[ofa-general] [PATCH 1/2] ehca: remove checkpatch.pl's warnings externs should be avoided in .c files
From b5d0336089b5ebe5b18acb94b2c94c2026cb95ee Mon Sep 17 00:00:00 2001
From: Hoang-Nam Nguyen hnguyen at de.ibm.com
Date: Fri, 27 Jul 2007 10:24:49 +0200
Subject: [PATCH] remove checkpatch.pl's warnings externs should be avoided in
.c files
Signed-off-by: Hoang-Nam Nguyen [EMAIL PROTECTED]
---
drivers/infiniband/hw/ehca/ehca_classes.h |1 +
drivers/infiniband/hw/ehca/ehca_mrmw.c|2 --
drivers/infiniband/hw/ehca/ehca_pd.c |1 -
drivers/infiniband/hw/ehca/hcp_if.c |1 -
drivers/infiniband/hw/ehca/ipz_pt_fn.h|2 ++
5 files changed, 3 insertions(+), 4 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_classes.h
b/drivers/infiniband/hw/ehca/ehca_classes.h
index 3725aa8..b5e9603 100644
--- a/drivers/infiniband/hw/ehca/ehca_classes.h
+++ b/drivers/infiniband/hw/ehca/ehca_classes.h
@@ -322,6 +322,7 @@ extern int ehca_static_rate;
extern int ehca_port_act_time;
extern int ehca_use_hp_mr;
extern int ehca_scaling_code;
+extern int ehca_mr_largepage;
struct ipzu_queue_resp {
u32 qe_size; /* queue entry size */
diff --git a/drivers/infiniband/hw/ehca/ehca_mrmw.c
b/drivers/infiniband/hw/ehca/ehca_mrmw.c
index c1b868b..773ac3f 100644
--- a/drivers/infiniband/hw/ehca/ehca_mrmw.c
+++ b/drivers/infiniband/hw/ehca/ehca_mrmw.c
@@ -64,8 +64,6 @@ enum ehca_mr_pgsize {
EHCA_MR_PGSIZE16M = 0x100L
};
-extern int ehca_mr_largepage;
-
static u32 ehca_encode_hwpage_size(u32 pgsize)
{
u32 idx = 0;
diff --git a/drivers/infiniband/hw/ehca/ehca_pd.c
b/drivers/infiniband/hw/ehca/ehca_pd.c
index 3dafd7f..43bcf08 100644
--- a/drivers/infiniband/hw/ehca/ehca_pd.c
+++ b/drivers/infiniband/hw/ehca/ehca_pd.c
@@ -88,7 +88,6 @@ int ehca_dealloc_pd(struct ib_pd *pd)
u32 cur_pid = current-tgid;
struct ehca_pd *my_pd = container_of(pd, struct ehca_pd, ib_pd);
int i, leftovers = 0;
- extern struct kmem_cache *small_qp_cache;
struct ipz_small_queue_page *page, *tmp;
if (my_pd-ib_pd.uobject my_pd-ib_pd.uobject-context
diff --git a/drivers/infiniband/hw/ehca/hcp_if.c
b/drivers/infiniband/hw/ehca/hcp_if.c
index fdbfebe..24f4541 100644
--- a/drivers/infiniband/hw/ehca/hcp_if.c
+++ b/drivers/infiniband/hw/ehca/hcp_if.c
@@ -758,7 +758,6 @@ u64 hipz_h_register_rpage_mr(const struct
ipz_adapter_handle adapter_handle,
const u64 logical_address_of_page,
const u64 count)
{
- extern int ehca_debug_level;
u64 ret;
if (unlikely(ehca_debug_level = 2)) {
diff --git a/drivers/infiniband/hw/ehca/ipz_pt_fn.h
b/drivers/infiniband/hw/ehca/ipz_pt_fn.h
index c6937a0..a801274 100644
--- a/drivers/infiniband/hw/ehca/ipz_pt_fn.h
+++ b/drivers/infiniband/hw/ehca/ipz_pt_fn.h
@@ -54,6 +54,8 @@
struct ehca_pd;
struct ipz_small_queue_page;
+extern struct kmem_cache *small_qp_cache;
+
/* struct generic ehca page */
struct ipz_page {
u8 entries[EHCA_PAGESIZE];
--
1.5.2
___
general mailing list
general@lists.openfabrics.org
