 if constructed; otherwise, return 0.
*/
-int hfi1_make_uc_req(struct hfi1_qp *qp)
+int hfi1_make_uc_req(struct rvt_qp *qp)
{
struct hfi1_qp_priv *priv = qp->priv;
struct hfi1_other_headers *ohdr;
- struct hfi1_swqe *wqe;
+ struct rvt_swqe *wqe;
unsigned long flags;
u32 hwords = 5;
u32 bth0 = 0;
@@ -267,7 +267,7 @@ void hfi1_uc_rcv(struct hfi1_packet *packet)
u32 rcv_flags = packet->rcv_flags;
void *data = packet->ebuf;
u32 tlen = packet->tlen;
- struct hfi1_qp *qp = packet->qp;
+ struct rvt_qp *qp = packet->qp;
struct hfi1_other_headers *ohdr = packet->ohdr;
u32 bth0, opcode;
u32 hdrsize = packet->hlen;
@@ -492,8 +492,8 @@ rdma_first:
int ok;
/* Check rkey */
- ok = hfi1_rkey_ok(qp, &qp->r_sge.sge, qp->r_len,
- vaddr, rkey, IB_ACCESS_REMOTE_WRITE);
+ ok = rvt_rkey_ok(qp, &qp->r_sge.sge, qp->r_len,
+ vaddr, rkey, IB_ACCESS_REMOTE_WRITE);
if (unlikely(!ok))
goto drop;
qp->r_sge.num_sge = 1;
diff --git a/drivers/staging/rdma/hfi1/ud.c b/drivers/staging/rdma/hfi1/ud.c
index 24b6077..b14641c 100644
--- a/drivers/staging/rdma/hfi1/ud.c
+++ b/drivers/staging/rdma/hfi1/ud.c
@@ -65,15 +65,15 @@
* Note that the receive interrupt handler may be calling hfi1_ud_rcv()
* while this is being called.
*/
-static void ud_loopback(struct hfi1_qp *sqp, struct hfi1_swqe *swqe)
+static void ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
{
struct hfi1_ibport *ibp = to_iport(sqp->ibqp.device, sqp->port_num);
struct hfi1_pportdata *ppd;
- struct hfi1_qp *qp;
+ struct rvt_qp *qp;
struct ib_ah_attr *ah_attr;
unsigned long flags;
- struct hfi1_sge_state ssge;
- struct hfi1_sge *sge;
+ struct rvt_sge_state ssge;
+ struct rvt_sge *sge;
struct ib_wc wc;
u32 length;
enum ib_qp_type sqptype, dqptype;
@@ -263,14 +263,14 @@ drop:
*
* Return 1 if constructed; otherwise, return 0.
*/
-int hfi1_make_ud_req(struct hfi1_qp *qp)
+int hfi1_make_ud_req(struct rvt_qp *qp)
{
struct hfi1_qp_priv *priv = qp->priv;
struct hfi1_other_headers *ohdr;
struct ib_ah_attr *ah_attr;
struct hfi1_pportdata *ppd;
struct hfi1_ibport *ibp;
- struct hfi1_swqe *wqe;
+ struct rvt_swqe *wqe;
unsigned long flags;
u32 nwords;
u32 extra_bytes;
@@ -478,7 +478,7 @@ int hfi1_lookup_pkey_idx(struct hfi1_ibport *ibp, u16 pkey)
return -1;
}
-void return_cnp(struct hfi1_ibport *ibp, struct hfi1_qp *qp, u32 remote_qpn,
+void return_cnp(struct hfi1_ibport *ibp, struct rvt_qp *qp, u32 remote_qpn,
u32 pkey, u32 slid, u32 dlid, u8 sc5,
const struct ib_grh *old_grh)
{
@@ -552,7 +552,7 @@ void return_cnp(struct hfi1_ibport *ibp, struct hfi1_qp *qp, u32 remote_qpn,
* opa_smp_check() returns 0 if all checks succeed, 1 otherwise.
*/
static int opa_smp_check(struct hfi1_ibport *ibp, u16 pkey, u8 sc5,
- struct hfi1_qp *qp, u16 slid, struct opa_smp *smp)
+ struct rvt_qp *qp, u16 slid, struct opa_smp *smp)
{
struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
@@ -656,7 +656,7 @@ void hfi1_ud_rcv(struct hfi1_packet *packet)
u32 rcv_flags = packet->rcv_flags;
void *data = packet->ebuf;
u32 tlen = packet->tlen;
- struct hfi1_qp *qp = packet->qp;
+ struct rvt_qp *qp = packet->qp;
bool has_grh = rcv_flags & HFI1_HAS_GRH;
bool sc4_bit = has_sc4_bit(packet);
u8 sc;
diff --git a/drivers/staging/rdma/hfi1/verbs.c b/drivers/staging/rdma/hfi1/verbs.c
index 6d69491..38069e6 100644
--- a/drivers/staging/rdma/hfi1/verbs.c
+++ b/drivers/staging/rdma/hfi1/verbs.c
@@ -65,7 +65,7 @@
#include "qp.h"
#include "sdma.h"
-unsigned int hfi1_lkey_table_size = 16;
+static unsigned int hfi1_lkey_table_size = 16;
module_param_named(lkey_table_size, hfi1_lkey_table_size, uint,
S_IRUGO);
MODULE_PARM_DESC(lkey_table_size,
@@ -162,7 +162,7 @@ static inline struct hfi1_ucontext *to_iucontext(struct ib_ucontext
return container_of(ibucontext, struct hfi1_ucontext, ibucontext);
}
-static inline void _hfi1_schedule_send(struct hfi1_qp *qp);
+static inline void _hfi1_schedule_send(struct rvt_qp *qp);
/*
* Translate ib_wr_opcode into ib_wc_opcode.
@@ -276,11 +276,11 @@ __be64 ib_hfi1_sys_image_guid;
* @length: the length of the data
*/
void hfi1_copy_sge(
- struct hfi1_sge_state *ss,
+ struct rvt_sge_state *ss,
void *data, u32 length,
int release)
{
- struct hfi1_sge *sge = &ss->sge;
+ struct rvt_sge *sge = &ss->sge;
while (length) {
u32 len = sge->length;
@@ -296,7 +296,7 @@ void hfi1_copy_sge(
sge->sge_length -= len;
if (sge->sge_length == 0) {
if (release)
- hfi1_put_mr(sge->mr);
+ rvt_put_mr(sge->mr);
if (--ss->num_sge)
*sge = *ss->sg_list++;
} else if (sge->length == 0 && sge->mr->lkey) {
@@ -320,9 +320,9 @@ void hfi1_copy_sge(
* @ss: the SGE state
* @length: the number of bytes to skip
*/
-void hfi1_skip_sge(struct hfi1_sge_state *ss, u32 length, int release)
+void hfi1_skip_sge(struct rvt_sge_state *ss, u32 length, int release)
{
- struct hfi1_sge *sge = &ss->sge;
+ struct rvt_sge *sge = &ss->sge;
while (length) {
u32 len = sge->length;
@@ -337,7 +337,7 @@ void hfi1_skip_sge(struct hfi1_sge_state *ss, u32 length, int release)
sge->sge_length -= len;
if (sge->sge_length == 0) {
if (release)
- hfi1_put_mr(sge->mr);
+ rvt_put_mr(sge->mr);
if (--ss->num_sge)
*sge = *ss->sg_list++;
} else if (sge->length == 0 && sge->mr->lkey) {
@@ -360,9 +360,9 @@ void hfi1_skip_sge(struct hfi1_sge_state *ss, u32 length, int release)
* @qp: the QP to post on
* @wr: the work request to send
*/
-static int post_one_send(struct hfi1_qp *qp, struct ib_send_wr *wr)
+static int post_one_send(struct rvt_qp *qp, struct ib_send_wr *wr)
{
- struct hfi1_swqe *wqe;
+ struct rvt_swqe *wqe;
u32 next;
int i;
int j;
@@ -412,7 +412,7 @@ static int post_one_send(struct hfi1_qp *qp, struct ib_send_wr *wr)
if (next == qp->s_last)
return -ENOMEM;
- rkt = &to_idev(qp->ibqp.device)->lk_table;
+ rkt = &to_idev(qp->ibqp.device)->rdi.lk_table;
pd = ibpd_to_rvtpd(qp->ibqp.pd);
wqe = get_swqe_ptr(qp, qp->s_head);
@@ -441,8 +441,8 @@ static int post_one_send(struct hfi1_qp *qp, struct ib_send_wr *wr)
if (length == 0)
continue;
- ok = hfi1_lkey_ok(rkt, pd, &wqe->sg_list[j],
- &wr->sg_list[i], acc);
+ ok = rvt_lkey_ok(rkt, pd, &wqe->sg_list[j],
+ &wr->sg_list[i], acc);
if (!ok)
goto bail_inval_free;
wqe->length += length;
@@ -465,9 +465,9 @@ static int post_one_send(struct hfi1_qp *qp, struct ib_send_wr *wr)
bail_inval_free:
/* release mr holds */
while (j) {
- struct hfi1_sge *sge = &wqe->sg_list[--j];
+ struct rvt_sge *sge = &wqe->sg_list[--j];
- hfi1_put_mr(sge->mr);
+ rvt_put_mr(sge->mr);
}
return -EINVAL;
}
@@ -483,7 +483,7 @@ bail_inval_free:
static int post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
struct ib_send_wr **bad_wr)
{
- struct hfi1_qp *qp = to_iqp(ibqp);
+ struct rvt_qp *qp = to_iqp(ibqp);
struct hfi1_qp_priv *priv = qp->priv;
int err = 0;
int call_send;
@@ -529,8 +529,8 @@ bail:
static int post_receive(struct ib_qp *ibqp, struct ib_recv_wr *wr,
struct ib_recv_wr **bad_wr)
{
- struct hfi1_qp *qp = to_iqp(ibqp);
- struct hfi1_rwq *wq = qp->r_rq.wq;
+ struct rvt_qp *qp = to_iqp(ibqp);
+ struct rvt_rwq *wq = qp->r_rq.wq;
unsigned long flags;
int ret;
@@ -542,7 +542,7 @@ static int post_receive(struct ib_qp *ibqp, struct ib_recv_wr *wr,
}
for (; wr; wr = wr->next) {
- struct hfi1_rwqe *wqe;
+ struct rvt_rwqe *wqe;
u32 next;
int i;
@@ -694,7 +694,7 @@ static void mem_timer(unsigned long data)
{
struct hfi1_ibdev *dev = (struct hfi1_ibdev *)data;
struct list_head *list = &dev->memwait;
- struct hfi1_qp *qp = NULL;
+ struct rvt_qp *qp = NULL;
struct iowait *wait;
unsigned long flags;
struct hfi1_qp_priv *priv;
@@ -715,9 +715,9 @@ static void mem_timer(unsigned long data)
hfi1_qp_wakeup(qp, HFI1_S_WAIT_KMEM);
}
-void update_sge(struct hfi1_sge_state *ss, u32 length)
+void update_sge(struct rvt_sge_state *ss, u32 length)
{
- struct hfi1_sge *sge = &ss->sge;
+ struct rvt_sge *sge = &ss->sge;
sge->vaddr += length;
sge->length -= length;
@@ -737,7 +737,7 @@ void update_sge(struct hfi1_sge_state *ss, u32 length)
}
static noinline struct verbs_txreq *__get_txreq(struct hfi1_ibdev *dev,
- struct hfi1_qp *qp)
+ struct rvt_qp *qp)
{
struct hfi1_qp_priv *priv = qp->priv;
struct verbs_txreq *tx;
@@ -764,7 +764,7 @@ static noinline struct verbs_txreq *__get_txreq(struct hfi1_ibdev *dev,
}
static inline struct verbs_txreq *get_txreq(struct hfi1_ibdev *dev,
- struct hfi1_qp *qp)
+ struct rvt_qp *qp)
{
struct verbs_txreq *tx;
@@ -782,7 +782,7 @@ static inline struct verbs_txreq *get_txreq(struct hfi1_ibdev *dev,
void hfi1_put_txreq(struct verbs_txreq *tx)
{
struct hfi1_ibdev *dev;
- struct hfi1_qp *qp;
+ struct rvt_qp *qp;
unsigned long flags;
unsigned int seq;
struct hfi1_qp_priv *priv;
@@ -791,7 +791,7 @@ void hfi1_put_txreq(struct verbs_txreq *tx)
dev = to_idev(qp->ibqp.device);
if (tx->mr) {
- hfi1_put_mr(tx->mr);
+ rvt_put_mr(tx->mr);
tx->mr = NULL;
}
sdma_txclean(dd_from_dev(dev), &tx->txreq);
@@ -830,7 +830,7 @@ static void verbs_sdma_complete(
{
struct verbs_txreq *tx =
container_of(cookie, struct verbs_txreq, txreq);
- struct hfi1_qp *qp = tx->qp;
+ struct rvt_qp *qp = tx->qp;
spin_lock(&qp->s_lock);
if (tx->wqe)
@@ -858,7 +858,7 @@ static void verbs_sdma_complete(
hfi1_put_txreq(tx);
}
-static int wait_kmem(struct hfi1_ibdev *dev, struct hfi1_qp *qp)
+static int wait_kmem(struct hfi1_ibdev *dev, struct rvt_qp *qp)
{
struct hfi1_qp_priv *priv = qp->priv;
unsigned long flags;
@@ -891,12 +891,12 @@ static int wait_kmem(struct hfi1_ibdev *dev, struct hfi1_qp *qp)
*/
static int build_verbs_ulp_payload(
struct sdma_engine *sde,
- struct hfi1_sge_state *ss,
+ struct rvt_sge_state *ss,
u32 length,
struct verbs_txreq *tx)
{
- struct hfi1_sge *sg_list = ss->sg_list;
- struct hfi1_sge sge = ss->sge;
+ struct rvt_sge *sg_list = ss->sg_list;
+ struct rvt_sge sge = ss->sge;
u8 num_sge = ss->num_sge;
u32 len;
int ret = 0;
@@ -939,7 +939,7 @@ bail_txadd:
/* New API */
static int build_verbs_tx_desc(
struct sdma_engine *sde,
- struct hfi1_sge_state *ss,
+ struct rvt_sge_state *ss,
u32 length,
struct verbs_txreq *tx,
struct ahg_ib_header *ahdr,
@@ -1006,13 +1006,13 @@ bail_txadd:
return ret;
}
-int hfi1_verbs_send_dma(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
+int hfi1_verbs_send_dma(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
u64 pbc)
{
struct hfi1_qp_priv *priv = qp->priv;
struct ahg_ib_header *ahdr = priv->s_hdr;
u32 hdrwords = qp->s_hdrwords;
- struct hfi1_sge_state *ss = qp->s_cur_sge;
+ struct rvt_sge_state *ss = qp->s_cur_sge;
u32 len = qp->s_cur_size;
u32 plen = hdrwords + ((len + 3) >> 2) + 2; /* includes pbc */
struct hfi1_ibdev *dev = ps->dev;
@@ -1079,7 +1079,7 @@ bail_tx:
* If we are now in the error state, return zero to flush the
* send work request.
*/
-static int no_bufs_available(struct hfi1_qp *qp, struct send_context *sc)
+static int no_bufs_available(struct rvt_qp *qp, struct send_context *sc)
{
struct hfi1_qp_priv *priv = qp->priv;
struct hfi1_devdata *dd = sc->dd;
@@ -1118,7 +1118,7 @@ static int no_bufs_available(struct hfi1_qp *qp, struct send_context *sc)
return ret;
}
-struct send_context *qp_to_send_context(struct hfi1_qp *qp, u8 sc5)
+struct send_context *qp_to_send_context(struct rvt_qp *qp, u8 sc5)
{
struct hfi1_devdata *dd = dd_from_ibdev(qp->ibqp.device);
struct hfi1_pportdata *ppd = dd->pport + (qp->port_num - 1);
@@ -1130,13 +1130,13 @@ struct send_context *qp_to_send_context(struct hfi1_qp *qp, u8 sc5)
return dd->vld[vl].sc;
}
-int hfi1_verbs_send_pio(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
+int hfi1_verbs_send_pio(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
u64 pbc)
{
struct hfi1_qp_priv *priv = qp->priv;
struct ahg_ib_header *ahdr = priv->s_hdr;
u32 hdrwords = qp->s_hdrwords;
- struct hfi1_sge_state *ss = qp->s_cur_sge;
+ struct rvt_sge_state *ss = qp->s_cur_sge;
u32 len = qp->s_cur_size;
u32 dwords = (len + 3) >> 2;
u32 plen = hdrwords + dwords + 2; /* includes pbc */
@@ -1208,7 +1208,7 @@ int hfi1_verbs_send_pio(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
trace_output_ibhdr(dd_from_ibdev(qp->ibqp.device), &ahdr->ibh);
if (qp->s_rdma_mr) {
- hfi1_put_mr(qp->s_rdma_mr);
+ rvt_put_mr(qp->s_rdma_mr);
qp->s_rdma_mr = NULL;
}
@@ -1255,7 +1255,7 @@ static inline int egress_pkey_matches_entry(u16 pkey, u16 ent)
*/
static inline int egress_pkey_check(struct hfi1_pportdata *ppd,
struct hfi1_ib_header *hdr,
- struct hfi1_qp *qp)
+ struct rvt_qp *qp)
{
struct hfi1_qp_priv *priv = qp->priv;
struct hfi1_other_headers *ohdr;
@@ -1318,7 +1318,7 @@ bad:
* Return zero if packet is sent or queued OK.
* Return non-zero and clear qp->s_flags HFI1_S_BUSY otherwise.
*/
-int hfi1_verbs_send(struct hfi1_qp *qp, struct hfi1_pkt_state *ps)
+int hfi1_verbs_send(struct rvt_qp *qp, struct hfi1_pkt_state *ps)
{
struct hfi1_devdata *dd = dd_from_ibdev(qp->ibqp.device);
struct hfi1_qp_priv *priv = qp->priv;
@@ -1401,8 +1401,8 @@ static int query_device(struct ib_device *ibdev,
props->max_cq = hfi1_max_cqs;
props->max_ah = hfi1_max_ahs;
props->max_cqe = hfi1_max_cqes;
- props->max_mr = dev->lk_table.max;
- props->max_fmr = dev->lk_table.max;
+ props->max_mr = dev->rdi.lk_table.max;
+ props->max_fmr = dev->rdi.lk_table.max;
props->max_map_per_fmr = 32767;
props->max_pd = dev->rdi.dparms.props.max_pd;
props->max_qp_rd_atom = HFI1_MAX_RDMA_ATOMIC;
@@ -1656,7 +1656,7 @@ struct ib_ah *hfi1_create_qp0_ah(struct hfi1_ibport *ibp, u16 dlid)
{
struct ib_ah_attr attr;
struct ib_ah *ah = ERR_PTR(-EINVAL);
- struct hfi1_qp *qp0;
+ struct rvt_qp *qp0;
memset(&attr, 0, sizeof(attr));
attr.dlid = dlid;
@@ -1771,7 +1771,7 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
struct hfi1_ibdev *dev = &dd->verbs_dev;
struct ib_device *ibdev = &dev->rdi.ibdev;
struct hfi1_pportdata *ppd = dd->pport;
- unsigned i, lk_tab_size;
+ unsigned i;
int ret;
size_t lcpysz = IB_DEVICE_NAME_MAX;
u16 descq_cnt;
@@ -1795,29 +1795,6 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
dev->mem_timer.function = mem_timer;
dev->mem_timer.data = (unsigned long) dev;
- /*
- * The top hfi1_lkey_table_size bits are used to index the
- * table. The lower 8 bits can be owned by the user (copied from
- * the LKEY). The remaining bits act as a generation number or tag.
- */
- spin_lock_init(&dev->lk_table.lock);
- dev->lk_table.max = 1 << hfi1_lkey_table_size;
- /* ensure generation is at least 4 bits (keys.c) */
- if (hfi1_lkey_table_size > RVT_MAX_LKEY_TABLE_BITS) {
- dd_dev_warn(dd, "lkey bits %u too large, reduced to %u\n",
- hfi1_lkey_table_size, RVT_MAX_LKEY_TABLE_BITS);
- hfi1_lkey_table_size = RVT_MAX_LKEY_TABLE_BITS;
- }
- lk_tab_size = dev->lk_table.max * sizeof(*dev->lk_table.table);
- dev->lk_table.table = (struct rvt_mregion __rcu **)
- vmalloc(lk_tab_size);
- if (dev->lk_table.table == NULL) {
- ret = -ENOMEM;
- goto err_lk;
- }
- RCU_INIT_POINTER(dev->dma_mr, NULL);
- for (i = 0; i < dev->lk_table.max; i++)
- RCU_INIT_POINTER(dev->lk_table.table[i], NULL);
INIT_LIST_HEAD(&dev->pending_mmaps);
spin_lock_init(&dev->pending_lock);
seqlock_init(&dev->iowait_lock);
@@ -1916,15 +1893,16 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
ibdev->resize_cq = hfi1_resize_cq;
ibdev->poll_cq = hfi1_poll_cq;
ibdev->req_notify_cq = hfi1_req_notify_cq;
- ibdev->get_dma_mr = hfi1_get_dma_mr;
- ibdev->reg_phys_mr = hfi1_reg_phys_mr;
- ibdev->reg_user_mr = hfi1_reg_user_mr;
- ibdev->dereg_mr = hfi1_dereg_mr;
- ibdev->alloc_mr = hfi1_alloc_mr;
- ibdev->alloc_fmr = hfi1_alloc_fmr;
- ibdev->map_phys_fmr = hfi1_map_phys_fmr;
- ibdev->unmap_fmr = hfi1_unmap_fmr;
- ibdev->dealloc_fmr = hfi1_dealloc_fmr;
+ ibdev->get_dma_mr = NULL;
+ ibdev->reg_phys_mr = NULL;
+ ibdev->reg_user_mr = NULL;
+ ibdev->dereg_mr = NULL;
+ ibdev->alloc_mr = NULL;
+ ibdev->map_mr_sg = NULL;
+ ibdev->alloc_fmr = NULL;
+ ibdev->map_phys_fmr = NULL;
+ ibdev->unmap_fmr = NULL;
+ ibdev->dealloc_fmr = NULL;
ibdev->attach_mcast = hfi1_multicast_attach;
ibdev->detach_mcast = hfi1_multicast_detach;
ibdev->process_mad = hfi1_process_mad;
@@ -1945,9 +1923,9 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
dd->verbs_dev.rdi.driver_f.notify_new_ah = hfi1_notify_new_ah;
dd->verbs_dev.rdi.dparms.props.max_ah = hfi1_max_ahs;
dd->verbs_dev.rdi.dparms.props.max_pd = hfi1_max_pds;
- dd->verbs_dev.rdi.flags = (RVT_FLAG_MR_INIT_DRIVER |
- RVT_FLAG_QP_INIT_DRIVER |
+ dd->verbs_dev.rdi.flags = (RVT_FLAG_QP_INIT_DRIVER |
RVT_FLAG_CQ_INIT_DRIVER);
+ dd->verbs_dev.rdi.dparms.lkey_table_size = hfi1_lkey_table_size;
ret = rvt_register_device(&dd->verbs_dev.rdi);
if (ret)
@@ -1970,8 +1948,6 @@ err_agents:
err_reg:
err_verbs_txreq:
kmem_cache_destroy(dev->verbs_txreq_cache);
- vfree(dev->lk_table.table);
-err_lk:
hfi1_qp_exit(dev);
err_qp_init:
dd_dev_err(dd, "cannot register verbs: %d!\n", -ret);
@@ -1993,13 +1969,10 @@ void hfi1_unregister_ib_device(struct hfi1_devdata *dd)
dd_dev_err(dd, "txwait list not empty!\n");
if (!list_empty(&dev->memwait))
dd_dev_err(dd, "memwait list not empty!\n");
- if (dev->dma_mr)
- dd_dev_err(dd, "DMA MR not NULL!\n");
hfi1_qp_exit(dev);
del_timer_sync(&dev->mem_timer);
kmem_cache_destroy(dev->verbs_txreq_cache);
- vfree(dev->lk_table.table);
}
void hfi1_cnp_rcv(struct hfi1_packet *packet)
@@ -2007,7 +1980,7 @@ void hfi1_cnp_rcv(struct hfi1_packet *packet)
struct hfi1_ibport *ibp = &packet->rcd->ppd->ibport_data;
struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
struct hfi1_ib_header *hdr = packet->hdr;
- struct hfi1_qp *qp = packet->qp;
+ struct rvt_qp *qp = packet->qp;
u32 lqpn, rqpn = 0;
u16 rlid = 0;
u8 sl, sc5, sc4_bit, svc_type;
diff --git a/drivers/staging/rdma/hfi1/verbs.h b/drivers/staging/rdma/hfi1/verbs.h
index 3f3815c..fec5e7b 100644
--- a/drivers/staging/rdma/hfi1/verbs.h
+++ b/drivers/staging/rdma/hfi1/verbs.h
@@ -222,7 +222,7 @@ struct tx_pio_header {
*/
struct hfi1_mcast_qp {
struct list_head list;
- struct hfi1_qp *qp;
+ struct rvt_qp *qp;
};
struct hfi1_mcast {
@@ -235,20 +235,6 @@ struct hfi1_mcast {
};
/*
- * This structure is used by hfi1_mmap() to validate an offset
- * when an mmap() request is made. The vm_area_struct then uses
- * this as its vm_private_data.
- */
-struct hfi1_mmap_info {
- struct list_head pending_mmaps;
- struct ib_ucontext *context;
- void *obj;
- __u64 offset;
- struct kref ref;
- unsigned size;
-};
-
-/*
* This structure is used to contain the head pointer, tail pointer,
* and completion queue entries as a single memory allocation so
* it can be mmap'ed into user space.
@@ -274,238 +260,26 @@ struct hfi1_cq {
u8 notify;
u8 triggered;
struct hfi1_cq_wc *queue;
- struct hfi1_mmap_info *ip;
-};
-
-/*
- * These keep track of the copy progress within a memory region.
- * Used by the verbs layer.
- */
-struct hfi1_sge {
- struct rvt_mregion *mr;
- void *vaddr; /* kernel virtual address of segment */
- u32 sge_length; /* length of the SGE */
- u32 length; /* remaining length of the segment */
- u16 m; /* current index: mr->map[m] */
- u16 n; /* current index: mr->map[m]->segs[n] */
-};
-
-/* Memory region */
-struct hfi1_mr {
- struct ib_mr ibmr;
- struct ib_umem *umem;
- struct rvt_mregion mr; /* must be last */
-};
-
-/*
- * Send work request queue entry.
- * The size of the sg_list is determined when the QP is created and stored
- * in qp->s_max_sge.
- */
-struct hfi1_swqe {
- union {
- struct ib_send_wr wr; /* don't use wr.sg_list */
- struct ib_rdma_wr rdma_wr;
- struct ib_atomic_wr atomic_wr;
- struct ib_ud_wr ud_wr;
- };
- u32 psn; /* first packet sequence number */
- u32 lpsn; /* last packet sequence number */
- u32 ssn; /* send sequence number */
- u32 length; /* total length of data in sg_list */
- struct hfi1_sge sg_list[0];
-};
-
-/*
- * Receive work request queue entry.
- * The size of the sg_list is determined when the QP (or SRQ) is created
- * and stored in qp->r_rq.max_sge (or srq->rq.max_sge).
- */
-struct hfi1_rwqe {
- u64 wr_id;
- u8 num_sge;
- struct ib_sge sg_list[0];
-};
-
-/*
- * This structure is used to contain the head pointer, tail pointer,
- * and receive work queue entries as a single memory allocation so
- * it can be mmap'ed into user space.
- * Note that the wq array elements are variable size so you can't
- * just index into the array to get the N'th element;
- * use get_rwqe_ptr() instead.
- */
-struct hfi1_rwq {
- u32 head; /* new work requests posted to the head */
- u32 tail; /* receives pull requests from here. */
- struct hfi1_rwqe wq[0];
-};
-
-struct hfi1_rq {
- struct hfi1_rwq *wq;
- u32 size; /* size of RWQE array */
- u8 max_sge;
- /* protect changes in this struct */
- spinlock_t lock ____cacheline_aligned_in_smp;
+ struct rvt_mmap_info *ip;
};
struct hfi1_srq {
struct ib_srq ibsrq;
- struct hfi1_rq rq;
- struct hfi1_mmap_info *ip;
+ struct rvt_rq rq;
+ struct rvt_mmap_info *ip;
/* send signal when number of RWQEs < limit */
u32 limit;
};
-struct hfi1_sge_state {
- struct hfi1_sge *sg_list; /* next SGE to be used if any */
- struct hfi1_sge sge; /* progress state for the current SGE */
- u32 total_len;
- u8 num_sge;
-};
-
-/*
- * This structure holds the information that the send tasklet needs
- * to send a RDMA read response or atomic operation.
- */
-struct hfi1_ack_entry {
- u8 opcode;
- u8 sent;
- u32 psn;
- u32 lpsn;
- union {
- struct hfi1_sge rdma_sge;
- u64 atomic_data;
- };
-};
-
/*
* hfi1 specific data structures that will be hidden from rvt after the queue
* pair is made common
*/
-struct hfi1_qp;
struct hfi1_qp_priv {
struct ahg_ib_header *s_hdr; /* next packet header to send */
u8 s_sc; /* SC[0..4] for next packet */
struct iowait s_iowait;
- struct hfi1_qp *owner;
-};
-
-/*
- * Variables prefixed with s_ are for the requester (sender).
- * Variables prefixed with r_ are for the responder (receiver).
- * Variables prefixed with ack_ are for responder replies.
- *
- * Common variables are protected by both r_rq.lock and s_lock in that order
- * which only happens in modify_qp() or changing the QP 'state'.
- */
-struct hfi1_qp {
- struct ib_qp ibqp;
- void *priv;
- /* read mostly fields above and below */
- struct ib_ah_attr remote_ah_attr;
- struct ib_ah_attr alt_ah_attr;
- struct hfi1_qp __rcu *next; /* link list for QPN hash table */
- struct hfi1_swqe *s_wq; /* send work queue */
- struct hfi1_mmap_info *ip;
- unsigned long timeout_jiffies; /* computed from timeout */
-
- enum ib_mtu path_mtu;
- int srate_mbps; /* s_srate (below) converted to Mbit/s */
- u32 remote_qpn;
- u32 pmtu; /* decoded from path_mtu */
- u32 qkey; /* QKEY for this QP (for UD or RD) */
- u32 s_size; /* send work queue size */
- u32 s_rnr_timeout; /* number of milliseconds for RNR timeout */
- u32 s_ahgpsn; /* set to the psn in the copy of the header */
-
- u8 state; /* QP state */
- u8 allowed_ops; /* high order bits of allowed opcodes */
- u8 qp_access_flags;
- u8 alt_timeout; /* Alternate path timeout for this QP */
- u8 timeout; /* Timeout for this QP */
- u8 s_srate;
- u8 s_mig_state;
- u8 port_num;
- u8 s_pkey_index; /* PKEY index to use */
- u8 s_alt_pkey_index; /* Alternate path PKEY index to use */
- u8 r_max_rd_atomic; /* max number of RDMA read/atomic to receive */
- u8 s_max_rd_atomic; /* max number of RDMA read/atomic to send */
- u8 s_retry_cnt; /* number of times to retry */
- u8 s_rnr_retry_cnt;
- u8 r_min_rnr_timer; /* retry timeout value for RNR NAKs */
- u8 s_max_sge; /* size of s_wq->sg_list */
- u8 s_draining;
-
- /* start of read/write fields */
- atomic_t refcount ____cacheline_aligned_in_smp;
- wait_queue_head_t wait;
-
-
- struct hfi1_ack_entry s_ack_queue[HFI1_MAX_RDMA_ATOMIC + 1]
- ____cacheline_aligned_in_smp;
- struct hfi1_sge_state s_rdma_read_sge;
-
- spinlock_t r_lock ____cacheline_aligned_in_smp; /* used for APM */
- unsigned long r_aflags;
- u64 r_wr_id; /* ID for current receive WQE */
- u32 r_ack_psn; /* PSN for next ACK or atomic ACK */
- u32 r_len; /* total length of r_sge */
- u32 r_rcv_len; /* receive data len processed */
- u32 r_psn; /* expected rcv packet sequence number */
- u32 r_msn; /* message sequence number */
-
- u8 r_adefered; /* number of acks defered */
- u8 r_state; /* opcode of last packet received */
- u8 r_flags;
- u8 r_head_ack_queue; /* index into s_ack_queue[] */
-
- struct list_head rspwait; /* link for waiting to respond */
-
- struct hfi1_sge_state r_sge; /* current receive data */
- struct hfi1_rq r_rq; /* receive work queue */
-
- spinlock_t s_lock ____cacheline_aligned_in_smp;
- struct hfi1_sge_state *s_cur_sge;
- u32 s_flags;
- struct hfi1_swqe *s_wqe;
- struct hfi1_sge_state s_sge; /* current send request data */
- struct rvt_mregion *s_rdma_mr;
- struct sdma_engine *s_sde; /* current sde */
- u32 s_cur_size; /* size of send packet in bytes */
- u32 s_len; /* total length of s_sge */
- u32 s_rdma_read_len; /* total length of s_rdma_read_sge */
- u32 s_next_psn; /* PSN for next request */
- u32 s_last_psn; /* last response PSN processed */
- u32 s_sending_psn; /* lowest PSN that is being sent */
- u32 s_sending_hpsn; /* highest PSN that is being sent */
- u32 s_psn; /* current packet sequence number */
- u32 s_ack_rdma_psn; /* PSN for sending RDMA read responses */
- u32 s_ack_psn; /* PSN for acking sends and RDMA writes */
- u32 s_head; /* new entries added here */
- u32 s_tail; /* next entry to process */
- u32 s_cur; /* current work queue entry */
- u32 s_acked; /* last un-ACK'ed entry */
- u32 s_last; /* last completed entry */
- u32 s_ssn; /* SSN of tail entry */
- u32 s_lsn; /* limit sequence number (credit) */
- u16 s_hdrwords; /* size of s_hdr in 32 bit words */
- u16 s_rdma_ack_cnt;
- s8 s_ahgidx;
- u8 s_state; /* opcode of last packet sent */
- u8 s_ack_state; /* opcode of packet to ACK */
- u8 s_nak_state; /* non-zero if NAK is pending */
- u8 r_nak_state; /* non-zero if NAK is pending */
- u8 s_retry; /* requester retry counter */
- u8 s_rnr_retry; /* requester RNR retry counter */
- u8 s_num_rd_atomic; /* number of RDMA read/atomic pending */
- u8 s_tail_ack_queue; /* index into s_ack_queue[] */
-
- struct hfi1_sge_state s_ack_rdma_sge;
- struct timer_list s_timer;
-
- struct hfi1_sge r_sg_list[0] /* verified SGEs */
- ____cacheline_aligned_in_smp;
+ struct rvt_qp *owner;
};
/*
@@ -599,27 +373,27 @@ struct hfi1_pkt_state {
#define HFI1_PSN_CREDIT 16
/*
- * Since struct hfi1_swqe is not a fixed size, we can't simply index into
+ * Since struct rvt_swqe is not a fixed size, we can't simply index into
* struct hfi1_qp.s_wq. This function does the array index computation.
*/
-static inline struct hfi1_swqe *get_swqe_ptr(struct hfi1_qp *qp,
- unsigned n)
+static inline struct rvt_swqe *get_swqe_ptr(struct rvt_qp *qp,
+ unsigned n)
{
- return (struct hfi1_swqe *)((char *)qp->s_wq +
- (sizeof(struct hfi1_swqe) +
+ return (struct rvt_swqe *)((char *)qp->s_wq +
+ (sizeof(struct rvt_swqe) +
qp->s_max_sge *
- sizeof(struct hfi1_sge)) * n);
+ sizeof(struct rvt_sge)) * n);
}
/*
- * Since struct hfi1_rwqe is not a fixed size, we can't simply index into
- * struct hfi1_rwq.wq. This function does the array index computation.
+ * Since struct rvt_rwqe is not a fixed size, we can't simply index into
+ * struct rvt_rwq.wq. This function does the array index computation.
*/
-static inline struct hfi1_rwqe *get_rwqe_ptr(struct hfi1_rq *rq, unsigned n)
+static inline struct rvt_rwqe *get_rwqe_ptr(struct rvt_rq *rq, unsigned n)
{
- return (struct hfi1_rwqe *)
+ return (struct rvt_rwqe *)
((char *) rq->wq->wq +
- (sizeof(struct hfi1_rwqe) +
+ (sizeof(struct rvt_rwqe) +
rq->max_sge * sizeof(struct ib_sge)) * n);
}
@@ -643,7 +417,7 @@ static inline void inc_opstats(
}
struct hfi1_ibport {
- struct hfi1_qp __rcu *qp[2];
+ struct rvt_qp __rcu *qp[2];
struct ib_mad_agent *send_agent; /* agent for SMI (traps) */
struct rvt_ah *sm_ah;
struct rvt_ah *smi_ah;
@@ -706,12 +480,10 @@ struct hfi1_ibdev {
struct list_head pending_mmaps;
spinlock_t mmap_offset_lock; /* protect mmap_offset */
u32 mmap_offset;
- struct rvt_mregion __rcu *dma_mr;
struct hfi1_qp_ibdev *qp_dev;
/* QP numbers are shared by all IB ports */
- struct rvt_lkey_table lk_table;
/* protect wait lists */
seqlock_t iowait_lock;
struct list_head txwait; /* list for wait verbs_txreq */
@@ -760,11 +532,6 @@ struct hfi1_verbs_counters {
u32 vl15_dropped;
};
-static inline struct hfi1_mr *to_imr(struct ib_mr *ibmr)
-{
- return container_of(ibmr, struct hfi1_mr, ibmr);
-}
-
static inline struct hfi1_cq *to_icq(struct ib_cq *ibcq)
{
return container_of(ibcq, struct hfi1_cq, ibcq);
@@ -775,9 +542,9 @@ static inline struct hfi1_srq *to_isrq(struct ib_srq *ibsrq)
return container_of(ibsrq, struct hfi1_srq, ibsrq);
}
-static inline struct hfi1_qp *to_iqp(struct ib_qp *ibqp)
+static inline struct rvt_qp *to_iqp(struct ib_qp *ibqp)
{
- return container_of(ibqp, struct hfi1_qp, ibqp);
+ return container_of(ibqp, struct rvt_qp, ibqp);
}
static inline struct hfi1_ibdev *to_idev(struct ib_device *ibdev)
@@ -788,7 +555,7 @@ static inline struct hfi1_ibdev *to_idev(struct ib_device *ibdev)
return container_of(rdi, struct hfi1_ibdev, rdi);
}
-static inline struct hfi1_qp *iowait_to_qp(struct iowait *s_iowait)
+static inline struct rvt_qp *iowait_to_qp(struct iowait *s_iowait)
{
struct hfi1_qp_priv *priv;
@@ -800,7 +567,7 @@ static inline struct hfi1_qp *iowait_to_qp(struct iowait *s_iowait)
* Send if not busy or waiting for I/O and either
* a RC response is pending or we can process send work requests.