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
[ofa-general] [PATCH 2/2] ehca: correction include order according kernel coding style
From a2794450cbee597cefd7b6e159257583c459d358 Mon Sep 17 00:00:00 2001 From: Hoang-Nam Nguyen hnguyen at de.ibm.com Date: Fri, 27 Jul 2007 10:26:40 +0200 Subject: [PATCH] correction include order according kernel coding style Signed-off-by: Hoang-Nam Nguyen [EMAIL PROTECTED] --- drivers/infiniband/hw/ehca/ehca_mrmw.c |3 +-- 1 files changed, 1 insertions(+), 2 deletions(-) diff --git a/drivers/infiniband/hw/ehca/ehca_mrmw.c b/drivers/infiniband/hw/ehca/ehca_mrmw.c index 773ac3f..1180b65 100644 --- a/drivers/infiniband/hw/ehca/ehca_mrmw.c +++ b/drivers/infiniband/hw/ehca/ehca_mrmw.c @@ -40,9 +40,8 @@ * POSSIBILITY OF SUCH DAMAGE. */ -#include rdma/ib_umem.h - #include asm/current.h +#include rdma/ib_umem.h #include ehca_iverbs.h #include ehca_mrmw.h -- 1.5.2 ___ general mailing list general@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
[PATCH 1/2] ehca: remove checkpatch.pl's warnings externs should be avoided in .c files
From b5d0336089b5ebe5b18acb94b2c94c2026cb95ee Mon Sep 17 00:00:00 2001
From: Hoang-Nam Nguyen hnguyen at de.ibm.com
Date: Fri, 27 Jul 2007 10:24:49 +0200
Subject: [PATCH] remove checkpatch.pl's warnings externs should be avoided in
.c files
Signed-off-by: Hoang-Nam Nguyen [EMAIL PROTECTED]
---
drivers/infiniband/hw/ehca/ehca_classes.h |1 +
drivers/infiniband/hw/ehca/ehca_mrmw.c|2 --
drivers/infiniband/hw/ehca/ehca_pd.c |1 -
drivers/infiniband/hw/ehca/hcp_if.c |1 -
drivers/infiniband/hw/ehca/ipz_pt_fn.h|2 ++
5 files changed, 3 insertions(+), 4 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_classes.h
b/drivers/infiniband/hw/ehca/ehca_classes.h
index 3725aa8..b5e9603 100644
--- a/drivers/infiniband/hw/ehca/ehca_classes.h
+++ b/drivers/infiniband/hw/ehca/ehca_classes.h
@@ -322,6 +322,7 @@ extern int ehca_static_rate;
extern int ehca_port_act_time;
extern int ehca_use_hp_mr;
extern int ehca_scaling_code;
+extern int ehca_mr_largepage;
struct ipzu_queue_resp {
u32 qe_size; /* queue entry size */
diff --git a/drivers/infiniband/hw/ehca/ehca_mrmw.c
b/drivers/infiniband/hw/ehca/ehca_mrmw.c
index c1b868b..773ac3f 100644
--- a/drivers/infiniband/hw/ehca/ehca_mrmw.c
+++ b/drivers/infiniband/hw/ehca/ehca_mrmw.c
@@ -64,8 +64,6 @@ enum ehca_mr_pgsize {
EHCA_MR_PGSIZE16M = 0x100L
};
-extern int ehca_mr_largepage;
-
static u32 ehca_encode_hwpage_size(u32 pgsize)
{
u32 idx = 0;
diff --git a/drivers/infiniband/hw/ehca/ehca_pd.c
b/drivers/infiniband/hw/ehca/ehca_pd.c
index 3dafd7f..43bcf08 100644
--- a/drivers/infiniband/hw/ehca/ehca_pd.c
+++ b/drivers/infiniband/hw/ehca/ehca_pd.c
@@ -88,7 +88,6 @@ int ehca_dealloc_pd(struct ib_pd *pd)
u32 cur_pid = current-tgid;
struct ehca_pd *my_pd = container_of(pd, struct ehca_pd, ib_pd);
int i, leftovers = 0;
- extern struct kmem_cache *small_qp_cache;
struct ipz_small_queue_page *page, *tmp;
if (my_pd-ib_pd.uobject my_pd-ib_pd.uobject-context
diff --git a/drivers/infiniband/hw/ehca/hcp_if.c
b/drivers/infiniband/hw/ehca/hcp_if.c
index fdbfebe..24f4541 100644
--- a/drivers/infiniband/hw/ehca/hcp_if.c
+++ b/drivers/infiniband/hw/ehca/hcp_if.c
@@ -758,7 +758,6 @@ u64 hipz_h_register_rpage_mr(const struct
ipz_adapter_handle adapter_handle,
const u64 logical_address_of_page,
const u64 count)
{
- extern int ehca_debug_level;
u64 ret;
if (unlikely(ehca_debug_level = 2)) {
diff --git a/drivers/infiniband/hw/ehca/ipz_pt_fn.h
b/drivers/infiniband/hw/ehca/ipz_pt_fn.h
index c6937a0..a801274 100644
--- a/drivers/infiniband/hw/ehca/ipz_pt_fn.h
+++ b/drivers/infiniband/hw/ehca/ipz_pt_fn.h
@@ -54,6 +54,8 @@
struct ehca_pd;
struct ipz_small_queue_page;
+extern struct kmem_cache *small_qp_cache;
+
/* struct generic ehca page */
struct ipz_page {
u8 entries[EHCA_PAGESIZE];
--
1.5.2
-
To unsubscribe from this list: send the line unsubscribe linux-kernel in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[PATCH 0/2] ehca: remove WARNING: externs should be avoided in .c files
Hello Roland! This small patch set fixes some coding-style related issues for ehca: [1/2] remove checkpatch.pl's warnings externs should be avoided in .c files [2/2] correction include order according kernel coding style Thanks Nam - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
[PATCH 2/2] ehca: correction include order according kernel coding style
From a2794450cbee597cefd7b6e159257583c459d358 Mon Sep 17 00:00:00 2001 From: Hoang-Nam Nguyen hnguyen at de.ibm.com Date: Fri, 27 Jul 2007 10:26:40 +0200 Subject: [PATCH] correction include order according kernel coding style Signed-off-by: Hoang-Nam Nguyen [EMAIL PROTECTED] --- drivers/infiniband/hw/ehca/ehca_mrmw.c |3 +-- 1 files changed, 1 insertions(+), 2 deletions(-) diff --git a/drivers/infiniband/hw/ehca/ehca_mrmw.c b/drivers/infiniband/hw/ehca/ehca_mrmw.c index 773ac3f..1180b65 100644 --- a/drivers/infiniband/hw/ehca/ehca_mrmw.c +++ b/drivers/infiniband/hw/ehca/ehca_mrmw.c @@ -40,9 +40,8 @@ * POSSIBILITY OF SUCH DAMAGE. */ -#include rdma/ib_umem.h - #include asm/current.h +#include rdma/ib_umem.h #include ehca_iverbs.h #include ehca_mrmw.h -- 1.5.2 - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
[PATCH 0/2] ehca: remove WARNING: externs should be avoided in .c files
Hello Roland! This small patch set fixes some coding-style related issues for ehca: [1/2] remove checkpatch.pl's warnings externs should be avoided in .c files [2/2] correction include order according kernel coding style Thanks Nam ___ Linuxppc-dev mailing list Linuxppc-dev@ozlabs.org https://ozlabs.org/mailman/listinfo/linuxppc-dev
[PATCH 1/2] ehca: remove checkpatch.pl's warnings externs should be avoided in .c files
From b5d0336089b5ebe5b18acb94b2c94c2026cb95ee Mon Sep 17 00:00:00 2001
From: Hoang-Nam Nguyen hnguyen at de.ibm.com