*/
-static inline int hfi1_send_ok(struct hfi1_qp *qp)
+static inline int hfi1_send_ok(struct rvt_qp *qp)
{
return !(qp->s_flags & (HFI1_S_BUSY | HFI1_S_ANY_WAIT_IO)) &&
(qp->s_hdrwords || (qp->s_flags & HFI1_S_RESP_PENDING) ||
@@ -890,12 +657,12 @@ int hfi1_mcast_tree_empty(struct hfi1_ibport *ibp);
struct verbs_txreq;
void hfi1_put_txreq(struct verbs_txreq *tx);
-int hfi1_verbs_send(struct hfi1_qp *qp, struct hfi1_pkt_state *ps);
+int hfi1_verbs_send(struct rvt_qp *qp, struct hfi1_pkt_state *ps);
-void hfi1_copy_sge(struct hfi1_sge_state *ss, void *data, u32 length,
+void hfi1_copy_sge(struct rvt_sge_state *ss, void *data, u32 length,
int release);
-void hfi1_skip_sge(struct hfi1_sge_state *ss, u32 length, int release);
+void hfi1_skip_sge(struct rvt_sge_state *ss, u32 length, int release);
void hfi1_cnp_rcv(struct hfi1_packet *packet);
@@ -907,7 +674,7 @@ void hfi1_rc_hdrerr(
struct hfi1_ctxtdata *rcd,
struct hfi1_ib_header *hdr,
u32 rcv_flags,
- struct hfi1_qp *qp);
+ struct rvt_qp *qp);
u8 ah_to_sc(struct ib_device *ibdev, struct ib_ah_attr *ah_attr);
@@ -915,24 +682,14 @@ struct ib_ah *hfi1_create_qp0_ah(struct hfi1_ibport *ibp, u16 dlid);
void hfi1_rc_rnr_retry(unsigned long arg);
-void hfi1_rc_send_complete(struct hfi1_qp *qp, struct hfi1_ib_header *hdr);
+void hfi1_rc_send_complete(struct rvt_qp *qp, struct hfi1_ib_header *hdr);
-void hfi1_rc_error(struct hfi1_qp *qp, enum ib_wc_status err);
+void hfi1_rc_error(struct rvt_qp *qp, enum ib_wc_status err);
void hfi1_ud_rcv(struct hfi1_packet *packet);
int hfi1_lookup_pkey_idx(struct hfi1_ibport *ibp, u16 pkey);
-int hfi1_alloc_lkey(struct rvt_mregion *mr, int dma_region);
-
-void hfi1_free_lkey(struct rvt_mregion *mr);
-
-int hfi1_lkey_ok(struct rvt_lkey_table *rkt, struct rvt_pd *pd,
- struct hfi1_sge *isge, struct ib_sge *sge, int acc);
-
-int hfi1_rkey_ok(struct hfi1_qp *qp, struct hfi1_sge *sge,
- u32 len, u64 vaddr, u32 rkey, int acc);
-
int hfi1_post_srq_receive(struct ib_srq *ibsrq, struct ib_recv_wr *wr,
struct ib_recv_wr **bad_wr);
@@ -970,47 +727,10 @@ int hfi1_req_notify_cq(
int hfi1_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata);
-struct ib_mr *hfi1_get_dma_mr(struct ib_pd *pd, int acc);
-
-struct ib_mr *hfi1_reg_phys_mr(struct ib_pd *pd,
- struct ib_phys_buf *buffer_list,
- int num_phys_buf, int acc, u64 *iova_start);
-
-struct ib_mr *hfi1_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
- u64 virt_addr, int mr_access_flags,
- struct ib_udata *udata);
-
-int hfi1_dereg_mr(struct ib_mr *ibmr);
-
-struct ib_mr *hfi1_alloc_mr(struct ib_pd *pd,
- enum ib_mr_type mr_type,
- u32 max_entries);
-
-struct ib_fmr *hfi1_alloc_fmr(struct ib_pd *pd, int mr_access_flags,
- struct ib_fmr_attr *fmr_attr);
-
-int hfi1_map_phys_fmr(struct ib_fmr *ibfmr, u64 *page_list,
- int list_len, u64 iova);
-
-int hfi1_unmap_fmr(struct list_head *fmr_list);
-
-int hfi1_dealloc_fmr(struct ib_fmr *ibfmr);
-
-static inline void hfi1_get_mr(struct rvt_mregion *mr)
-{
- atomic_inc(&mr->refcount);
-}
-
-static inline void hfi1_put_mr(struct rvt_mregion *mr)
-{
- if (unlikely(atomic_dec_and_test(&mr->refcount)))
- complete(&mr->comp);
-}
-
-static inline void hfi1_put_ss(struct hfi1_sge_state *ss)
+static inline void hfi1_put_ss(struct rvt_sge_state *ss)
{
while (ss->num_sge) {
- hfi1_put_mr(ss->sge.mr);
+ rvt_put_mr(ss->sge.mr);
if (--ss->num_sge)
ss->sge = *ss->sg_list++;
}
@@ -1018,38 +738,40 @@ static inline void hfi1_put_ss(struct hfi1_sge_state *ss)
void hfi1_release_mmap_info(struct kref *ref);
-struct hfi1_mmap_info *hfi1_create_mmap_info(struct hfi1_ibdev *dev, u32 size,
- struct ib_ucontext *context,
- void *obj);
+struct rvt_mmap_info *hfi1_create_mmap_info(struct hfi1_ibdev *dev, u32 size,
+ struct ib_ucontext *context,
+ void *obj);
-void hfi1_update_mmap_info(struct hfi1_ibdev *dev, struct hfi1_mmap_info *ip,
+void hfi1_update_mmap_info(struct hfi1_ibdev *dev, struct rvt_mmap_info *ip,
u32 size, void *obj);
int hfi1_mmap(struct ib_ucontext *context, struct vm_area_struct *vma);
-int hfi1_get_rwqe(struct hfi1_qp *qp, int wr_id_only);
+int hfi1_get_rwqe(struct rvt_qp *qp, int wr_id_only);
int hfi1_ruc_check_hdr(struct hfi1_ibport *ibp, struct hfi1_ib_header *hdr,
- int has_grh, struct hfi1_qp *qp, u32 bth0);
+ int has_grh, struct rvt_qp *qp, u32 bth0);
u32 hfi1_make_grh(struct hfi1_ibport *ibp, struct ib_grh *hdr,
struct ib_global_route *grh, u32 hwords, u32 nwords);
-void hfi1_make_ruc_header(struct hfi1_qp *qp, struct hfi1_other_headers *ohdr,
+void clear_ahg(struct rvt_qp *qp);
+
+void hfi1_make_ruc_header(struct rvt_qp *qp, struct hfi1_other_headers *ohdr,
u32 bth0, u32 bth2, int middle);
void hfi1_do_send(struct work_struct *work);
-void hfi1_send_complete(struct hfi1_qp *qp, struct hfi1_swqe *wqe,
+void hfi1_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe,
enum ib_wc_status status);
-void hfi1_send_rc_ack(struct hfi1_ctxtdata *, struct hfi1_qp *qp, int is_fecn);
+void hfi1_send_rc_ack(struct hfi1_ctxtdata *, struct rvt_qp *qp, int is_fecn);
-int hfi1_make_rc_req(struct hfi1_qp *qp);
+int hfi1_make_rc_req(struct rvt_qp *qp);
-int hfi1_make_uc_req(struct hfi1_qp *qp);
+int hfi1_make_uc_req(struct rvt_qp *qp);
-int hfi1_make_ud_req(struct hfi1_qp *qp);
+int hfi1_make_ud_req(struct rvt_qp *qp);
int hfi1_register_ib_device(struct hfi1_devdata *);
@@ -1059,13 +781,13 @@ void hfi1_ib_rcv(struct hfi1_packet *packet);
unsigned hfi1_get_npkeys(struct hfi1_devdata *);
-int hfi1_verbs_send_dma(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
+int hfi1_verbs_send_dma(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
u64 pbc);
-int hfi1_verbs_send_pio(struct hfi1_qp *qp, struct hfi1_pkt_state *ps,
+int hfi1_verbs_send_pio(struct rvt_qp *qp, struct hfi1_pkt_state *ps,
u64 pbc);
-struct send_context *qp_to_send_context(struct hfi1_qp *qp, u8 sc5);
+struct send_context *qp_to_send_context(struct rvt_qp *qp, u8 sc5);
extern const enum ib_wc_opcode ib_hfi1_wc_opcode[];
@@ -1075,8 +797,6 @@ extern const int ib_hfi1_state_ops[];
extern __be64 ib_hfi1_sys_image_guid; /* in network order */
-extern unsigned int hfi1_lkey_table_size;
-
extern unsigned int hfi1_max_cqes;
extern unsigned int hfi1_max_cqs;
diff --git a/drivers/staging/rdma/hfi1/verbs_mcast.c b/drivers/staging/rdma/hfi1/verbs_mcast.c
index afc6b4c..49954b9 100644
--- a/drivers/staging/rdma/hfi1/verbs_mcast.c
+++ b/drivers/staging/rdma/hfi1/verbs_mcast.c
@@ -56,7 +56,7 @@
* mcast_qp_alloc - alloc a struct to link a QP to mcast GID struct
* @qp: the QP to link
*/
-static struct hfi1_mcast_qp *mcast_qp_alloc(struct hfi1_qp *qp)
+static struct hfi1_mcast_qp *mcast_qp_alloc(struct rvt_qp *qp)
{
struct hfi1_mcast_qp *mqp;
@@ -73,7 +73,7 @@ bail:
static void mcast_qp_free(struct hfi1_mcast_qp *mqp)
{
- struct hfi1_qp *qp = mqp->qp;
+ struct rvt_qp *qp = mqp->qp;
/* Notify hfi1_destroy_qp() if it is waiting. */
if (atomic_dec_and_test(&qp->refcount))
@@ -241,7 +241,7 @@ bail:
int hfi1_multicast_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
{
- struct hfi1_qp *qp = to_iqp(ibqp);
+ struct rvt_qp *qp = to_iqp(ibqp);
struct hfi1_ibdev *dev = to_idev(ibqp->device);
struct hfi1_ibport *ibp;
struct hfi1_mcast *mcast;
@@ -299,7 +299,7 @@ bail:
int hfi1_multicast_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
{
- struct hfi1_qp *qp = to_iqp(ibqp);
+ struct rvt_qp *qp = to_iqp(ibqp);
struct hfi1_ibdev *dev = to_idev(ibqp->device);
struct hfi1_ibport *ibp = to_iport(ibqp->device, qp->port_num);
struct hfi1_mcast *mcast = NULL;
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [RFC PATCH 12/15] IB/hfi1: Remove srq from hfi1
[not found] ` <20151214172509.11576.19722.stgit-K+u1se/DcYrLESAwzcoQNrvm/XP+8Wra@public.gmane.org>
` (8 preceding siblings ...)
2015-12-14 17:29 ` [RFC PATCH 11/15] IB/hfi1: Remove hfi1 MR and hfi1 specific qp type Dennis Dalessandro
@ 2015-12-14 17:29 ` Dennis Dalessandro
2015-12-14 17:29 ` [RFC PATCH 13/15] IB/hfi1: Remove ibport and use rdmavt version Dennis Dalessandro
` (2 subsequent siblings)
12 siblings, 0 replies; 25+ messages in thread
From: Dennis Dalessandro @ 2015-12-14 17:29 UTC (permalink / raw)
To: devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-rdma-u79uwXL29TY76Z2rM5mHXA
Cc: gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r, dledford-H+wXaHxf7aLQT0dZR+AlfA
SRQ data structure has been moved to rdmavt. Make use of it.
Reviewed-by: Harish Chegondi <harish.chegondi-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
drivers/staging/rdma/hfi1/qp.c | 2 +-
drivers/staging/rdma/hfi1/ruc.c | 4 ++--
drivers/staging/rdma/hfi1/srq.c | 10 +++++-----
drivers/staging/rdma/hfi1/verbs.h | 13 -------------
4 files changed, 8 insertions(+), 21 deletions(-)
diff --git a/drivers/staging/rdma/hfi1/qp.c b/drivers/staging/rdma/hfi1/qp.c
index b82855f..045ee74 100644
--- a/drivers/staging/rdma/hfi1/qp.c
+++ b/drivers/staging/rdma/hfi1/qp.c
@@ -1092,7 +1092,7 @@ struct ib_qp *hfi1_create_qp(struct ib_pd *ibpd,
sz = sizeof(*qp);
sg_list_sz = 0;
if (init_attr->srq) {
- struct hfi1_srq *srq = to_isrq(init_attr->srq);
+ struct rvt_srq *srq = ibsrq_to_rvtsrq(init_attr->srq);
if (srq->rq.max_sge > 1)
sg_list_sz = sizeof(*qp->r_sg_list) *
diff --git a/drivers/staging/rdma/hfi1/ruc.c b/drivers/staging/rdma/hfi1/ruc.c
index 558dadb..9841e89 100644
--- a/drivers/staging/rdma/hfi1/ruc.c
+++ b/drivers/staging/rdma/hfi1/ruc.c
@@ -159,14 +159,14 @@ int hfi1_get_rwqe(struct rvt_qp *qp, int wr_id_only)
unsigned long flags;
struct rvt_rq *rq;
struct rvt_rwq *wq;
- struct hfi1_srq *srq;
+ struct rvt_srq *srq;
struct rvt_rwqe *wqe;
void (*handler)(struct ib_event *, void *);
u32 tail;
int ret;
if (qp->ibqp.srq) {
- srq = to_isrq(qp->ibqp.srq);
+ srq = ibsrq_to_rvtsrq(qp->ibqp.srq);
handler = srq->ibsrq.event_handler;
rq = &srq->rq;
} else {
diff --git a/drivers/staging/rdma/hfi1/srq.c b/drivers/staging/rdma/hfi1/srq.c
index 932bd96..78f190a 100644
--- a/drivers/staging/rdma/hfi1/srq.c
+++ b/drivers/staging/rdma/hfi1/srq.c
@@ -65,7 +65,7 @@
int hfi1_post_srq_receive(struct ib_srq *ibsrq, struct ib_recv_wr *wr,
struct ib_recv_wr **bad_wr)
{
- struct hfi1_srq *srq = to_isrq(ibsrq);
+ struct rvt_srq *srq = ibsrq_to_rvtsrq(ibsrq);
struct rvt_rwq *wq;
unsigned long flags;
int ret;
@@ -120,7 +120,7 @@ struct ib_srq *hfi1_create_srq(struct ib_pd *ibpd,
struct ib_udata *udata)
{
struct hfi1_ibdev *dev = to_idev(ibpd->device);
- struct hfi1_srq *srq;
+ struct rvt_srq *srq;
u32 sz;
struct ib_srq *ret;
@@ -229,7 +229,7 @@ int hfi1_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
enum ib_srq_attr_mask attr_mask,
struct ib_udata *udata)
{
- struct hfi1_srq *srq = to_isrq(ibsrq);
+ struct rvt_srq *srq = ibsrq_to_rvtsrq(ibsrq);
struct rvt_rwq *wq;
int ret = 0;
@@ -367,7 +367,7 @@ bail:
int hfi1_query_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr)
{
- struct hfi1_srq *srq = to_isrq(ibsrq);
+ struct rvt_srq *srq = ibsrq_to_rvtsrq(ibsrq);
attr->max_wr = srq->rq.size - 1;
attr->max_sge = srq->rq.max_sge;
@@ -381,7 +381,7 @@ int hfi1_query_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr)
*/
int hfi1_destroy_srq(struct ib_srq *ibsrq)
{
- struct hfi1_srq *srq = to_isrq(ibsrq);
+ struct rvt_srq *srq = ibsrq_to_rvtsrq(ibsrq);
struct hfi1_ibdev *dev = to_idev(ibsrq->device);
spin_lock(&dev->n_srqs_lock);
diff --git a/drivers/staging/rdma/hfi1/verbs.h b/drivers/staging/rdma/hfi1/verbs.h
index fec5e7b..f4ec83c 100644
--- a/drivers/staging/rdma/hfi1/verbs.h
+++ b/drivers/staging/rdma/hfi1/verbs.h
@@ -263,14 +263,6 @@ struct hfi1_cq {
struct rvt_mmap_info *ip;
};
-struct hfi1_srq {
- struct ib_srq ibsrq;
- struct rvt_rq rq;
- struct rvt_mmap_info *ip;
- /* send signal when number of RWQEs < limit */
- u32 limit;
-};
-
/*
* hfi1 specific data structures that will be hidden from rvt after the queue
* pair is made common
@@ -537,11 +529,6 @@ static inline struct hfi1_cq *to_icq(struct ib_cq *ibcq)
return container_of(ibcq, struct hfi1_cq, ibcq);
}
-static inline struct hfi1_srq *to_isrq(struct ib_srq *ibsrq)
-{
- return container_of(ibsrq, struct hfi1_srq, ibsrq);
-}
-
static inline struct rvt_qp *to_iqp(struct ib_qp *ibqp)
{
return container_of(ibqp, struct rvt_qp, ibqp);
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [RFC PATCH 13/15] IB/hfi1: Remove ibport and use rdmavt version
[not found] ` <20151214172509.11576.19722.stgit-K+u1se/DcYrLESAwzcoQNrvm/XP+8Wra@public.gmane.org>
` (9 preceding siblings ...)
2015-12-14 17:29 ` [RFC PATCH 12/15] IB/hfi1: Remove srq from hfi1 Dennis Dalessandro
@ 2015-12-14 17:29 ` Dennis Dalessandro
2015-12-14 17:29 ` [RFC PATCH 14/15] IB/hfi1: Remove mmap from hfi1 Dennis Dalessandro
2015-12-21 6:12 ` [RFC PATCH 00/15] staging/rdma/hfi1: Initial patches to add rdmavt support in HFI1 ira.weiny
12 siblings, 0 replies; 25+ messages in thread
From: Dennis Dalessandro @ 2015-12-14 17:29 UTC (permalink / raw)
To: devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-rdma-u79uwXL29TY76Z2rM5mHXA
Cc: gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r, dledford-H+wXaHxf7aLQT0dZR+AlfA
Remove most of the ibport members from hfi1 and use the rdmavt version.
Also register the port with rdmavt.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Reviewed-by: Harish Chegondi <harish.chegondi-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Jubin John <jubin.john-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
drivers/staging/rdma/hfi1/chip.c | 34 +++----
drivers/staging/rdma/hfi1/driver.c | 2
drivers/staging/rdma/hfi1/hfi.h | 8 +-
drivers/staging/rdma/hfi1/mad.c | 152 ++++++++++++++++---------------
drivers/staging/rdma/hfi1/qp.c | 23 ++---
drivers/staging/rdma/hfi1/qp.h | 2
drivers/staging/rdma/hfi1/rc.c | 32 +++----
drivers/staging/rdma/hfi1/ruc.c | 14 ++-
drivers/staging/rdma/hfi1/uc.c | 2
drivers/staging/rdma/hfi1/ud.c | 16 ++-
drivers/staging/rdma/hfi1/verbs.c | 61 +++++++-----
drivers/staging/rdma/hfi1/verbs.h | 51 +---------
drivers/staging/rdma/hfi1/verbs_mcast.c | 28 +++---
13 files changed, 198 insertions(+), 227 deletions(-)
diff --git a/drivers/staging/rdma/hfi1/chip.c b/drivers/staging/rdma/hfi1/chip.c
index f799b86..6d916d0 100644
--- a/drivers/staging/rdma/hfi1/chip.c
+++ b/drivers/staging/rdma/hfi1/chip.c
@@ -1543,8 +1543,8 @@ static u64 access_sw_cpu_##cntr(const struct cntr_entry *entry, \
void *context, int vl, int mode, u64 data) \
{ \
struct hfi1_pportdata *ppd = (struct hfi1_pportdata *)context; \
- return read_write_cpu(ppd->dd, &ppd->ibport_data.z_ ##cntr, \
- ppd->ibport_data.cntr, vl, \
+ return read_write_cpu(ppd->dd, &ppd->ibport_data.rvp.z_ ##cntr, \
+ ppd->ibport_data.rvp.cntr, vl, \
mode, data); \
}
@@ -1561,7 +1561,7 @@ static u64 access_ibp_##cntr(const struct cntr_entry *entry, \
if (vl != CNTR_INVALID_VL) \
return 0; \
\
- return read_write_sw(ppd->dd, &ppd->ibport_data.n_ ##cntr, \
+ return read_write_sw(ppd->dd, &ppd->ibport_data.rvp.n_ ##cntr, \
mode, data); \
}
@@ -5947,14 +5947,14 @@ static inline int init_cpu_counters(struct hfi1_devdata *dd)
ppd = (struct hfi1_pportdata *)(dd + 1);
for (i = 0; i < dd->num_pports; i++, ppd++) {
- ppd->ibport_data.rc_acks = NULL;
- ppd->ibport_data.rc_qacks = NULL;
- ppd->ibport_data.rc_acks = alloc_percpu(u64);
- ppd->ibport_data.rc_qacks = alloc_percpu(u64);
- ppd->ibport_data.rc_delayed_comp = alloc_percpu(u64);
- if ((ppd->ibport_data.rc_acks == NULL) ||
- (ppd->ibport_data.rc_delayed_comp == NULL) ||
- (ppd->ibport_data.rc_qacks == NULL))
+ ppd->ibport_data.rvp.rc_acks = NULL;
+ ppd->ibport_data.rvp.rc_qacks = NULL;
+ ppd->ibport_data.rvp.rc_acks = alloc_percpu(u64);
+ ppd->ibport_data.rvp.rc_qacks = alloc_percpu(u64);
+ ppd->ibport_data.rvp.rc_delayed_comp = alloc_percpu(u64);
+ if (!ppd->ibport_data.rvp.rc_acks ||
+ !ppd->ibport_data.rvp.rc_delayed_comp ||
+ !ppd->ibport_data.rvp.rc_qacks)
return -ENOMEM;
}
@@ -8010,14 +8010,14 @@ static void free_cntrs(struct hfi1_devdata *dd)
for (i = 0; i < dd->num_pports; i++, ppd++) {
kfree(ppd->cntrs);
kfree(ppd->scntrs);
- free_percpu(ppd->ibport_data.rc_acks);
- free_percpu(ppd->ibport_data.rc_qacks);
- free_percpu(ppd->ibport_data.rc_delayed_comp);
+ free_percpu(ppd->ibport_data.rvp.rc_acks);
+ free_percpu(ppd->ibport_data.rvp.rc_qacks);
+ free_percpu(ppd->ibport_data.rvp.rc_delayed_comp);
ppd->cntrs = NULL;
ppd->scntrs = NULL;
- ppd->ibport_data.rc_acks = NULL;
- ppd->ibport_data.rc_qacks = NULL;
- ppd->ibport_data.rc_delayed_comp = NULL;
+ ppd->ibport_data.rvp.rc_acks = NULL;
+ ppd->ibport_data.rvp.rc_qacks = NULL;
+ ppd->ibport_data.rvp.rc_delayed_comp = NULL;
}
kfree(dd->portcntrnames);
dd->portcntrnames = NULL;
diff --git a/drivers/staging/rdma/hfi1/driver.c b/drivers/staging/rdma/hfi1/driver.c
index 182e05f..3b17913 100644
--- a/drivers/staging/rdma/hfi1/driver.c
+++ b/drivers/staging/rdma/hfi1/driver.c
@@ -337,7 +337,7 @@ static void rcv_hdrerr(struct hfi1_ctxtdata *rcd, struct hfi1_pportdata *ppd,
/* Check for valid receive state. */
if (!(ib_hfi1_state_ops[qp->state] &
HFI1_PROCESS_RECV_OK)) {
- ibp->n_pkt_drops++;
+ ibp->rvp.n_pkt_drops++;
}
switch (qp->ibqp.qp_type) {
diff --git a/drivers/staging/rdma/hfi1/hfi.h b/drivers/staging/rdma/hfi1/hfi.h
index 414d920..a893789 100644
--- a/drivers/staging/rdma/hfi1/hfi.h
+++ b/drivers/staging/rdma/hfi1/hfi.h
@@ -1795,10 +1795,10 @@ static inline void hfi1_reset_cpu_counters(struct hfi1_devdata *dd)
ppd = (struct hfi1_pportdata *)(dd + 1);
for (i = 0; i < dd->num_pports; i++, ppd++) {
- ppd->ibport_data.z_rc_acks =
- get_all_cpu_total(ppd->ibport_data.rc_acks);
- ppd->ibport_data.z_rc_qacks =
- get_all_cpu_total(ppd->ibport_data.rc_qacks);
+ ppd->ibport_data.rvp.z_rc_acks =
+ get_all_cpu_total(ppd->ibport_data.rvp.rc_acks);
+ ppd->ibport_data.rvp.z_rc_qacks =
+ get_all_cpu_total(ppd->ibport_data.rvp.rc_qacks);
}
}
diff --git a/drivers/staging/rdma/hfi1/mad.c b/drivers/staging/rdma/hfi1/mad.c
index 8e9d1e7..7172d42 100644
--- a/drivers/staging/rdma/hfi1/mad.c
+++ b/drivers/staging/rdma/hfi1/mad.c
@@ -91,7 +91,7 @@ static void send_trap(struct hfi1_ibport *ibp, void *data, unsigned len)
int pkey_idx;
u32 qpn = ppd_from_ibp(ibp)->sm_trap_qp;
- agent = ibp->send_agent;
+ agent = ibp->rvp.send_agent;
if (!agent)
return;
@@ -100,7 +100,8 @@ static void send_trap(struct hfi1_ibport *ibp, void *data, unsigned len)
return;
/* o14-2 */
- if (ibp->trap_timeout && time_before(jiffies, ibp->trap_timeout))
+ if (ibp->rvp.trap_timeout && time_before(jiffies,
+ ibp->rvp.trap_timeout))
return;
pkey_idx = hfi1_lookup_pkey_idx(ibp, LIM_MGMT_P_KEY);
@@ -121,18 +122,18 @@ static void send_trap(struct hfi1_ibport *ibp, void *data, unsigned len)
smp->mgmt_class = IB_MGMT_CLASS_SUBN_LID_ROUTED;
smp->class_version = 1;
smp->method = IB_MGMT_METHOD_TRAP;
- ibp->tid++;
- smp->tid = cpu_to_be64(ibp->tid);
+ ibp->rvp.tid++;
+ smp->tid = cpu_to_be64(ibp->rvp.tid);
smp->attr_id = IB_SMP_ATTR_NOTICE;
/* o14-1: smp->mkey = 0; */
memcpy(smp->data, data, len);
- spin_lock_irqsave(&ibp->lock, flags);
+ spin_lock_irqsave(&ibp->rvp.lock, flags);
if (!ibp->sm_ah) {
- if (ibp->sm_lid != be16_to_cpu(IB_LID_PERMISSIVE)) {
+ if (ibp->rvp.sm_lid != be16_to_cpu(IB_LID_PERMISSIVE)) {
struct ib_ah *ah;
- ah = hfi1_create_qp0_ah(ibp, ibp->sm_lid);
+ ah = hfi1_create_qp0_ah(ibp, ibp->rvp.sm_lid);
if (IS_ERR(ah))
ret = PTR_ERR(ah);
else {
@@ -146,17 +147,17 @@ static void send_trap(struct hfi1_ibport *ibp, void *data, unsigned len)
send_buf->ah = &ibp->sm_ah->ibah;
ret = 0;
}
- spin_unlock_irqrestore(&ibp->lock, flags);
+ spin_unlock_irqrestore(&ibp->rvp.lock, flags);
if (!ret)
ret = ib_post_send_mad(send_buf, NULL);
if (!ret) {
/* 4.096 usec. */
- timeout = (4096 * (1UL << ibp->subnet_timeout)) / 1000;
- ibp->trap_timeout = jiffies + usecs_to_jiffies(timeout);
+ timeout = (4096 * (1UL << ibp->rvp.subnet_timeout)) / 1000;
+ ibp->rvp.trap_timeout = jiffies + usecs_to_jiffies(timeout);
} else {
ib_free_send_mad(send_buf);
- ibp->trap_timeout = 0;
+ ibp->rvp.trap_timeout = 0;
}
}
@@ -169,10 +170,10 @@ void hfi1_bad_pqkey(struct hfi1_ibport *ibp, __be16 trap_num, u32 key, u32 sl,
struct ib_mad_notice_attr data;
if (trap_num == IB_NOTICE_TRAP_BAD_PKEY)
- ibp->pkey_violations++;
+ ibp->rvp.pkey_violations++;
else
- ibp->qkey_violations++;
- ibp->n_pkt_drops++;
+ ibp->rvp.qkey_violations++;
+ ibp->rvp.n_pkt_drops++;
/* Send violation trap */
data.generic_type = IB_NOTICE_TYPE_SECURITY;
@@ -244,7 +245,8 @@ void hfi1_cap_mask_chg(struct hfi1_ibport *ibp)
data.toggle_count = 0;
memset(&data.details, 0, sizeof(data.details));
data.details.ntc_144.lid = data.issuer_lid;
- data.details.ntc_144.new_cap_mask = cpu_to_be32(ibp->port_cap_flags);
+ data.details.ntc_144.new_cap_mask =
+ cpu_to_be32(ibp->rvp.port_cap_flags);
send_trap(ibp, &data, sizeof(data));
}
@@ -406,37 +408,38 @@ static int check_mkey(struct hfi1_ibport *ibp, struct ib_mad_hdr *mad,
int ret = 0;
/* Is the mkey in the process of expiring? */
- if (ibp->mkey_lease_timeout &&
- time_after_eq(jiffies, ibp->mkey_lease_timeout)) {
+ if (ibp->rvp.mkey_lease_timeout &&
+ time_after_eq(jiffies, ibp->rvp.mkey_lease_timeout)) {
/* Clear timeout and mkey protection field. */
- ibp->mkey_lease_timeout = 0;
- ibp->mkeyprot = 0;
+ ibp->rvp.mkey_lease_timeout = 0;
+ ibp->rvp.mkeyprot = 0;
}
- if ((mad_flags & IB_MAD_IGNORE_MKEY) || ibp->mkey == 0 ||
- ibp->mkey == mkey)
+ if ((mad_flags & IB_MAD_IGNORE_MKEY) || ibp->rvp.mkey == 0 ||
+ ibp->rvp.mkey == mkey)
valid_mkey = 1;
/* Unset lease timeout on any valid Get/Set/TrapRepress */
- if (valid_mkey && ibp->mkey_lease_timeout &&
+ if (valid_mkey && ibp->rvp.mkey_lease_timeout &&
(mad->method == IB_MGMT_METHOD_GET ||
mad->method == IB_MGMT_METHOD_SET ||
mad->method == IB_MGMT_METHOD_TRAP_REPRESS))
- ibp->mkey_lease_timeout = 0;
+ ibp->rvp.mkey_lease_timeout = 0;
if (!valid_mkey) {
switch (mad->method) {
case IB_MGMT_METHOD_GET:
/* Bad mkey not a violation below level 2 */
- if (ibp->mkeyprot < 2)
+ if (ibp->rvp.mkeyprot < 2)
break;
case IB_MGMT_METHOD_SET:
case IB_MGMT_METHOD_TRAP_REPRESS:
- if (ibp->mkey_violations != 0xFFFF)
- ++ibp->mkey_violations;
- if (!ibp->mkey_lease_timeout && ibp->mkey_lease_period)
- ibp->mkey_lease_timeout = jiffies +
- ibp->mkey_lease_period * HZ;
+ if (ibp->rvp.mkey_violations != 0xFFFF)
+ ++ibp->rvp.mkey_violations;
+ if (!ibp->rvp.mkey_lease_timeout &&
+ ibp->rvp.mkey_lease_period)
+ ibp->rvp.mkey_lease_timeout = jiffies +
+ ibp->rvp.mkey_lease_period * HZ;
/* Generate a trap notice. */
bad_mkey(ibp, mad, mkey, dr_slid, return_path,
hop_cnt);
@@ -547,14 +550,14 @@ static int __subn_get_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
/* Only return the mkey if the protection field allows it. */
if (!(smp->method == IB_MGMT_METHOD_GET &&
- ibp->mkey != smp->mkey &&
- ibp->mkeyprot == 1))
- pi->mkey = ibp->mkey;
-
- pi->subnet_prefix = ibp->gid_prefix;
- pi->sm_lid = cpu_to_be32(ibp->sm_lid);
- pi->ib_cap_mask = cpu_to_be32(ibp->port_cap_flags);
- pi->mkey_lease_period = cpu_to_be16(ibp->mkey_lease_period);
+ ibp->rvp.mkey != smp->mkey &&
+ ibp->rvp.mkeyprot == 1))
+ pi->mkey = ibp->rvp.mkey;
+
+ pi->subnet_prefix = ibp->rvp.gid_prefix;
+ pi->sm_lid = cpu_to_be32(ibp->rvp.sm_lid);
+ pi->ib_cap_mask = cpu_to_be32(ibp->rvp.port_cap_flags);
+ pi->mkey_lease_period = cpu_to_be16(ibp->rvp.mkey_lease_period);
pi->sm_trap_qp = cpu_to_be32(ppd->sm_trap_qp);
pi->sa_qp = cpu_to_be32(ppd->sa_qp);
@@ -598,7 +601,7 @@ static int __subn_get_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
pi->port_states.portphysstate_portstate =
(hfi1_ibphys_portstate(ppd) << 4) | state;
- pi->mkeyprotect_lmc = (ibp->mkeyprot << 6) | ppd->lmc;
+ pi->mkeyprotect_lmc = (ibp->rvp.mkeyprot << 6) | ppd->lmc;
memset(pi->neigh_mtu.pvlx_to_mtu, 0, sizeof(pi->neigh_mtu.pvlx_to_mtu));
for (i = 0; i < ppd->vls_supported; i++) {
@@ -611,7 +614,7 @@ static int __subn_get_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
/* don't forget VL 15 */
mtu = mtu_to_enum(dd->vld[15].mtu, 2048);
pi->neigh_mtu.pvlx_to_mtu[15/2] |= mtu;
- pi->smsl = ibp->sm_sl & OPA_PI_MASK_SMSL;
+ pi->smsl = ibp->rvp.sm_sl & OPA_PI_MASK_SMSL;
pi->operational_vls = hfi1_get_ib_cfg(ppd, HFI1_IB_CFG_OP_VLS);
pi->partenforce_filterraw |=
(ppd->linkinit_reason & OPA_PI_MASK_LINKINIT_REASON);
@@ -619,17 +622,17 @@ static int __subn_get_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
pi->partenforce_filterraw |= OPA_PI_MASK_PARTITION_ENFORCE_IN;
if (ppd->part_enforce & HFI1_PART_ENFORCE_OUT)
pi->partenforce_filterraw |= OPA_PI_MASK_PARTITION_ENFORCE_OUT;
- pi->mkey_violations = cpu_to_be16(ibp->mkey_violations);
+ pi->mkey_violations = cpu_to_be16(ibp->rvp.mkey_violations);
/* P_KeyViolations are counted by hardware. */
- pi->pkey_violations = cpu_to_be16(ibp->pkey_violations);
- pi->qkey_violations = cpu_to_be16(ibp->qkey_violations);
+ pi->pkey_violations = cpu_to_be16(ibp->rvp.pkey_violations);
+ pi->qkey_violations = cpu_to_be16(ibp->rvp.qkey_violations);
pi->vl.cap = ppd->vls_supported;
- pi->vl.high_limit = cpu_to_be16(ibp->vl_high_limit);
+ pi->vl.high_limit = cpu_to_be16(ibp->rvp.vl_high_limit);
pi->vl.arb_high_cap = (u8)hfi1_get_ib_cfg(ppd, HFI1_IB_CFG_VL_HIGH_CAP);
pi->vl.arb_low_cap = (u8)hfi1_get_ib_cfg(ppd, HFI1_IB_CFG_VL_LOW_CAP);
- pi->clientrereg_subnettimeout = ibp->subnet_timeout;
+ pi->clientrereg_subnettimeout = ibp->rvp.subnet_timeout;
pi->port_link_mode = cpu_to_be16(OPA_PORT_LINK_MODE_OPA << 10 |
OPA_PORT_LINK_MODE_OPA << 5 |
@@ -1090,9 +1093,9 @@ static int __subn_set_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
ls_old = driver_lstate(ppd);
- ibp->mkey = pi->mkey;
- ibp->gid_prefix = pi->subnet_prefix;
- ibp->mkey_lease_period = be16_to_cpu(pi->mkey_lease_period);
+ ibp->rvp.mkey = pi->mkey;
+ ibp->rvp.gid_prefix = pi->subnet_prefix;
+ ibp->rvp.mkey_lease_period = be16_to_cpu(pi->mkey_lease_period);
/* Must be a valid unicast LID address. */
if ((lid == 0 && ls_old > IB_PORT_INIT) ||
@@ -1132,20 +1135,20 @@ static int __subn_set_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
smlid >= be16_to_cpu(IB_MULTICAST_LID_BASE)) {
smp->status |= IB_SMP_INVALID_FIELD;
pr_warn("SubnSet(OPA_PortInfo) smlid invalid 0x%x\n", smlid);
- } else if (smlid != ibp->sm_lid || msl != ibp->sm_sl) {
+ } else if (smlid != ibp->rvp.sm_lid || msl != ibp->rvp.sm_sl) {
pr_warn("SubnSet(OPA_PortInfo) smlid 0x%x\n", smlid);
- spin_lock_irqsave(&ibp->lock, flags);
+ spin_lock_irqsave(&ibp->rvp.lock, flags);
if (ibp->sm_ah) {
- if (smlid != ibp->sm_lid)
+ if (smlid != ibp->rvp.sm_lid)
ibp->sm_ah->attr.dlid = smlid;
- if (msl != ibp->sm_sl)
+ if (msl != ibp->rvp.sm_sl)
ibp->sm_ah->attr.sl = msl;
}
- spin_unlock_irqrestore(&ibp->lock, flags);
- if (smlid != ibp->sm_lid)
- ibp->sm_lid = smlid;
- if (msl != ibp->sm_sl)
- ibp->sm_sl = msl;
+ spin_unlock_irqrestore(&ibp->rvp.lock, flags);
+ if (smlid != ibp->rvp.sm_lid)
+ ibp->rvp.sm_lid = smlid;
+ if (msl != ibp->rvp.sm_sl)
+ ibp->rvp.sm_sl = msl;
event.event = IB_EVENT_SM_CHANGE;
ib_dispatch_event(&event);
}
@@ -1197,10 +1200,11 @@ static int __subn_set_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
smp->status |= IB_SMP_INVALID_FIELD;
}
- ibp->mkeyprot = (pi->mkeyprotect_lmc & OPA_PI_MASK_MKEY_PROT_BIT) >> 6;
- ibp->vl_high_limit = be16_to_cpu(pi->vl.high_limit) & 0xFF;
+ ibp->rvp.mkeyprot =
+ (pi->mkeyprotect_lmc & OPA_PI_MASK_MKEY_PROT_BIT) >> 6;
+ ibp->rvp.vl_high_limit = be16_to_cpu(pi->vl.high_limit) & 0xFF;
(void)hfi1_set_ib_cfg(ppd, HFI1_IB_CFG_VL_HIGH_LIMIT,
- ibp->vl_high_limit);
+ ibp->rvp.vl_high_limit);
if (ppd->vls_supported/2 > ARRAY_SIZE(pi->neigh_mtu.pvlx_to_mtu) ||
ppd->vls_supported > ARRAY_SIZE(dd->vld)) {
@@ -1259,15 +1263,15 @@ static int __subn_set_opa_portinfo(struct opa_smp *smp, u32 am, u8 *data,
}
if (pi->mkey_violations == 0)
- ibp->mkey_violations = 0;
+ ibp->rvp.mkey_violations = 0;
if (pi->pkey_violations == 0)
- ibp->pkey_violations = 0;
+ ibp->rvp.pkey_violations = 0;
if (pi->qkey_violations == 0)
- ibp->qkey_violations = 0;
+ ibp->rvp.qkey_violations = 0;
- ibp->subnet_timeout =
+ ibp->rvp.subnet_timeout =
pi->clientrereg_subnettimeout & OPA_PI_MASK_SUBNET_TIMEOUT;
crc_enabled = be16_to_cpu(pi->port_ltp_crc_mode);
@@ -3570,9 +3574,9 @@ static int subn_get_opa_sma(__be16 attr_id, struct opa_smp *smp, u32 am,
resp_len);
break;
case IB_SMP_ATTR_SM_INFO:
- if (ibp->port_cap_flags & IB_PORT_SM_DISABLED)
+ if (ibp->rvp.port_cap_flags & IB_PORT_SM_DISABLED)
return IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_CONSUMED;
- if (ibp->port_cap_flags & IB_PORT_SM)
+ if (ibp->rvp.port_cap_flags & IB_PORT_SM)
return IB_MAD_RESULT_SUCCESS;
/* FALLTHROUGH */
default:
@@ -3640,9 +3644,9 @@ static int subn_set_opa_sma(__be16 attr_id, struct opa_smp *smp, u32 am,
resp_len);
break;
case IB_SMP_ATTR_SM_INFO:
- if (ibp->port_cap_flags & IB_PORT_SM_DISABLED)
+ if (ibp->rvp.port_cap_flags & IB_PORT_SM_DISABLED)
return IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_CONSUMED;
- if (ibp->port_cap_flags & IB_PORT_SM)
+ if (ibp->rvp.port_cap_flags & IB_PORT_SM)
return IB_MAD_RESULT_SUCCESS;
/* FALLTHROUGH */
default:
@@ -4218,7 +4222,7 @@ int hfi1_create_agents(struct hfi1_ibdev *dev)
goto err;
}
- ibp->send_agent = agent;
+ ibp->rvp.send_agent = agent;
}
return 0;
@@ -4226,9 +4230,9 @@ int hfi1_create_agents(struct hfi1_ibdev *dev)
err:
for (p = 0; p < dd->num_pports; p++) {
ibp = &dd->pport[p].ibport_data;
- if (ibp->send_agent) {
- agent = ibp->send_agent;
- ibp->send_agent = NULL;
+ if (ibp->rvp.send_agent) {
+ agent = ibp->rvp.send_agent;
+ ibp->rvp.send_agent = NULL;
ib_unregister_mad_agent(agent);
}
}
@@ -4245,9 +4249,9 @@ void hfi1_free_agents(struct hfi1_ibdev *dev)
for (p = 0; p < dd->num_pports; p++) {
ibp = &dd->pport[p].ibport_data;
- if (ibp->send_agent) {
- agent = ibp->send_agent;
- ibp->send_agent = NULL;
+ if (ibp->rvp.send_agent) {
+ agent = ibp->rvp.send_agent;
+ ibp->rvp.send_agent = NULL;
ib_unregister_mad_agent(agent);
}
if (ibp->sm_ah) {
diff --git a/drivers/staging/rdma/hfi1/qp.c b/drivers/staging/rdma/hfi1/qp.c
index 045ee74..18e7276 100644
--- a/drivers/staging/rdma/hfi1/qp.c
+++ b/drivers/staging/rdma/hfi1/qp.c
@@ -238,7 +238,7 @@ static void insert_qp(struct hfi1_ibdev *dev, struct rvt_qp *qp)
spin_lock_irqsave(&dev->qp_dev->qpt_lock, flags);
if (qp->ibqp.qp_num <= 1) {
- rcu_assign_pointer(ibp->qp[qp->ibqp.qp_num], qp);
+ rcu_assign_pointer(ibp->rvp.qp[qp->ibqp.qp_num], qp);
} else {
u32 n = qpn_hash(dev->qp_dev, qp->ibqp.qp_num);
@@ -263,12 +263,13 @@ static void remove_qp(struct hfi1_ibdev *dev, struct rvt_qp *qp)
spin_lock_irqsave(&dev->qp_dev->qpt_lock, flags);
- if (rcu_dereference_protected(ibp->qp[0],
+ if (rcu_dereference_protected(ibp->rvp.qp[0],
+ lockdep_is_held(
+ &dev->qp_dev->qpt_lock)) == qp) {
+ RCU_INIT_POINTER(ibp->rvp.qp[0], NULL);
+ } else if (rcu_dereference_protected(ibp->rvp.qp[1],
lockdep_is_held(&dev->qp_dev->qpt_lock)) == qp) {
- RCU_INIT_POINTER(ibp->qp[0], NULL);
- } else if (rcu_dereference_protected(ibp->qp[1],
- lockdep_is_held(&dev->qp_dev->qpt_lock)) == qp) {
- RCU_INIT_POINTER(ibp->qp[1], NULL);
+ RCU_INIT_POINTER(ibp->rvp.qp[1], NULL);
} else {
struct rvt_qp *q;
struct rvt_qp __rcu **qpp;
@@ -317,9 +318,9 @@ static unsigned free_all_qps(struct hfi1_devdata *dd)
if (!hfi1_mcast_tree_empty(ibp))
qp_inuse++;
rcu_read_lock();
- if (rcu_dereference(ibp->qp[0]))
+ if (rcu_dereference(ibp->rvp.qp[0]))
qp_inuse++;
- if (rcu_dereference(ibp->qp[1]))
+ if (rcu_dereference(ibp->rvp.qp[1]))
qp_inuse++;
rcu_read_unlock();
}
@@ -1467,7 +1468,7 @@ static int iowait_sleep(
struct hfi1_ibport *ibp =
to_iport(qp->ibqp.device, qp->port_num);
- ibp->n_dmawait++;
+ ibp->rvp.n_dmawait++;
qp->s_flags |= HFI1_S_WAIT_DMA_DESC;
list_add_tail(&priv->s_iowait.list, &sde->dmawait);
trace_hfi1_qpsleep(qp, HFI1_S_WAIT_DMA_DESC);
@@ -1636,9 +1637,9 @@ int qp_iter_next(struct qp_iter *iter)
ibp = &ppd->ibport_data;
if (!(n & 1))
- qp = rcu_dereference(ibp->qp[0]);
+ qp = rcu_dereference(ibp->rvp.qp[0]);
else
- qp = rcu_dereference(ibp->qp[1]);
+ qp = rcu_dereference(ibp->rvp.qp[1]);
} else {
qp = rcu_dereference(
dev->qp_dev->qp_table[
diff --git a/drivers/staging/rdma/hfi1/qp.h b/drivers/staging/rdma/hfi1/qp.h
index fff61f9..f735825 100644
--- a/drivers/staging/rdma/hfi1/qp.h
+++ b/drivers/staging/rdma/hfi1/qp.h
@@ -104,7 +104,7 @@ static inline struct rvt_qp *hfi1_lookup_qpn(struct hfi1_ibport *ibp,
struct rvt_qp *qp = NULL;
if (unlikely(qpn <= 1)) {
- qp = rcu_dereference(ibp->qp[qpn]);
+ qp = rcu_dereference(ibp->rvp.qp[qpn]);
} else {
struct hfi1_ibdev *dev = &ppd_from_ibp(ibp)->dd->verbs_dev;
u32 n = qpn_hash(dev->qp_dev, qpn);
diff --git a/drivers/staging/rdma/hfi1/rc.c b/drivers/staging/rdma/hfi1/rc.c
index 752643d..833f00f 100644
--- a/drivers/staging/rdma/hfi1/rc.c
+++ b/drivers/staging/rdma/hfi1/rc.c
@@ -772,7 +772,7 @@ void hfi1_send_rc_ack(struct hfi1_ctxtdata *rcd, struct rvt_qp *qp,
return;
queue_ack:
- this_cpu_inc(*ibp->rc_qacks);
+ this_cpu_inc(*ibp->rvp.rc_qacks);
spin_lock_irqsave(&qp->s_lock, flags);
qp->s_flags |= HFI1_S_ACK_PENDING | HFI1_S_RESP_PENDING;
qp->s_nak_state = qp->r_nak_state;
@@ -900,9 +900,9 @@ static void restart_rc(struct rvt_qp *qp, u32 psn, int wait)
ibp = to_iport(qp->ibqp.device, qp->port_num);
if (wqe->wr.opcode == IB_WR_RDMA_READ)
- ibp->n_rc_resends++;
+ ibp->rvp.n_rc_resends++;
else
- ibp->n_rc_resends += delta_psn(qp->s_psn, psn);
+ ibp->rvp.n_rc_resends += delta_psn(qp->s_psn, psn);
qp->s_flags &= ~(HFI1_S_WAIT_FENCE | HFI1_S_WAIT_RDMAR |
HFI1_S_WAIT_SSN_CREDIT | HFI1_S_WAIT_PSN |
@@ -925,7 +925,7 @@ static void rc_timeout(unsigned long arg)
spin_lock(&qp->s_lock);
if (qp->s_flags & HFI1_S_TIMER) {
ibp = to_iport(qp->ibqp.device, qp->port_num);
- ibp->n_rc_timeouts++;
+ ibp->rvp.n_rc_timeouts++;
qp->s_flags &= ~HFI1_S_TIMER;
del_timer(&qp->s_timer);
trace_hfi1_rc_timeout(qp, qp->s_last_psn + 1);
@@ -1104,7 +1104,7 @@ static struct rvt_swqe *do_rc_completion(struct rvt_qp *qp,
} else {
struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
- this_cpu_inc(*ibp->rc_delayed_comp);
+ this_cpu_inc(*ibp->rvp.rc_delayed_comp);
/*
* If send progress not running attempt to progress
* SDMA queue.