Date: Fri, 27 Jul 2007 10:24:49 +0200
Subject: [PATCH] remove checkpatch.pl's warnings externs should be avoided in
.c files
Signed-off-by: Hoang-Nam Nguyen [EMAIL PROTECTED]
---
drivers/infiniband/hw/ehca/ehca_classes.h |1 +
drivers/infiniband/hw/ehca/ehca_mrmw.c|2 --
drivers/infiniband/hw/ehca/ehca_pd.c |1 -
drivers/infiniband/hw/ehca/hcp_if.c |1 -
drivers/infiniband/hw/ehca/ipz_pt_fn.h|2 ++
5 files changed, 3 insertions(+), 4 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_classes.h
b/drivers/infiniband/hw/ehca/ehca_classes.h
index 3725aa8..b5e9603 100644
--- a/drivers/infiniband/hw/ehca/ehca_classes.h
+++ b/drivers/infiniband/hw/ehca/ehca_classes.h
@@ -322,6 +322,7 @@ extern int ehca_static_rate;
extern int ehca_port_act_time;
extern int ehca_use_hp_mr;
extern int ehca_scaling_code;
+extern int ehca_mr_largepage;
struct ipzu_queue_resp {
u32 qe_size; /* queue entry size */
diff --git a/drivers/infiniband/hw/ehca/ehca_mrmw.c
b/drivers/infiniband/hw/ehca/ehca_mrmw.c
index c1b868b..773ac3f 100644
--- a/drivers/infiniband/hw/ehca/ehca_mrmw.c
+++ b/drivers/infiniband/hw/ehca/ehca_mrmw.c
@@ -64,8 +64,6 @@ enum ehca_mr_pgsize {
EHCA_MR_PGSIZE16M = 0x100L
};
-extern int ehca_mr_largepage;
-
static u32 ehca_encode_hwpage_size(u32 pgsize)
{
u32 idx = 0;
diff --git a/drivers/infiniband/hw/ehca/ehca_pd.c
b/drivers/infiniband/hw/ehca/ehca_pd.c
index 3dafd7f..43bcf08 100644
--- a/drivers/infiniband/hw/ehca/ehca_pd.c
+++ b/drivers/infiniband/hw/ehca/ehca_pd.c
@@ -88,7 +88,6 @@ int ehca_dealloc_pd(struct ib_pd *pd)
u32 cur_pid = current-tgid;
struct ehca_pd *my_pd = container_of(pd, struct ehca_pd, ib_pd);
int i, leftovers = 0;
- extern struct kmem_cache *small_qp_cache;
struct ipz_small_queue_page *page, *tmp;
if (my_pd-ib_pd.uobject my_pd-ib_pd.uobject-context
diff --git a/drivers/infiniband/hw/ehca/hcp_if.c
b/drivers/infiniband/hw/ehca/hcp_if.c
index fdbfebe..24f4541 100644
--- a/drivers/infiniband/hw/ehca/hcp_if.c
+++ b/drivers/infiniband/hw/ehca/hcp_if.c
@@ -758,7 +758,6 @@ u64 hipz_h_register_rpage_mr(const struct
ipz_adapter_handle adapter_handle,
const u64 logical_address_of_page,
const u64 count)
{
- extern int ehca_debug_level;
u64 ret;
if (unlikely(ehca_debug_level = 2)) {
diff --git a/drivers/infiniband/hw/ehca/ipz_pt_fn.h
b/drivers/infiniband/hw/ehca/ipz_pt_fn.h
index c6937a0..a801274 100644
--- a/drivers/infiniband/hw/ehca/ipz_pt_fn.h
+++ b/drivers/infiniband/hw/ehca/ipz_pt_fn.h
@@ -54,6 +54,8 @@
struct ehca_pd;
struct ipz_small_queue_page;
+extern struct kmem_cache *small_qp_cache;
+
/* struct generic ehca page */
struct ipz_page {
u8 entries[EHCA_PAGESIZE];
--
1.5.2
___
Linuxppc-dev mailing list
Linuxppc-dev@ozlabs.org
https://ozlabs.org/mailman/listinfo/linuxppc-dev
[PATCH 2/2] ehca: correction include order according kernel coding style