@@ -1263,7 +1263,7 @@ static int do_rc_ack(struct rvt_qp *qp, u32 aeth, u32 psn, int opcode,
switch (aeth >> 29) {
case 0: /* ACK */
- this_cpu_inc(*ibp->rc_acks);
+ this_cpu_inc(*ibp->rvp.rc_acks);
if (qp->s_acked != qp->s_tail) {
/*
* We are expecting more ACKs so
@@ -1292,7 +1292,7 @@ static int do_rc_ack(struct rvt_qp *qp, u32 aeth, u32 psn, int opcode,
goto bail;
case 1: /* RNR NAK */
- ibp->n_rnr_naks++;
+ ibp->rvp.n_rnr_naks++;
if (qp->s_acked == qp->s_tail)
goto bail;
if (qp->s_flags & HFI1_S_WAIT_RNR)
@@ -1307,7 +1307,7 @@ static int do_rc_ack(struct rvt_qp *qp, u32 aeth, u32 psn, int opcode,
/* The last valid PSN is the previous PSN. */
update_last_psn(qp, psn - 1);
- ibp->n_rc_resends += delta_psn(qp->s_psn, psn);
+ ibp->rvp.n_rc_resends += delta_psn(qp->s_psn, psn);
reset_psn(qp, psn);
@@ -1328,7 +1328,7 @@ static int do_rc_ack(struct rvt_qp *qp, u32 aeth, u32 psn, int opcode,
switch ((aeth >> HFI1_AETH_CREDIT_SHIFT) &
HFI1_AETH_CREDIT_MASK) {
case 0: /* PSN sequence error */
- ibp->n_seq_naks++;
+ ibp->rvp.n_seq_naks++;
/*
* Back up to the responder's expected PSN.
* Note that we might get a NAK in the middle of an
@@ -1341,17 +1341,17 @@ static int do_rc_ack(struct rvt_qp *qp, u32 aeth, u32 psn, int opcode,
case 1: /* Invalid Request */
status = IB_WC_REM_INV_REQ_ERR;
- ibp->n_other_naks++;
+ ibp->rvp.n_other_naks++;
goto class_b;
case 2: /* Remote Access Error */
status = IB_WC_REM_ACCESS_ERR;
- ibp->n_other_naks++;
+ ibp->rvp.n_other_naks++;
goto class_b;
case 3: /* Remote Operation Error */
status = IB_WC_REM_OP_ERR;
- ibp->n_other_naks++;
+ ibp->rvp.n_other_naks++;
class_b:
if (qp->s_last == qp->s_acked) {
hfi1_send_complete(qp, wqe, status);
@@ -1402,7 +1402,7 @@ static void rdma_seq_err(struct rvt_qp *qp, struct hfi1_ibport *ibp, u32 psn,
wqe = do_rc_completion(qp, wqe, ibp);
}
- ibp->n_rdma_seq++;
+ ibp->rvp.n_rdma_seq++;
qp->r_flags |= HFI1_R_RDMAR_SEQ;
restart_rc(qp, qp->s_last_psn + 1, 0);
if (list_empty(&qp->rspwait)) {
@@ -1663,7 +1663,7 @@ static noinline int rc_rcv_error(struct hfi1_other_headers *ohdr, void *data,
* Don't queue the NAK if we already sent one.
*/
if (!qp->r_nak_state) {
- ibp->n_rc_seqnak++;
+ ibp->rvp.n_rc_seqnak++;
qp->r_nak_state = IB_NAK_PSN_ERROR;
/* Use the expected PSN. */
qp->r_ack_psn = qp->r_psn;
@@ -1695,7 +1695,7 @@ static noinline int rc_rcv_error(struct hfi1_other_headers *ohdr, void *data,
*/
e = NULL;
old_req = 1;
- ibp->n_rc_dupreq++;
+ ibp->rvp.n_rc_dupreq++;
spin_lock_irqsave(&qp->s_lock, flags);
@@ -2429,7 +2429,7 @@ void hfi1_rc_hdrerr(
if (opcode < IB_OPCODE_RC_RDMA_READ_RESPONSE_FIRST) {
diff = delta_psn(psn, qp->r_psn);
if (!qp->r_nak_state && diff >= 0) {
- ibp->n_rc_seqnak++;
+ ibp->rvp.n_rc_seqnak++;
qp->r_nak_state = IB_NAK_PSN_ERROR;
/* Use the expected PSN. */
qp->r_ack_psn = qp->r_psn;
diff --git a/drivers/staging/rdma/hfi1/ruc.c b/drivers/staging/rdma/hfi1/ruc.c
index 9841e89..adc28b6 100644
--- a/drivers/staging/rdma/hfi1/ruc.c
+++ b/drivers/staging/rdma/hfi1/ruc.c
@@ -279,7 +279,8 @@ int hfi1_ruc_check_hdr(struct hfi1_ibport *ibp, struct hfi1_ib_header *hdr,
if (!(qp->alt_ah_attr.ah_flags & IB_AH_GRH))
goto err;
guid = get_sguid(ibp, qp->alt_ah_attr.grh.sgid_index);
- if (!gid_ok(&hdr->u.l.grh.dgid, ibp->gid_prefix, guid))
+ if (!gid_ok(&hdr->u.l.grh.dgid, ibp->rvp.gid_prefix,
+ guid))
goto err;
if (!gid_ok(&hdr->u.l.grh.sgid,
qp->alt_ah_attr.grh.dgid.global.subnet_prefix,
@@ -311,7 +312,8 @@ int hfi1_ruc_check_hdr(struct hfi1_ibport *ibp, struct hfi1_ib_header *hdr,
goto err;
guid = get_sguid(ibp,
qp->remote_ah_attr.grh.sgid_index);
- if (!gid_ok(&hdr->u.l.grh.dgid, ibp->gid_prefix, guid))
+ if (!gid_ok(&hdr->u.l.grh.dgid, ibp->rvp.gid_prefix,
+ guid))
goto err;
if (!gid_ok(&hdr->u.l.grh.sgid,
qp->remote_ah_attr.grh.dgid.global.subnet_prefix,
@@ -411,7 +413,7 @@ again:
if (!qp || !(ib_hfi1_state_ops[qp->state] & HFI1_PROCESS_RECV_OK) ||
qp->ibqp.qp_type != sqp->ibqp.qp_type) {
- ibp->n_pkt_drops++;
+ ibp->rvp.n_pkt_drops++;
/*
* For RC, the requester would timeout and retry so
* shortcut the timeouts and just signal too many retries.
@@ -567,7 +569,7 @@ again:
send_comp:
spin_lock_irqsave(&sqp->s_lock, flags);
- ibp->n_loop_pkts++;
+ ibp->rvp.n_loop_pkts++;
flush_send:
sqp->s_rnr_retry = sqp->s_rnr_retry_cnt;
hfi1_send_complete(sqp, wqe, send_status);
@@ -577,7 +579,7 @@ rnr_nak:
/* Handle RNR NAK */
if (qp->ibqp.qp_type == IB_QPT_UC)
goto send_comp;
- ibp->n_rnr_naks++;
+ ibp->rvp.n_rnr_naks++;
/*
* Note: we don't need the s_lock held since the BUSY flag
* makes this single threaded.
@@ -663,7 +665,7 @@ u32 hfi1_make_grh(struct hfi1_ibport *ibp, struct ib_grh *hdr,
hdr->next_hdr = IB_GRH_NEXT_HDR;
hdr->hop_limit = grh->hop_limit;
/* The SGID is 32-bit aligned. */
- hdr->sgid.global.subnet_prefix = ibp->gid_prefix;
+ hdr->sgid.global.subnet_prefix = ibp->rvp.gid_prefix;
hdr->sgid.global.interface_id =
grh->sgid_index && grh->sgid_index < ARRAY_SIZE(ibp->guids) ?
ibp->guids[grh->sgid_index - 1] :
diff --git a/drivers/staging/rdma/hfi1/uc.c b/drivers/staging/rdma/hfi1/uc.c
index 6686331..cac3724 100644
--- a/drivers/staging/rdma/hfi1/uc.c
+++ b/drivers/staging/rdma/hfi1/uc.c
@@ -579,7 +579,7 @@ rewind:
set_bit(HFI1_R_REWIND_SGE, &qp->r_aflags);
qp->r_sge.num_sge = 0;
drop:
- ibp->n_pkt_drops++;
+ ibp->rvp.n_pkt_drops++;
return;
op_err:
diff --git a/drivers/staging/rdma/hfi1/ud.c b/drivers/staging/rdma/hfi1/ud.c
index b14641c..f1d43b3 100644
--- a/drivers/staging/rdma/hfi1/ud.c
+++ b/drivers/staging/rdma/hfi1/ud.c
@@ -82,7 +82,7 @@ static void ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
qp = hfi1_lookup_qpn(ibp, swqe->ud_wr.remote_qpn);
if (!qp) {
- ibp->n_pkt_drops++;
+ ibp->rvp.n_pkt_drops++;
rcu_read_unlock();
return;
}
@@ -94,7 +94,7 @@ static void ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
if (dqptype != sqptype ||
!(ib_hfi1_state_ops[qp->state] & HFI1_PROCESS_RECV_OK)) {
- ibp->n_pkt_drops++;
+ ibp->rvp.n_pkt_drops++;
goto drop;
}
@@ -174,14 +174,14 @@ static void ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
}
if (!ret) {
if (qp->ibqp.qp_num == 0)
- ibp->n_vl15_dropped++;
+ ibp->rvp.n_vl15_dropped++;
goto bail_unlock;
}
}
/* Silently drop packets which are too big. */
if (unlikely(wc.byte_len > qp->r_len)) {
qp->r_flags |= HFI1_R_REUSE_SGE;
- ibp->n_pkt_drops++;
+ ibp->rvp.n_pkt_drops++;
goto bail_unlock;
}
@@ -250,7 +250,7 @@ static void ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe)
/* Signal completion event if the solicited bit is set. */
hfi1_cq_enter(to_icq(qp->ibqp.recv_cq), &wc,
swqe->wr.send_flags & IB_SEND_SOLICITED);
- ibp->n_loop_pkts++;
+ ibp->rvp.n_loop_pkts++;
bail_unlock:
spin_unlock_irqrestore(&qp->r_lock, flags);
drop:
@@ -609,7 +609,7 @@ static int opa_smp_check(struct hfi1_ibport *ibp, u16 pkey, u8 sc5,
case IB_MGMT_METHOD_TRAP:
case IB_MGMT_METHOD_GET_RESP:
case IB_MGMT_METHOD_REPORT_RESP:
- if (ibp->port_cap_flags & IB_PORT_SM)
+ if (ibp->rvp.port_cap_flags & IB_PORT_SM)
return 0;
if (pkey == FULL_MGMT_P_KEY) {
smp->status |= IB_SMP_UNSUP_METHOD;
@@ -823,7 +823,7 @@ void hfi1_ud_rcv(struct hfi1_packet *packet)
}
if (!ret) {
if (qp->ibqp.qp_num == 0)
- ibp->n_vl15_dropped++;
+ ibp->rvp.n_vl15_dropped++;
return;
}
}
@@ -883,5 +883,5 @@ void hfi1_ud_rcv(struct hfi1_packet *packet)
return;
drop:
- ibp->n_pkt_drops++;
+ ibp->rvp.n_pkt_drops++;
}
diff --git a/drivers/staging/rdma/hfi1/verbs.c b/drivers/staging/rdma/hfi1/verbs.c
index 38069e6..895d07a 100644
--- a/drivers/staging/rdma/hfi1/verbs.c
+++ b/drivers/staging/rdma/hfi1/verbs.c
@@ -593,7 +593,7 @@ static inline int qp_ok(int opcode, struct hfi1_packet *packet)
return 1;
dropit:
ibp = &packet->rcd->ppd->ibport_data;
- ibp->n_pkt_drops++;
+ ibp->rvp.n_pkt_drops++;
return 0;
}
@@ -683,7 +683,7 @@ void hfi1_ib_rcv(struct hfi1_packet *packet)
return;
drop:
- ibp->n_pkt_drops++;
+ ibp->rvp.n_pkt_drops++;
}
/*
@@ -1464,17 +1464,17 @@ static int query_port(struct ib_device *ibdev, u8 port,
memset(props, 0, sizeof(*props));
props->lid = lid ? lid : 0;
props->lmc = ppd->lmc;
- props->sm_lid = ibp->sm_lid;
- props->sm_sl = ibp->sm_sl;
+ props->sm_lid = ibp->rvp.sm_lid;
+ props->sm_sl = ibp->rvp.sm_sl;
/* OPA logical states match IB logical states */
props->state = driver_lstate(ppd);
props->phys_state = hfi1_ibphys_portstate(ppd);
- props->port_cap_flags = ibp->port_cap_flags;
+ props->port_cap_flags = ibp->rvp.port_cap_flags;
props->gid_tbl_len = HFI1_GUIDS_PER_PORT;
props->max_msg_sz = 0x80000000;
props->pkey_tbl_len = hfi1_get_npkeys(dd);
- props->bad_pkey_cntr = ibp->pkey_violations;
- props->qkey_viol_cntr = ibp->qkey_violations;
+ props->bad_pkey_cntr = ibp->rvp.pkey_violations;
+ props->qkey_viol_cntr = ibp->rvp.qkey_violations;
props->active_width = (u8)opa_width_to_ib(ppd->link_width_active);
/* see rate_show() in ib core/sysfs.c */
props->active_speed = (u8)opa_speed_to_ib(ppd->link_speed_active);
@@ -1493,7 +1493,7 @@ static int query_port(struct ib_device *ibdev, u8 port,
4096 : hfi1_max_mtu), IB_MTU_4096);
props->active_mtu = !valid_ib_mtu(ppd->ibmtu) ? props->max_mtu :
mtu_to_enum(ppd->ibmtu, IB_MTU_2048);
- props->subnet_timeout = ibp->subnet_timeout;
+ props->subnet_timeout = ibp->rvp.subnet_timeout;
return 0;
}
@@ -1564,8 +1564,8 @@ static int modify_port(struct ib_device *ibdev, u8 port,
struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
int ret = 0;
- ibp->port_cap_flags |= props->set_port_cap_mask;
- ibp->port_cap_flags &= ~props->clr_port_cap_mask;
+ ibp->rvp.port_cap_flags |= props->set_port_cap_mask;
+ ibp->rvp.port_cap_flags &= ~props->clr_port_cap_mask;
if (props->set_port_cap_mask || props->clr_port_cap_mask)
hfi1_cap_mask_chg(ibp);
if (port_modify_mask & IB_PORT_SHUTDOWN) {
@@ -1574,7 +1574,7 @@ static int modify_port(struct ib_device *ibdev, u8 port,
ret = set_link_state(ppd, HLS_DN_DOWNDEF);
}
if (port_modify_mask & IB_PORT_RESET_QKEY_CNTR)
- ibp->qkey_violations = 0;
+ ibp->rvp.qkey_violations = 0;
return ret;
}
@@ -1590,7 +1590,7 @@ static int query_gid(struct ib_device *ibdev, u8 port,
struct hfi1_ibport *ibp = to_iport(ibdev, port);
struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
- gid->global.subnet_prefix = ibp->gid_prefix;
+ gid->global.subnet_prefix = ibp->rvp.gid_prefix;
if (index == 0)
gid->global.interface_id = cpu_to_be64(ppd->guid);
else if (index < HFI1_GUIDS_PER_PORT)
@@ -1662,7 +1662,7 @@ struct ib_ah *hfi1_create_qp0_ah(struct hfi1_ibport *ibp, u16 dlid)
attr.dlid = dlid;
attr.port_num = ppd_from_ibp(ibp)->port;
rcu_read_lock();
- qp0 = rcu_dereference(ibp->qp[0]);
+ qp0 = rcu_dereference(ibp->rvp.qp[0]);
if (qp0)
ah = ib_create_ah(qp0->ibqp.pd, &attr);
rcu_read_unlock();
@@ -1737,21 +1737,21 @@ static void init_ibport(struct hfi1_pportdata *ppd)
ibp->sc_to_sl[i] = i;
}
- spin_lock_init(&ibp->lock);
+ spin_lock_init(&ibp->rvp.lock);
/* Set the prefix to the default value (see ch. 4.1.1) */
- ibp->gid_prefix = IB_DEFAULT_GID_PREFIX;
- ibp->sm_lid = 0;
+ ibp->rvp.gid_prefix = IB_DEFAULT_GID_PREFIX;
+ ibp->rvp.sm_lid = 0;
/* Below should only set bits defined in OPA PortInfo.CapabilityMask */
- ibp->port_cap_flags = IB_PORT_AUTO_MIGR_SUP |
+ ibp->rvp.port_cap_flags = IB_PORT_AUTO_MIGR_SUP |
IB_PORT_CAP_MASK_NOTICE_SUP;
- ibp->pma_counter_select[0] = IB_PMA_PORT_XMIT_DATA;
- ibp->pma_counter_select[1] = IB_PMA_PORT_RCV_DATA;
- ibp->pma_counter_select[2] = IB_PMA_PORT_XMIT_PKTS;
- ibp->pma_counter_select[3] = IB_PMA_PORT_RCV_PKTS;
- ibp->pma_counter_select[4] = IB_PMA_PORT_XMIT_WAIT;
-
- RCU_INIT_POINTER(ibp->qp[0], NULL);
- RCU_INIT_POINTER(ibp->qp[1], NULL);
+ ibp->rvp.pma_counter_select[0] = IB_PMA_PORT_XMIT_DATA;
+ ibp->rvp.pma_counter_select[1] = IB_PMA_PORT_RCV_DATA;
+ ibp->rvp.pma_counter_select[2] = IB_PMA_PORT_XMIT_PKTS;
+ ibp->rvp.pma_counter_select[3] = IB_PMA_PORT_RCV_PKTS;
+ ibp->rvp.pma_counter_select[4] = IB_PMA_PORT_XMIT_WAIT;
+
+ RCU_INIT_POINTER(ibp->rvp.qp[0], NULL);
+ RCU_INIT_POINTER(ibp->rvp.qp[1], NULL);
}
static void verbs_txreq_kmem_cache_ctor(void *obj)
@@ -1926,6 +1926,15 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
dd->verbs_dev.rdi.flags = (RVT_FLAG_QP_INIT_DRIVER |
RVT_FLAG_CQ_INIT_DRIVER);
dd->verbs_dev.rdi.dparms.lkey_table_size = hfi1_lkey_table_size;
+ dd->verbs_dev.rdi.dparms.nports = dd->num_pports;
+ dd->verbs_dev.rdi.dparms.npkeys = hfi1_get_npkeys(dd);
+
+ ppd = dd->pport;
+ for (i = 0; i < dd->num_pports; i++, ppd++)
+ rvt_init_port(&dd->verbs_dev.rdi,
+ &ppd->ibport_data.rvp,
+ i,
+ ppd->pkeys);
ret = rvt_register_device(&dd->verbs_dev.rdi);
if (ret)
@@ -2003,7 +2012,7 @@ void hfi1_cnp_rcv(struct hfi1_packet *packet)
svc_type = IB_CC_SVCTYPE_UD;
break;
default:
- ibp->n_pkt_drops++;
+ ibp->rvp.n_pkt_drops++;
return;
}
diff --git a/drivers/staging/rdma/hfi1/verbs.h b/drivers/staging/rdma/hfi1/verbs.h
index f4ec83c..bc21f1a 100644
--- a/drivers/staging/rdma/hfi1/verbs.h
+++ b/drivers/staging/rdma/hfi1/verbs.h
@@ -410,62 +410,17 @@ static inline void inc_opstats(
struct hfi1_ibport {
struct rvt_qp __rcu *qp[2];
- struct ib_mad_agent *send_agent; /* agent for SMI (traps) */
+ struct rvt_ibport rvp;
struct rvt_ah *sm_ah;
struct rvt_ah *smi_ah;
- struct rb_root mcast_tree;
- spinlock_t lock; /* protect changes in this struct */
-
- /* non-zero when timer is set */
- unsigned long mkey_lease_timeout;
- unsigned long trap_timeout;
- __be64 gid_prefix; /* in network order */
- __be64 mkey;
+
__be64 guids[HFI1_GUIDS_PER_PORT - 1]; /* writable GUIDs */
- u64 tid; /* TID for traps */
- u64 n_rc_resends;
- u64 n_seq_naks;
- u64 n_rdma_seq;
- u64 n_rnr_naks;
- u64 n_other_naks;
- u64 n_loop_pkts;
- u64 n_pkt_drops;
- u64 n_vl15_dropped;
- u64 n_rc_timeouts;
- u64 n_dmawait;
- u64 n_unaligned;
- u64 n_rc_dupreq;
- u64 n_rc_seqnak;
-
- /* Hot-path per CPU counters to avoid cacheline trading to update */
- u64 z_rc_acks;
- u64 z_rc_qacks;
- u64 z_rc_delayed_comp;
- u64 __percpu *rc_acks;
- u64 __percpu *rc_qacks;
- u64 __percpu *rc_delayed_comp;
-
- u32 port_cap_flags;
- u32 pma_sample_start;
- u32 pma_sample_interval;
- __be16 pma_counter_select[5];
- u16 pma_tag;
- u16 pkey_violations;
- u16 qkey_violations;
- u16 mkey_violations;
- u16 mkey_lease_period;
- u16 sm_lid;
- u16 repress_traps;
- u8 sm_sl;
- u8 mkeyprot;
- u8 subnet_timeout;
- u8 vl_high_limit;
+
/* the first 16 entries are sl_to_vl for !OPA */
u8 sl_to_sc[32];
u8 sc_to_sl[32];
};
-
struct hfi1_qp_ibdev;
struct hfi1_ibdev {
struct rvt_dev_info rdi; /* Must be first */
diff --git a/drivers/staging/rdma/hfi1/verbs_mcast.c b/drivers/staging/rdma/hfi1/verbs_mcast.c
index 49954b9..aa3f560 100644
--- a/drivers/staging/rdma/hfi1/verbs_mcast.c
+++ b/drivers/staging/rdma/hfi1/verbs_mcast.c
@@ -131,8 +131,8 @@ struct hfi1_mcast *hfi1_mcast_find(struct hfi1_ibport *ibp, union ib_gid *mgid)
unsigned long flags;
struct hfi1_mcast *mcast;
- spin_lock_irqsave(&ibp->lock, flags);
- n = ibp->mcast_tree.rb_node;
+ spin_lock_irqsave(&ibp->rvp.lock, flags);
+ n = ibp->rvp.mcast_tree.rb_node;
while (n) {
int ret;
@@ -146,11 +146,11 @@ struct hfi1_mcast *hfi1_mcast_find(struct hfi1_ibport *ibp, union ib_gid *mgid)
n = n->rb_right;
else {
atomic_inc(&mcast->refcount);
- spin_unlock_irqrestore(&ibp->lock, flags);
+ spin_unlock_irqrestore(&ibp->rvp.lock, flags);
goto bail;
}
}
- spin_unlock_irqrestore(&ibp->lock, flags);
+ spin_unlock_irqrestore(&ibp->rvp.lock, flags);
mcast = NULL;
@@ -170,11 +170,11 @@ bail:
static int mcast_add(struct hfi1_ibdev *dev, struct hfi1_ibport *ibp,
struct hfi1_mcast *mcast, struct hfi1_mcast_qp *mqp)
{
- struct rb_node **n = &ibp->mcast_tree.rb_node;
+ struct rb_node **n = &ibp->rvp.mcast_tree.rb_node;
struct rb_node *pn = NULL;
int ret;
- spin_lock_irq(&ibp->lock);
+ spin_lock_irq(&ibp->rvp.lock);
while (*n) {
struct hfi1_mcast *tmcast;
@@ -229,12 +229,12 @@ static int mcast_add(struct hfi1_ibdev *dev, struct hfi1_ibport *ibp,
atomic_inc(&mcast->refcount);
rb_link_node(&mcast->rb_node, pn, n);
- rb_insert_color(&mcast->rb_node, &ibp->mcast_tree);
+ rb_insert_color(&mcast->rb_node, &ibp->rvp.mcast_tree);
ret = 0;
bail:
- spin_unlock_irq(&ibp->lock);
+ spin_unlock_irq(&ibp->rvp.lock);
return ret;
}
@@ -313,13 +313,13 @@ int hfi1_multicast_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
goto bail;
}
- spin_lock_irq(&ibp->lock);
+ spin_lock_irq(&ibp->rvp.lock);
/* Find the GID in the mcast table. */
- n = ibp->mcast_tree.rb_node;
+ n = ibp->rvp.mcast_tree.rb_node;
while (1) {
if (n == NULL) {
- spin_unlock_irq(&ibp->lock);
+ spin_unlock_irq(&ibp->rvp.lock);
ret = -EINVAL;
goto bail;
}
@@ -348,13 +348,13 @@ int hfi1_multicast_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
/* If this was the last attached QP, remove the GID too. */
if (list_empty(&mcast->qp_list)) {
- rb_erase(&mcast->rb_node, &ibp->mcast_tree);
+ rb_erase(&mcast->rb_node, &ibp->rvp.mcast_tree);
last = 1;
}
break;
}
- spin_unlock_irq(&ibp->lock);
+ spin_unlock_irq(&ibp->rvp.lock);
if (p) {
/*
@@ -381,5 +381,5 @@ bail:
int hfi1_mcast_tree_empty(struct hfi1_ibport *ibp)
{
- return ibp->mcast_tree.rb_node == NULL;
+ return !ibp->rvp.mcast_tree.rb_node;
}
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [RFC PATCH 14/15] IB/hfi1: Remove mmap from hfi1
[not found] ` <20151214172509.11576.19722.stgit-K+u1se/DcYrLESAwzcoQNrvm/XP+8Wra@public.gmane.org>
` (10 preceding siblings ...)
2015-12-14 17:29 ` [RFC PATCH 13/15] IB/hfi1: Remove ibport and use rdmavt version Dennis Dalessandro
@ 2015-12-14 17:29 ` Dennis Dalessandro
2015-12-21 6:12 ` [RFC PATCH 00/15] staging/rdma/hfi1: Initial patches to add rdmavt support in HFI1 ira.weiny
12 siblings, 0 replies; 25+ messages in thread
From: Dennis Dalessandro @ 2015-12-14 17:29 UTC (permalink / raw)
To: devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-rdma-u79uwXL29TY76Z2rM5mHXA
Cc: gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r, dledford-H+wXaHxf7aLQT0dZR+AlfA
Mmap data structure has already been moved to rdmavt and hfi1 supports
it. Now that the mmap functionality has also been moved to rdmavt its
time for hfi1 to use that as well.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Jubin John <jubin.john-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
drivers/staging/rdma/hfi1/Makefile | 2
drivers/staging/rdma/hfi1/cq.c | 18 ++-
drivers/staging/rdma/hfi1/mmap.c | 192 ------------------------------------
drivers/staging/rdma/hfi1/qp.c | 12 +-
drivers/staging/rdma/hfi1/srq.c | 20 ++--
drivers/staging/rdma/hfi1/verbs.c | 6 -
drivers/staging/rdma/hfi1/verbs.h | 17 ---
7 files changed, 27 insertions(+), 240 deletions(-)
delete mode 100644 drivers/staging/rdma/hfi1/mmap.c
diff --git a/drivers/staging/rdma/hfi1/Makefile b/drivers/staging/rdma/hfi1/Makefile
index 3ba64fe..ff663b3 100644
--- a/drivers/staging/rdma/hfi1/Makefile
+++ b/drivers/staging/rdma/hfi1/Makefile
@@ -8,7 +8,7 @@
obj-$(CONFIG_INFINIBAND_HFI1) += hfi1.o
hfi1-y := chip.o cq.o device.o diag.o driver.o eprom.o file_ops.o firmware.o \
- init.o intr.o mad.o mmap.o pcie.o pio.o pio_copy.o \
+ init.o intr.o mad.o pcie.o pio.o pio_copy.o \
qp.o qsfp.o rc.o ruc.o sdma.o srq.o sysfs.o trace.o twsi.o \
uc.o ud.o user_pages.o user_sdma.o verbs_mcast.o verbs.o
hfi1-$(CONFIG_DEBUG_FS) += debugfs.o
diff --git a/drivers/staging/rdma/hfi1/cq.c b/drivers/staging/rdma/hfi1/cq.c
index ffd0e7a..25d1a2a 100644
--- a/drivers/staging/rdma/hfi1/cq.c
+++ b/drivers/staging/rdma/hfi1/cq.c
@@ -277,7 +277,7 @@ struct ib_cq *hfi1_create_cq(
if (udata && udata->outlen >= sizeof(__u64)) {
int err;
- cq->ip = hfi1_create_mmap_info(dev, sz, context, wc);
+ cq->ip = rvt_create_mmap_info(&dev->rdi, sz, context, wc);
if (!cq->ip) {
ret = ERR_PTR(-ENOMEM);
goto bail_wc;
@@ -303,9 +303,9 @@ struct ib_cq *hfi1_create_cq(
spin_unlock(&dev->n_cqs_lock);
if (cq->ip) {
- spin_lock_irq(&dev->pending_lock);
- list_add(&cq->ip->pending_mmaps, &dev->pending_mmaps);
- spin_unlock_irq(&dev->pending_lock);
+ spin_lock_irq(&dev->rdi.pending_lock);
+ list_add(&cq->ip->pending_mmaps, &dev->rdi.pending_mmaps);
+ spin_unlock_irq(&dev->rdi.pending_lock);
}
/*
@@ -355,7 +355,7 @@ int hfi1_destroy_cq(struct ib_cq *ibcq)
dev->n_cqs_allocated--;
spin_unlock(&dev->n_cqs_lock);
if (cq->ip)
- kref_put(&cq->ip->ref, hfi1_release_mmap_info);
+ kref_put(&cq->ip->ref, rvt_release_mmap_info);
else
vfree(cq->queue);
kfree(cq);
@@ -481,7 +481,7 @@ int hfi1_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata)
struct hfi1_ibdev *dev = to_idev(ibcq->device);
struct rvt_mmap_info *ip = cq->ip;
- hfi1_update_mmap_info(dev, ip, sz, wc);
+ rvt_update_mmap_info(&dev->rdi, ip, sz, wc);
/*
* Return the offset to mmap.
@@ -494,10 +494,10 @@ int hfi1_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata)
goto bail;
}
- spin_lock_irq(&dev->pending_lock);
+ spin_lock_irq(&dev->rdi.pending_lock);
if (list_empty(&ip->pending_mmaps))
- list_add(&ip->pending_mmaps, &dev->pending_mmaps);
- spin_unlock_irq(&dev->pending_lock);
+ list_add(&ip->pending_mmaps, &dev->rdi.pending_mmaps);
+ spin_unlock_irq(&dev->rdi.pending_lock);
}
ret = 0;
diff --git a/drivers/staging/rdma/hfi1/mmap.c b/drivers/staging/rdma/hfi1/mmap.c
deleted file mode 100644
index 4ce6be6..0000000
--- a/drivers/staging/rdma/hfi1/mmap.c
+++ /dev/null
@@ -1,192 +0,0 @@
-/*
- *
- * This file is provided under a dual BSD/GPLv2 license. When using or
- * redistributing this file, you may do so under either license.
- *
- * GPL LICENSE SUMMARY
- *
- * Copyright(c) 2015 Intel Corporation.
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of version 2 of the GNU General Public License as
- * published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- * General Public License for more details.
- *
- * BSD LICENSE
- *
- * Copyright(c) 2015 Intel Corporation.
- *
- * 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.
- * - Neither the name of Intel Corporation nor the names of its
- * contributors may be used to endorse or promote products derived
- * from this software without specific prior written permission.
- *
- * 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.
- *
- */
-
-#include <linux/module.h>
-#include <linux/slab.h>
-#include <linux/vmalloc.h>
-#include <linux/mm.h>
-#include <linux/errno.h>
-#include <asm/pgtable.h>
-
-#include "verbs.h"
-
-/**
- * hfi1_release_mmap_info - free mmap info structure
- * @ref: a pointer to the kref within struct rvt_mmap_info
- */
-void hfi1_release_mmap_info(struct kref *ref)
-{
- struct rvt_mmap_info *ip =
- container_of(ref, struct rvt_mmap_info, ref);
- struct hfi1_ibdev *dev = to_idev(ip->context->device);
-
- spin_lock_irq(&dev->pending_lock);
- list_del(&ip->pending_mmaps);
- spin_unlock_irq(&dev->pending_lock);
-
- vfree(ip->obj);
- kfree(ip);
-}
-
-/*
- * open and close keep track of how many times the CQ is mapped,
- * to avoid releasing it.
- */
-static void hfi1_vma_open(struct vm_area_struct *vma)
-{
- struct rvt_mmap_info *ip = vma->vm_private_data;
-
- kref_get(&ip->ref);
-}
-
-static void hfi1_vma_close(struct vm_area_struct *vma)
-{
- struct rvt_mmap_info *ip = vma->vm_private_data;
-
- kref_put(&ip->ref, hfi1_release_mmap_info);
-}
-
-static struct vm_operations_struct hfi1_vm_ops = {
- .open = hfi1_vma_open,
- .close = hfi1_vma_close,
-};
-
-/**
- * hfi1_mmap - create a new mmap region
- * @context: the IB user context of the process making the mmap() call
- * @vma: the VMA to be initialized
- * Return zero if the mmap is OK. Otherwise, return an errno.
- */
-int hfi1_mmap(struct ib_ucontext *context, struct vm_area_struct *vma)
-{
- struct hfi1_ibdev *dev = to_idev(context->device);
- unsigned long offset = vma->vm_pgoff << PAGE_SHIFT;
- unsigned long size = vma->vm_end - vma->vm_start;
- struct rvt_mmap_info *ip, *pp;
- int ret = -EINVAL;
-
- /*
- * Search the device's list of objects waiting for a mmap call.
- * Normally, this list is very short since a call to create a
- * CQ, QP, or SRQ is soon followed by a call to mmap().
- */
- spin_lock_irq(&dev->pending_lock);
- list_for_each_entry_safe(ip, pp, &dev->pending_mmaps,
- pending_mmaps) {
- /* Only the creator is allowed to mmap the object */
- if (context != ip->context || (__u64) offset != ip->offset)
- continue;
- /* Don't allow a mmap larger than the object. */
- if (size > ip->size)
- break;
-
- list_del_init(&ip->pending_mmaps);
- spin_unlock_irq(&dev->pending_lock);
-
- ret = remap_vmalloc_range(vma, ip->obj, 0);
- if (ret)
- goto done;
- vma->vm_ops = &hfi1_vm_ops;
- vma->vm_private_data = ip;
- hfi1_vma_open(vma);
- goto done;
- }
- spin_unlock_irq(&dev->pending_lock);
-done:
- return ret;
-}
-
-/*
- * Allocate information for hfi1_mmap
- */
-struct rvt_mmap_info *hfi1_create_mmap_info(struct hfi1_ibdev *dev,
- u32 size,
- struct ib_ucontext *context,
- void *obj) {
- struct rvt_mmap_info *ip;
-
- ip = kmalloc(sizeof(*ip), GFP_KERNEL);
- if (!ip)
- goto bail;
-
- size = PAGE_ALIGN(size);
-
- spin_lock_irq(&dev->mmap_offset_lock);
- if (dev->mmap_offset == 0)
- dev->mmap_offset = PAGE_SIZE;
- ip->offset = dev->mmap_offset;
- dev->mmap_offset += size;
- spin_unlock_irq(&dev->mmap_offset_lock);
-
- INIT_LIST_HEAD(&ip->pending_mmaps);
- ip->size = size;
- ip->context = context;
- ip->obj = obj;
- kref_init(&ip->ref);
-
-bail:
- return ip;
-}
-
-void hfi1_update_mmap_info(struct hfi1_ibdev *dev, struct rvt_mmap_info *ip,
- u32 size, void *obj)
-{
- size = PAGE_ALIGN(size);
-
- spin_lock_irq(&dev->mmap_offset_lock);
- if (dev->mmap_offset == 0)
- dev->mmap_offset = PAGE_SIZE;
- ip->offset = dev->mmap_offset;
- dev->mmap_offset += size;
- spin_unlock_irq(&dev->mmap_offset_lock);
-
- ip->size = size;
- ip->obj = obj;
-}
diff --git a/drivers/staging/rdma/hfi1/qp.c b/drivers/staging/rdma/hfi1/qp.c
index 18e7276..b09aaa3 100644
--- a/drivers/staging/rdma/hfi1/qp.c
+++ b/drivers/staging/rdma/hfi1/qp.c
@@ -1195,7 +1195,7 @@ struct ib_qp *hfi1_create_qp(struct ib_pd *ibpd,
} else {
u32 s = sizeof(struct rvt_rwq) + qp->r_rq.size * sz;
- qp->ip = hfi1_create_mmap_info(dev, s,
+ qp->ip = rvt_create_mmap_info(&dev->rdi, s,
ibpd->uobject->context,
qp->r_rq.wq);
if (!qp->ip) {
@@ -1223,9 +1223,9 @@ struct ib_qp *hfi1_create_qp(struct ib_pd *ibpd,
spin_unlock(&dev->n_qps_lock);
if (qp->ip) {
- spin_lock_irq(&dev->pending_lock);
- list_add(&qp->ip->pending_mmaps, &dev->pending_mmaps);
- spin_unlock_irq(&dev->pending_lock);
+ spin_lock_irq(&dev->rdi.pending_lock);
+ list_add(&qp->ip->pending_mmaps, &dev->rdi.pending_mmaps);
+ spin_unlock_irq(&dev->rdi.pending_lock);
}
ret = &qp->ibqp;
@@ -1256,7 +1256,7 @@ struct ib_qp *hfi1_create_qp(struct ib_pd *ibpd,
bail_ip:
if (qp->ip)
- kref_put(&qp->ip->ref, hfi1_release_mmap_info);
+ kref_put(&qp->ip->ref, rvt_release_mmap_info);
else
vfree(qp->r_rq.wq);
free_qpn(&dev->qp_dev->qpn_table, qp->ibqp.qp_num);
@@ -1316,7 +1316,7 @@ int hfi1_destroy_qp(struct ib_qp *ibqp)
spin_unlock(&dev->n_qps_lock);
if (qp->ip)
- kref_put(&qp->ip->ref, hfi1_release_mmap_info);
+ kref_put(&qp->ip->ref, rvt_release_mmap_info);
else
vfree(qp->r_rq.wq);
vfree(qp->s_wq);
diff --git a/drivers/staging/rdma/hfi1/srq.c b/drivers/staging/rdma/hfi1/srq.c
index 78f190a..c53b378 100644
--- a/drivers/staging/rdma/hfi1/srq.c
+++ b/drivers/staging/rdma/hfi1/srq.c
@@ -165,8 +165,8 @@ struct ib_srq *hfi1_create_srq(struct ib_pd *ibpd,
u32 s = sizeof(struct rvt_rwq) + srq->rq.size * sz;
srq->ip =
- hfi1_create_mmap_info(dev, s, ibpd->uobject->context,
- srq->rq.wq);
+ rvt_create_mmap_info(&dev->rdi, s, ibpd->uobject->context,
+ srq->rq.wq);
if (!srq->ip) {
ret = ERR_PTR(-ENOMEM);
goto bail_wq;
@@ -200,9 +200,9 @@ struct ib_srq *hfi1_create_srq(struct ib_pd *ibpd,
spin_unlock(&dev->n_srqs_lock);
if (srq->ip) {
- spin_lock_irq(&dev->pending_lock);
- list_add(&srq->ip->pending_mmaps, &dev->pending_mmaps);
- spin_unlock_irq(&dev->pending_lock);
+ spin_lock_irq(&dev->rdi.pending_lock);
+ list_add(&srq->ip->pending_mmaps, &dev->rdi.pending_mmaps);
+ spin_unlock_irq(&dev->rdi.pending_lock);
}
ret = &srq->ibsrq;
@@ -324,7 +324,7 @@ int hfi1_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
struct hfi1_ibdev *dev = to_idev(srq->ibsrq.device);
u32 s = sizeof(struct rvt_rwq) + size * sz;
- hfi1_update_mmap_info(dev, ip, s, wq);
+ rvt_update_mmap_info(&dev->rdi, ip, s, wq);
/*
* Return the offset to mmap.
@@ -341,11 +341,11 @@ int hfi1_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
* Put user mapping info onto the pending list
* unless it already is on the list.
*/
- spin_lock_irq(&dev->pending_lock);
+ spin_lock_irq(&dev->rdi.pending_lock);
if (list_empty(&ip->pending_mmaps))
list_add(&ip->pending_mmaps,
- &dev->pending_mmaps);
- spin_unlock_irq(&dev->pending_lock);
+ &dev->rdi.pending_mmaps);
+ spin_unlock_irq(&dev->rdi.pending_lock);
}
} else if (attr_mask & IB_SRQ_LIMIT) {
spin_lock_irq(&srq->rq.lock);
@@ -388,7 +388,7 @@ int hfi1_destroy_srq(struct ib_srq *ibsrq)
dev->n_srqs_allocated--;
spin_unlock(&dev->n_srqs_lock);
if (srq->ip)
- kref_put(&srq->ip->ref, hfi1_release_mmap_info);
+ kref_put(&srq->ip->ref, rvt_release_mmap_info);
else
vfree(srq->rq.wq);
kfree(srq);
diff --git a/drivers/staging/rdma/hfi1/verbs.c b/drivers/staging/rdma/hfi1/verbs.c
index 895d07a..d208717 100644
--- a/drivers/staging/rdma/hfi1/verbs.c
+++ b/drivers/staging/rdma/hfi1/verbs.c
@@ -1795,11 +1795,7 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
dev->mem_timer.function = mem_timer;
dev->mem_timer.data = (unsigned long) dev;
- INIT_LIST_HEAD(&dev->pending_mmaps);
- spin_lock_init(&dev->pending_lock);
seqlock_init(&dev->iowait_lock);
- dev->mmap_offset = PAGE_SIZE;
- spin_lock_init(&dev->mmap_offset_lock);
INIT_LIST_HEAD(&dev->txwait);
INIT_LIST_HEAD(&dev->memwait);
@@ -1906,7 +1902,7 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
ibdev->attach_mcast = hfi1_multicast_attach;
ibdev->detach_mcast = hfi1_multicast_detach;
ibdev->process_mad = hfi1_process_mad;
- ibdev->mmap = hfi1_mmap;
+ ibdev->mmap = NULL;
ibdev->dma_ops = NULL;
ibdev->get_port_immutable = port_immutable;
diff --git a/drivers/staging/rdma/hfi1/verbs.h b/drivers/staging/rdma/hfi1/verbs.h
index bc21f1a..cbbdfbd 100644
--- a/drivers/staging/rdma/hfi1/verbs.h
+++ b/drivers/staging/rdma/hfi1/verbs.h
@@ -424,9 +424,6 @@ struct hfi1_ibport {
struct hfi1_qp_ibdev;
struct hfi1_ibdev {
struct rvt_dev_info rdi; /* Must be first */
- struct list_head pending_mmaps;
- spinlock_t mmap_offset_lock; /* protect mmap_offset */
- u32 mmap_offset;
struct hfi1_qp_ibdev *qp_dev;
@@ -439,9 +436,6 @@ struct hfi1_ibdev {
struct kmem_cache *verbs_txreq_cache;
struct timer_list mem_timer;
- /* other waiters */
- spinlock_t pending_lock;
-
u64 n_piowait;
u64 n_txwait;
u64 n_kmem_wait;
@@ -678,17 +672,6 @@ static inline void hfi1_put_ss(struct rvt_sge_state *ss)
}
}
-void hfi1_release_mmap_info(struct kref *ref);
-
-struct rvt_mmap_info *hfi1_create_mmap_info(struct hfi1_ibdev *dev, u32 size,
- struct ib_ucontext *context,
- void *obj);
-
-void hfi1_update_mmap_info(struct hfi1_ibdev *dev, struct rvt_mmap_info *ip,
- u32 size, void *obj);
-
-int hfi1_mmap(struct ib_ucontext *context, struct vm_area_struct *vma);
-
int hfi1_get_rwqe(struct rvt_qp *qp, int wr_id_only);
int hfi1_ruc_check_hdr(struct hfi1_ibport *ibp, struct hfi1_ib_header *hdr,
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related [flat|nested] 25+ messages in thread
* [RFC PATCH 15/15] IB/hfi1: Use rdmavt pkey verbs function
2015-12-14 17:27 [RFC PATCH 00/15] staging/rdma/hfi1: Initial patches to add rdmavt support in HFI1 Dennis Dalessandro
` (2 preceding siblings ...)
[not found] ` <20151214172509.11576.19722.stgit-K+u1se/DcYrLESAwzcoQNrvm/XP+8Wra@public.gmane.org>
@ 2015-12-14 17:29 ` Dennis Dalessandro
3 siblings, 0 replies; 25+ messages in thread
From: Dennis Dalessandro @ 2015-12-14 17:29 UTC (permalink / raw)
To: devel, linux-rdma; +Cc: gregkh, dledford
No need to keep providing the query pkey function. This is now being
done in rdmavt. Remove support from hfi1. The allocation and
maintenance of the list still resides in the driver.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
---
drivers/staging/rdma/hfi1/verbs.c | 20 +-------------------
1 files changed, 1 insertions(+), 19 deletions(-)
diff --git a/drivers/staging/rdma/hfi1/verbs.c b/drivers/staging/rdma/hfi1/verbs.c
index d208717..e1f249a 100644
--- a/drivers/staging/rdma/hfi1/verbs.c
+++ b/drivers/staging/rdma/hfi1/verbs.c
@@ -1678,24 +1678,6 @@ unsigned hfi1_get_npkeys(struct hfi1_devdata *dd)
return ARRAY_SIZE(dd->pport[0].pkeys);
}
-static int query_pkey(struct ib_device *ibdev, u8 port, u16 index,
- u16 *pkey)
-{
- struct hfi1_devdata *dd = dd_from_ibdev(ibdev);
- int ret;
-
- if (index >= hfi1_get_npkeys(dd)) {
- ret = -EINVAL;
- goto bail;
- }
-
- *pkey = hfi1_get_pkey(to_iport(ibdev, port), index);
- ret = 0;
-
-bail:
- return ret;
-}
-
/**
* alloc_ucontext - allocate a ucontest
* @ibdev: the infiniband device
@@ -1863,7 +1845,7 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
ibdev->modify_device = modify_device;
ibdev->query_port = query_port;
ibdev->modify_port = modify_port;
- ibdev->query_pkey = query_pkey;
+ ibdev->query_pkey = NULL;
ibdev->query_gid = query_gid;
ibdev->alloc_ucontext = alloc_ucontext;
ibdev->dealloc_ucontext = dealloc_ucontext;
^ permalink raw reply related [flat|nested] 25+ messages in thread
* Re: [RFC PATCH 00/15] staging/rdma/hfi1: Initial patches to add rdmavt support in HFI1
[not found] ` <20151214172509.11576.19722.stgit-K+u1se/DcYrLESAwzcoQNrvm/XP+8Wra@public.gmane.org>
` (11 preceding siblings ...)
2015-12-14 17:29 ` [RFC PATCH 14/15] IB/hfi1: Remove mmap from hfi1 Dennis Dalessandro
@ 2015-12-21 6:12 ` ira.weiny
[not found] ` <20151221061214.GD3860-W4f6Xiosr+yv7QzWx2u06xL4W9x8LtSr@public.gmane.org>
12 siblings, 1 reply; 25+ messages in thread
From: ira.weiny @ 2015-12-21 6:12 UTC (permalink / raw)
To: gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r, Doug Ledford
Cc: devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-rdma-u79uwXL29TY76Z2rM5mHXA, Dalessandro, Dennis
Greg, Doug,
As mentioned below, these patches depend on the new rdmavt library submitted to
Doug on linux-rdma.
We continue to identify (and rework) patches by our other developers which can
be submitted without conflicts with this series. Furthermore, We have, as much
as possible, placed fixes directly into rdmavt such that those changes can be
dropped from hfi1. But at this point, we need to know if and where these are
going to land so that we can start reworking as appropriate.
Therefore, I would like to discuss plans to get hfi1 under the same maintainer
to work through this transitional period.
Basically, At what point should we stop submitting patches to Greg and start
submitting to Doug?
Should we consider the merge window itself as the swap over point and submit
changes to Doug at that point? If so, should we continue to submit what we can
to Greg until then (and continue rebase'ing the series below on that work)? Or
given Gregs backlog, should we stop submitting to Greg sometime prior to the
merge window?
That brings up my final question, at the point of swap over I assume anything
not accepted by Greg should be considered rejected and we need to resubmit to
Doug?
Thanks in advance for any guidance,
Ira
On Mon, Dec 14, 2015 at 12:27:49PM -0500, Dalessandro, Dennis wrote:
> This patch series is being submitted as a Request For Comment only. It depends
> on code submitted to the rdma subsystem [1].
>
> This work is the first submission aimed to satisfy the TODO item for removing
> duplicated code in hfi1. At the time of submission hfi1 and qib contained alot
> of duplicated verbs processing code. The qib driver is having similar changes
> made to use rdmavt. This will result in a common code base that both drivers and
> future drivers such as soft-roce can use.
>
> Note that due to the ongoing submission of hfi1 improvement patches, there will
> likely be a number of conflicts which will still need to be resolved.
>
> We also are still faced with the issue of separate trees for this work as was
> discussed previously [2]. The result of that conversation was to keep the
> drivers in separate trees until the 4.5 merge window. We are hoping that after
> this merge window a single maintainer can take control of hfi1, qib, and rdmavt
> so that these patches can move forward and be applied.
>
> For now though we would like to get feedback on these patches with more to
> follow.
>
> [1] https://www.mail-archive.com/linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org/msg30074.html
> [2] https://www.mail-archive.com/linux-rdma%40vger.kernel.org/msg29360.html
>
> ---
>
> Dennis Dalessandro (15):
> IB/hfi1: Begin to use rdmavt for verbs
> IB/hfi1: Add basic rdmavt capability flags for hfi1
> IB/hfi1: Consolidate dma ops for hfi1
> IB/hfi1: Use rdmavt protection domain
> IB/hfi1: Remove MR data structures from hfi1
> IB/hfi1: Remove driver specific members from hfi1 qp type
> IB/hfi1: Add device specific info prints
> IB/hfi1: Use correct rdmavt header files after move.
> IB/hfi1: Use address handle in rdmavt and remove from hfi1
> IB/hfi1: Implement hfi1 support for AH notification
> IB/hfi1: Remove hfi1 MR and hfi1 specific qp type
> IB/hfi1: Remove srq from hfi1
> IB/hfi1: Remove ibport and use rdmavt version
> IB/hfi1: Remove mmap from hfi1
> IB/hfi1: Use rdmavt pkey verbs function
>
>
> drivers/staging/rdma/hfi1/Kconfig | 2
> drivers/staging/rdma/hfi1/Makefile | 4
> drivers/staging/rdma/hfi1/chip.c | 36 +-
> drivers/staging/rdma/hfi1/common.h | 2
> drivers/staging/rdma/hfi1/cq.c | 20 +
> drivers/staging/rdma/hfi1/diag.c | 13 -
> drivers/staging/rdma/hfi1/driver.c | 31 +-
> drivers/staging/rdma/hfi1/hfi.h | 27 +-
> drivers/staging/rdma/hfi1/init.c | 5
> drivers/staging/rdma/hfi1/intr.c | 2
> drivers/staging/rdma/hfi1/keys.c | 356 ---------------------
> drivers/staging/rdma/hfi1/mad.c | 163 +++++-----
> drivers/staging/rdma/hfi1/mmap.c | 192 -----------
> drivers/staging/rdma/hfi1/mr.c | 522 ------------------------------
> drivers/staging/rdma/hfi1/pio.c | 10 -
> drivers/staging/rdma/hfi1/qp.c | 214 +++++++-----
> drivers/staging/rdma/hfi1/qp.h | 44 +--
> drivers/staging/rdma/hfi1/rc.c | 155 +++++----
> drivers/staging/rdma/hfi1/ruc.c | 161 +++++----
> drivers/staging/rdma/hfi1/sdma.h | 8
> drivers/staging/rdma/hfi1/srq.c | 58 ++-
> drivers/staging/rdma/hfi1/sysfs.c | 18 +
> drivers/staging/rdma/hfi1/trace.h | 22 +
> drivers/staging/rdma/hfi1/uc.c | 19 +
> drivers/staging/rdma/hfi1/ud.c | 91 +++--
> drivers/staging/rdma/hfi1/verbs.c | 526 +++++++++++--------------------
> drivers/staging/rdma/hfi1/verbs.h | 531 ++++---------------------------
> drivers/staging/rdma/hfi1/verbs_mcast.c | 36 +-
> 28 files changed, 843 insertions(+), 2425 deletions(-)
> delete mode 100644 drivers/staging/rdma/hfi1/keys.c
> delete mode 100644 drivers/staging/rdma/hfi1/mmap.c
> delete mode 100644 drivers/staging/rdma/hfi1/mr.c
>
> --
> -Denny
> --
> To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
> the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 25+ messages in thread
* Re: [RFC PATCH 00/15] staging/rdma/hfi1: Initial patches to add rdmavt support in HFI1
[not found] ` <20151221061214.GD3860-W4f6Xiosr+yv7QzWx2u06xL4W9x8LtSr@public.gmane.org>
@ 2015-12-21 22:02 ` gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r
[not found] ` <20151221220235.GA14142-U8xfFu+wG4EAvxtiuMwx3w@public.gmane.org>
0 siblings, 1 reply; 25+ messages in thread
From: gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r @ 2015-12-21 22:02 UTC (permalink / raw)
To: ira.weiny
Cc: Doug Ledford, devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-rdma-u79uwXL29TY76Z2rM5mHXA
On Mon, Dec 21, 2015 at 01:12:14AM -0500, ira.weiny wrote:
> Greg, Doug,
>
> As mentioned below, these patches depend on the new rdmavt library submitted to
> Doug on linux-rdma.
>
> We continue to identify (and rework) patches by our other developers which can
> be submitted without conflicts with this series. Furthermore, We have, as much
> as possible, placed fixes directly into rdmavt such that those changes can be
> dropped from hfi1. But at this point, we need to know if and where these are
> going to land so that we can start reworking as appropriate.
>
> Therefore, I would like to discuss plans to get hfi1 under the same maintainer
> to work through this transitional period.
>
> Basically, At what point should we stop submitting patches to Greg and start
> submitting to Doug?
>
> Should we consider the merge window itself as the swap over point and submit
> changes to Doug at that point? If so, should we continue to submit what we can
> to Greg until then (and continue rebase'ing the series below on that work)? Or
> given Gregs backlog, should we stop submitting to Greg sometime prior to the
> merge window?
>
> That brings up my final question, at the point of swap over I assume anything
> not accepted by Greg should be considered rejected and we need to resubmit to
> Doug?
If Doug accepts the library changes, let me know that public git commit
and I can pull it into the staging-next branch and you can continue to
send me staging patches that way.
That's the easiest thing to do usually.
thanks,
greg k-h
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 25+ messages in thread
* Re: [RFC PATCH 00/15] staging/rdma/hfi1: Initial patches to add rdmavt support in HFI1
[not found] ` <20151221220235.GA14142-U8xfFu+wG4EAvxtiuMwx3w@public.gmane.org>
@ 2015-12-22 0:19 ` ira.weiny
[not found] ` <20151222001943.GM3860-W4f6Xiosr+yv7QzWx2u06xL4W9x8LtSr@public.gmane.org>
0 siblings, 1 reply; 25+ messages in thread
From: ira.weiny @ 2015-12-22 0:19 UTC (permalink / raw)
To: gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r
Cc: Doug Ledford, devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-rdma-u79uwXL29TY76Z2rM5mHXA
On Mon, Dec 21, 2015 at 02:02:35PM -0800, gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r@public.gmane.org wrote:
> On Mon, Dec 21, 2015 at 01:12:14AM -0500, ira.weiny wrote:
> > Greg, Doug,
> >
> > As mentioned below, these patches depend on the new rdmavt library submitted to
> > Doug on linux-rdma.
> >
> > We continue to identify (and rework) patches by our other developers which can
> > be submitted without conflicts with this series. Furthermore, We have, as much
> > as possible, placed fixes directly into rdmavt such that those changes can be
> > dropped from hfi1. But at this point, we need to know if and where these are
> > going to land so that we can start reworking as appropriate.
> >
> > Therefore, I would like to discuss plans to get hfi1 under the same maintainer
> > to work through this transitional period.
> >
> > Basically, At what point should we stop submitting patches to Greg and start
> > submitting to Doug?
> >
> > Should we consider the merge window itself as the swap over point and submit
> > changes to Doug at that point? If so, should we continue to submit what we can
> > to Greg until then (and continue rebase'ing the series below on that work)? Or
> > given Gregs backlog, should we stop submitting to Greg sometime prior to the
> > merge window?
> >
> > That brings up my final question, at the point of swap over I assume anything
> > not accepted by Greg should be considered rejected and we need to resubmit to
> > Doug?
>
> If Doug accepts the library changes, let me know that public git commit
> and I can pull it into the staging-next branch and you can continue to
> send me staging patches that way.
Won't this cause a conflict during the merge window?
How do we handle changes which affect both qib and hfi1?
Ira
>
> That's the easiest thing to do usually.
>
> thanks,
>
> greg k-h
> --
> To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
> the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 25+ messages in thread
* Re: [RFC PATCH 00/15] staging/rdma/hfi1: Initial patches to add rdmavt support in HFI1
[not found] ` <20151222001943.GM3860-W4f6Xiosr+yv7QzWx2u06xL4W9x8LtSr@public.gmane.org>
@ 2015-12-22 1:01 ` gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r
[not found] ` <20151222010148.GB2145-U8xfFu+wG4EAvxtiuMwx3w@public.gmane.org>
0 siblings, 1 reply; 25+ messages in thread
From: gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r @ 2015-12-22 1:01 UTC (permalink / raw)
To: ira.weiny
Cc: Doug Ledford, devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-rdma-u79uwXL29TY76Z2rM5mHXA
On Mon, Dec 21, 2015 at 07:19:43PM -0500, ira.weiny wrote:
> On Mon, Dec 21, 2015 at 02:02:35PM -0800, gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r@public.gmane.org wrote:
> > On Mon, Dec 21, 2015 at 01:12:14AM -0500, ira.weiny wrote:
> > > Greg, Doug,
> > >
> > > As mentioned below, these patches depend on the new rdmavt library submitted to
> > > Doug on linux-rdma.
> > >
> > > We continue to identify (and rework) patches by our other developers which can
> > > be submitted without conflicts with this series. Furthermore, We have, as much
> > > as possible, placed fixes directly into rdmavt such that those changes can be
> > > dropped from hfi1. But at this point, we need to know if and where these are
> > > going to land so that we can start reworking as appropriate.
> > >
> > > Therefore, I would like to discuss plans to get hfi1 under the same maintainer
> > > to work through this transitional period.
> > >
> > > Basically, At what point should we stop submitting patches to Greg and start
> > > submitting to Doug?
> > >
> > > Should we consider the merge window itself as the swap over point and submit
> > > changes to Doug at that point? If so, should we continue to submit what we can
> > > to Greg until then (and continue rebase'ing the series below on that work)? Or
> > > given Gregs backlog, should we stop submitting to Greg sometime prior to the
> > > merge window?
> > >
> > > That brings up my final question, at the point of swap over I assume anything
> > > not accepted by Greg should be considered rejected and we need to resubmit to
> > > Doug?
> >
> > If Doug accepts the library changes, let me know that public git commit
> > and I can pull it into the staging-next branch and you can continue to
> > send me staging patches that way.
>
> Won't this cause a conflict during the merge window?
No, git is good :)
> How do we handle changes which affect both qib and hfi1?
I don't know, now this gets messy...
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 25+ messages in thread
* Re: [RFC PATCH 00/15] staging/rdma/hfi1: Initial patches to add rdmavt support in HFI1
[not found] ` <20151222010148.GB2145-U8xfFu+wG4EAvxtiuMwx3w@public.gmane.org>
@ 2015-12-22 19:15 ` ira.weiny
[not found] ` <20151222191508.GO3860-W4f6Xiosr+yv7QzWx2u06xL4W9x8LtSr@public.gmane.org>
0 siblings, 1 reply; 25+ messages in thread
From: ira.weiny @ 2015-12-22 19:15 UTC (permalink / raw)
To: gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r
Cc: Doug Ledford, devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-rdma-u79uwXL29TY76Z2rM5mHXA
On Mon, Dec 21, 2015 at 05:01:48PM -0800, gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r@public.gmane.org wrote:
> On Mon, Dec 21, 2015 at 07:19:43PM -0500, ira.weiny wrote:
> > On Mon, Dec 21, 2015 at 02:02:35PM -0800, gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r@public.gmane.org wrote:
> > > On Mon, Dec 21, 2015 at 01:12:14AM -0500, ira.weiny wrote:
> > > > Greg, Doug,
> > > >
> > > > As mentioned below, these patches depend on the new rdmavt library submitted to
> > > > Doug on linux-rdma.
> > > >
> > > > We continue to identify (and rework) patches by our other developers which can
> > > > be submitted without conflicts with this series. Furthermore, We have, as much
> > > > as possible, placed fixes directly into rdmavt such that those changes can be
> > > > dropped from hfi1. But at this point, we need to know if and where these are
> > > > going to land so that we can start reworking as appropriate.
> > > >
> > > > Therefore, I would like to discuss plans to get hfi1 under the same maintainer
> > > > to work through this transitional period.
> > > >
> > > > Basically, At what point should we stop submitting patches to Greg and start
> > > > submitting to Doug?
> > > >
> > > > Should we consider the merge window itself as the swap over point and submit
> > > > changes to Doug at that point? If so, should we continue to submit what we can
> > > > to Greg until then (and continue rebase'ing the series below on that work)? Or
> > > > given Gregs backlog, should we stop submitting to Greg sometime prior to the
> > > > merge window?
> > > >
> > > > That brings up my final question, at the point of swap over I assume anything
> > > > not accepted by Greg should be considered rejected and we need to resubmit to
> > > > Doug?
> > >
> > > If Doug accepts the library changes, let me know that public git commit
> > > and I can pull it into the staging-next branch and you can continue to
> > > send me staging patches that way.
> >
> > Won't this cause a conflict during the merge window?
>
> No, git is good :)
>
> > How do we handle changes which affect both qib and hfi1?
>
> I don't know, now this gets messy...
>
Agreed and this is what we are worried about.
Can we do what Dan and Doug have proposed in the past and have Doug take over
the staging/rdma sub-tree?
http://driverdev.linuxdriverproject.org/pipermail/driverdev-devel/2015-November/081922.html
I think the upcoming merge window is a reasonable time for him to do that.
Ira
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 25+ messages in thread
* Re: [RFC PATCH 00/15] staging/rdma/hfi1: Initial patches to add rdmavt support in HFI1
[not found] ` <20151222191508.GO3860-W4f6Xiosr+yv7QzWx2u06xL4W9x8LtSr@public.gmane.org>
@ 2015-12-23 2:27 ` gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r
[not found] ` <20151223022757.GA5562-U8xfFu+wG4EAvxtiuMwx3w@public.gmane.org>
0 siblings, 1 reply; 25+ messages in thread
From: gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r @ 2015-12-23 2:27 UTC (permalink / raw)
To: ira.weiny
Cc: Doug Ledford, devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-rdma-u79uwXL29TY76Z2rM5mHXA
On Tue, Dec 22, 2015 at 02:15:08PM -0500, ira.weiny wrote:
> On Mon, Dec 21, 2015 at 05:01:48PM -0800, gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r@public.gmane.org wrote:
> > On Mon, Dec 21, 2015 at 07:19:43PM -0500, ira.weiny wrote:
> > > On Mon, Dec 21, 2015 at 02:02:35PM -0800, gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r@public.gmane.org wrote:
> > > > On Mon, Dec 21, 2015 at 01:12:14AM -0500, ira.weiny wrote:
> > > > > Greg, Doug,
> > > > >
> > > > > As mentioned below, these patches depend on the new rdmavt library submitted to
> > > > > Doug on linux-rdma.
> > > > >
> > > > > We continue to identify (and rework) patches by our other developers which can
> > > > > be submitted without conflicts with this series. Furthermore, We have, as much
> > > > > as possible, placed fixes directly into rdmavt such that those changes can be
> > > > > dropped from hfi1. But at this point, we need to know if and where these are
> > > > > going to land so that we can start reworking as appropriate.
> > > > >
> > > > > Therefore, I would like to discuss plans to get hfi1 under the same maintainer
> > > > > to work through this transitional period.
> > > > >
> > > > > Basically, At what point should we stop submitting patches to Greg and start
> > > > > submitting to Doug?
> > > > >
> > > > > Should we consider the merge window itself as the swap over point and submit
> > > > > changes to Doug at that point? If so, should we continue to submit what we can
> > > > > to Greg until then (and continue rebase'ing the series below on that work)? Or
> > > > > given Gregs backlog, should we stop submitting to Greg sometime prior to the
> > > > > merge window?
> > > > >
> > > > > That brings up my final question, at the point of swap over I assume anything
> > > > > not accepted by Greg should be considered rejected and we need to resubmit to
> > > > > Doug?
> > > >
> > > > If Doug accepts the library changes, let me know that public git commit
> > > > and I can pull it into the staging-next branch and you can continue to
> > > > send me staging patches that way.
> > >
> > > Won't this cause a conflict during the merge window?
> >
> > No, git is good :)
> >
> > > How do we handle changes which affect both qib and hfi1?
> >
> > I don't know, now this gets messy...
> >
>
> Agreed and this is what we are worried about.
>
> Can we do what Dan and Doug have proposed in the past and have Doug take over
> the staging/rdma sub-tree?
>
> http://driverdev.linuxdriverproject.org/pipermail/driverdev-devel/2015-November/081922.html
>
> I think the upcoming merge window is a reasonable time for him to do that.
Ok, but keeping on top of all of the generic staging patches that come
in is a tough thing to do, that's up to Doug, if he is ready for it...
greg k-h
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 25+ messages in thread
* Re: [RFC PATCH 00/15] staging/rdma/hfi1: Initial patches to add rdmavt support in HFI1
[not found] ` <20151223022757.GA5562-U8xfFu+wG4EAvxtiuMwx3w@public.gmane.org>
@ 2015-12-23 4:21 ` Doug Ledford
2015-12-24 17:50 ` ira.weiny
2016-01-05 22:50 ` ira.weiny
2 siblings, 0 replies; 25+ messages in thread
From: Doug Ledford @ 2015-12-23 4:21 UTC (permalink / raw)
To: gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r, ira.weiny
Cc: devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-rdma-u79uwXL29TY76Z2rM5mHXA
[-- Attachment #1: Type: text/plain, Size: 3017 bytes --]
On 12/22/2015 09:27 PM, gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r@public.gmane.org wrote:
> On Tue, Dec 22, 2015 at 02:15:08PM -0500, ira.weiny wrote:
>> On Mon, Dec 21, 2015 at 05:01:48PM -0800, gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r@public.gmane.org wrote:
>>> On Mon, Dec 21, 2015 at 07:19:43PM -0500, ira.weiny wrote:
>>>> On Mon, Dec 21, 2015 at 02:02:35PM -0800, gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r@public.gmane.org wrote:
>>>>> On Mon, Dec 21, 2015 at 01:12:14AM -0500, ira.weiny wrote:
>>>>>> Greg, Doug,
>>>>>>
>>>>>> As mentioned below, these patches depend on the new rdmavt library submitted to
>>>>>> Doug on linux-rdma.
>>>>>>
>>>>>> We continue to identify (and rework) patches by our other developers which can
>>>>>> be submitted without conflicts with this series. Furthermore, We have, as much
>>>>>> as possible, placed fixes directly into rdmavt such that those changes can be
>>>>>> dropped from hfi1. But at this point, we need to know if and where these are
>>>>>> going to land so that we can start reworking as appropriate.
>>>>>>
>>>>>> Therefore, I would like to discuss plans to get hfi1 under the same maintainer
>>>>>> to work through this transitional period.
>>>>>>
>>>>>> Basically, At what point should we stop submitting patches to Greg and start
>>>>>> submitting to Doug?
>>>>>>
>>>>>> Should we consider the merge window itself as the swap over point and submit
>>>>>> changes to Doug at that point? If so, should we continue to submit what we can
>>>>>> to Greg until then (and continue rebase'ing the series below on that work)? Or
>>>>>> given Gregs backlog, should we stop submitting to Greg sometime prior to the
>>>>>> merge window?
>>>>>>
>>>>>> That brings up my final question, at the point of swap over I assume anything
>>>>>> not accepted by Greg should be considered rejected and we need to resubmit to
>>>>>> Doug?
>>>>>
>>>>> If Doug accepts the library changes, let me know that public git commit
>>>>> and I can pull it into the staging-next branch and you can continue to
>>>>> send me staging patches that way.
>>>>
>>>> Won't this cause a conflict during the merge window?
>>>
>>> No, git is good :)
>>>
>>>> How do we handle changes which affect both qib and hfi1?
>>>
>>> I don't know, now this gets messy...
>>>
>>
>> Agreed and this is what we are worried about.
>>
>> Can we do what Dan and Doug have proposed in the past and have Doug take over
>> the staging/rdma sub-tree?
>>
>> http://driverdev.linuxdriverproject.org/pipermail/driverdev-devel/2015-November/081922.html
>>
>> I think the upcoming merge window is a reasonable time for him to do that.
>
> Ok, but keeping on top of all of the generic staging patches that come
> in is a tough thing to do, that's up to Doug, if he is ready for it...
I'm not worried about that. Patchworks makes the workflow reasonable.
--
Doug Ledford <dledford-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
GPG KeyID: 0E572FDD
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 884 bytes --]
^ permalink raw reply [flat|nested] 25+ messages in thread
* Re: [RFC PATCH 00/15] staging/rdma/hfi1: Initial patches to add rdmavt support in HFI1
[not found] ` <20151223022757.GA5562-U8xfFu+wG4EAvxtiuMwx3w@public.gmane.org>
2015-12-23 4:21 ` Doug Ledford
@ 2015-12-24 17:50 ` ira.weiny
2016-01-05 22:50 ` ira.weiny
2 siblings, 0 replies; 25+ messages in thread
From: ira.weiny @ 2015-12-24 17:50 UTC (permalink / raw)
To: gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r
Cc: Doug Ledford, devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-rdma-u79uwXL29TY76Z2rM5mHXA
On Tue, Dec 22, 2015 at 06:27:57PM -0800, gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r@public.gmane.org wrote:
> On Tue, Dec 22, 2015 at 02:15:08PM -0500, ira.weiny wrote:
> > On Mon, Dec 21, 2015 at 05:01:48PM -0800, gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r@public.gmane.org wrote:
[snip]
> > >
> > > No, git is good :)
> > >
> > > > How do we handle changes which affect both qib and hfi1?
> > >
> > > I don't know, now this gets messy...
> > >
> >
> > Agreed and this is what we are worried about.
> >
> > Can we do what Dan and Doug have proposed in the past and have Doug take over
> > the staging/rdma sub-tree?
> >
> > http://driverdev.linuxdriverproject.org/pipermail/driverdev-devel/2015-November/081922.html
> >
> > I think the upcoming merge window is a reasonable time for him to do that.
>
> Ok, but keeping on top of all of the generic staging patches that come
> in is a tough thing to do, that's up to Doug, if he is ready for it...
>
To help this process, once the change over happens, we will help to monitor
driverdev-devel for anything submitted to staging/rdma. If something is
submitted which was not to Doug and linux-rdma we can handle alerting the
submitter to make sure it gets submitted to Doug as per the current MAINTAINERS
file.
Hope this helps,
Ira
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 25+ messages in thread
* Re: [RFC PATCH 00/15] staging/rdma/hfi1: Initial patches to add rdmavt support in HFI1
[not found] ` <20151223022757.GA5562-U8xfFu+wG4EAvxtiuMwx3w@public.gmane.org>
2015-12-23 4:21 ` Doug Ledford
2015-12-24 17:50 ` ira.weiny
@ 2016-01-05 22:50 ` ira.weiny
2 siblings, 0 replies; 25+ messages in thread
From: ira.weiny @ 2016-01-05 22:50 UTC (permalink / raw)
To: gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r, Doug Ledford
Cc: devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-rdma-u79uwXL29TY76Z2rM5mHXA, Mike Marciniszyn, Jubin John,
Dalessandro, Dennis
> > > > >
> > > > > If Doug accepts the library changes, let me know that public git commit
> > > > > and I can pull it into the staging-next branch and you can continue to
> > > > > send me staging patches that way.
> > > >
> > > > Won't this cause a conflict during the merge window?
> > >
> > > No, git is good :)
> > >
> > > > How do we handle changes which affect both qib and hfi1?
> > >
> > > I don't know, now this gets messy...
> > >
> >
> > Agreed and this is what we are worried about.
> >
> > Can we do what Dan and Doug have proposed in the past and have Doug take over
> > the staging/rdma sub-tree?
> >
> > http://driverdev.linuxdriverproject.org/pipermail/driverdev-devel/2015-November/081922.html
> >
> > I think the upcoming merge window is a reasonable time for him to do that.
>
> Ok, but keeping on top of all of the generic staging patches that come
> in is a tough thing to do, that's up to Doug, if he is ready for it...
>
Greg,
Forgive me for not knowing how multiple maintainers deal with hand offs like
this. I'm hoping you, and/or Doug, can answer some questions for me.
Am I correct in assuming the merge window will be open on Monday? If so, when
will Linus pull the staging tree? Then at what point will Doug get the hfi1
changes which have already been accepted?
Are you going to be able to review the outstanding patches for
staging/rdma/hfi1 before the merge window? Or should we consider them dropped
and resubmit to Doug to apply after he has merged the latest hfi1 code from
Linus?
Doug,
At what point should we start submitting additional patches to you which we
have queued up but are not yet submitted to Greg?
So far we have been cross posting to linux-rdma for feedback and I see that the
patches have been dropped from patchworks. I assume you dropped them from
patchworks because you knew that Greg was handling them?
Thanks,
Ira
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 25+ messages in thread
end of thread, other threads:[~2016-01-05 22:50 UTC | newest]
Thread overview: 25+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2015-12-14 17:27 [RFC PATCH 00/15] staging/rdma/hfi1: Initial patches to add rdmavt support in HFI1 Dennis Dalessandro
2015-12-14 17:28 ` [RFC PATCH 06/15] IB/hfi1: Remove driver specific members from hfi1 qp type Dennis Dalessandro
2015-12-14 17:28 ` [RFC PATCH 07/15] IB/hfi1: Add device specific info prints Dennis Dalessandro
[not found] ` <20151214172509.11576.19722.stgit-K+u1se/DcYrLESAwzcoQNrvm/XP+8Wra@public.gmane.org>
2015-12-14 17:27 ` [RFC PATCH 01/15] IB/hfi1: Begin to use rdmavt for verbs Dennis Dalessandro
2015-12-14 17:28 ` [RFC PATCH 02/15] IB/hfi1: Add basic rdmavt capability flags for hfi1 Dennis Dalessandro
2015-12-14 17:28 ` [RFC PATCH 03/15] IB/hfi1: Consolidate dma ops " Dennis Dalessandro
2015-12-14 17:28 ` [RFC PATCH 04/15] IB/hfi1: Use rdmavt protection domain Dennis Dalessandro
2015-12-14 17:28 ` [RFC PATCH 05/15] IB/hfi1: Remove MR data structures from hfi1 Dennis Dalessandro
2015-12-14 17:28 ` [RFC PATCH 08/15] IB/hfi1: Use correct rdmavt header files after move Dennis Dalessandro
2015-12-14 17:28 ` [RFC PATCH 09/15] IB/hfi1: Use address handle in rdmavt and remove from hfi1 Dennis Dalessandro
2015-12-14 17:29 ` [RFC PATCH 10/15] IB/hfi1: Implement hfi1 support for AH notification Dennis Dalessandro
2015-12-14 17:29 ` [RFC PATCH 11/15] IB/hfi1: Remove hfi1 MR and hfi1 specific qp type Dennis Dalessandro
2015-12-14 17:29 ` [RFC PATCH 12/15] IB/hfi1: Remove srq from hfi1 Dennis Dalessandro
2015-12-14 17:29 ` [RFC PATCH 13/15] IB/hfi1: Remove ibport and use rdmavt version Dennis Dalessandro
2015-12-14 17:29 ` [RFC PATCH 14/15] IB/hfi1: Remove mmap from hfi1 Dennis Dalessandro
2015-12-21 6:12 ` [RFC PATCH 00/15] staging/rdma/hfi1: Initial patches to add rdmavt support in HFI1 ira.weiny
[not found] ` <20151221061214.GD3860-W4f6Xiosr+yv7QzWx2u06xL4W9x8LtSr@public.gmane.org>
2015-12-21 22:02 ` gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r
[not found] ` <20151221220235.GA14142-U8xfFu+wG4EAvxtiuMwx3w@public.gmane.org>
2015-12-22 0:19 ` ira.weiny
[not found] ` <20151222001943.GM3860-W4f6Xiosr+yv7QzWx2u06xL4W9x8LtSr@public.gmane.org>
2015-12-22 1:01 ` gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r
[not found] ` <20151222010148.GB2145-U8xfFu+wG4EAvxtiuMwx3w@public.gmane.org>
2015-12-22 19:15 ` ira.weiny
[not found] ` <20151222191508.GO3860-W4f6Xiosr+yv7QzWx2u06xL4W9x8LtSr@public.gmane.org>
2015-12-23 2:27 ` gregkh-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r
[not found] ` <20151223022757.GA5562-U8xfFu+wG4EAvxtiuMwx3w@public.gmane.org>
2015-12-23 4:21 ` Doug Ledford
2015-12-24 17:50 ` ira.weiny
2016-01-05 22:50 ` ira.weiny
2015-12-14 17:29 ` [RFC PATCH 15/15] IB/hfi1: Use rdmavt pkey verbs function Dennis Dalessandro
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.