From a2794450cbee597cefd7b6e159257583c459d358 Mon Sep 17 00:00:00 2001 From: Hoang-Nam Nguyen hnguyen at de.ibm.com Date: Fri, 27 Jul 2007 10:26:40 +0200 Subject: [PATCH] correction include order according kernel coding style Signed-off-by: Hoang-Nam Nguyen [EMAIL PROTECTED] --- drivers/infiniband/hw/ehca/ehca_mrmw.c |3 +-- 1 files changed, 1 insertions(+), 2 deletions(-) diff --git a/drivers/infiniband/hw/ehca/ehca_mrmw.c b/drivers/infiniband/hw/ehca/ehca_mrmw.c index 773ac3f..1180b65 100644 --- a/drivers/infiniband/hw/ehca/ehca_mrmw.c +++ b/drivers/infiniband/hw/ehca/ehca_mrmw.c @@ -40,9 +40,8 @@ * POSSIBILITY OF SUCH DAMAGE. */ -#include rdma/ib_umem.h - #include asm/current.h +#include rdma/ib_umem.h #include ehca_iverbs.h #include ehca_mrmw.h -- 1.5.2 ___ Linuxppc-dev mailing list Linuxppc-dev@ozlabs.org https://ozlabs.org/mailman/listinfo/linuxppc-dev
[ofa-general] Re: [PATCH ofed-1.2-rc3 2/4] ehca: backport for rhel-4.5 - mmap functonality
Hi Michael,
Below is the version without conflicts. And it should compile.
As soon as the build scripts are ready, I'll test the whole backport.
Thanks
Nam
From 6fa28219914394064a49c34030a09e23d160231c Mon Sep 17 00:00:00 2001
From: [EMAIL PROTECTED]
Date: Wed, 25 Jul 2007 17:16:53 +0200
Subject: [PATCH ofed-1.3-alpha] ehca: backport_ehca_2_rhel45_umap.patch
---
drivers/infiniband/hw/ehca/ehca_classes.h | 29 ++-
drivers/infiniband/hw/ehca/ehca_cq.c | 66 -
drivers/infiniband/hw/ehca/ehca_iverbs.h |8 +
drivers/infiniband/hw/ehca/ehca_qp.c | 92 +--
drivers/infiniband/hw/ehca/ehca_uverbs.c | 423 +
5 files changed, 395 insertions(+), 223 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_classes.h
b/drivers/infiniband/hw/ehca/ehca_classes.h
index 3725aa8..49d6155 100644
--- a/drivers/infiniband/hw/ehca/ehca_classes.h
+++ b/drivers/infiniband/hw/ehca/ehca_classes.h
@@ -160,14 +160,13 @@ struct ehca_qp {
struct ipz_qp_handle ipz_qp_handle;
struct ehca_pfqp pf;
struct ib_qp_init_attr init_attr;
+ u64 uspace_squeue;
+ u64 uspace_rqueue;
+ u64 uspace_fwh;
struct ehca_cq *send_cq;
struct ehca_cq *recv_cq;
unsigned int sqerr_purgeflag;
struct hlist_node list_entries;
- /* mmap counter for resources mapped into user space */
- u32 mm_count_squeue;
- u32 mm_count_rqueue;
- u32 mm_count_galpa;
};
#define IS_SRQ(qp) (qp-ext_type == EQPT_SRQ)
@@ -188,6 +187,8 @@ struct ehca_cq {
struct ipz_cq_handle ipz_cq_handle;
struct ehca_pfcq pf;
spinlock_t cb_lock;
+ u64 uspace_queue;
+ u64 uspace_fwh;
struct hlist_head qp_hashtab[QP_HASHTAB_LEN];
struct list_head entry;
u32 nr_callbacks; /* #events assigned to cpu by scaling code */
@@ -195,9 +196,6 @@ struct ehca_cq {
wait_queue_head_t wait_completion;
spinlock_t task_lock;
u32 ownpid;
- /* mmap counter for resources mapped into user space */
- u32 mm_count_queue;
- u32 mm_count_galpa;
};
enum ehca_mr_flag {
@@ -300,6 +298,20 @@ struct ehca_ucontext {
struct ib_ucontext ib_ucontext;
};
+struct ehca_module *ehca_module_new(void);
+
+int ehca_module_delete(struct ehca_module *me);
+
+int ehca_eq_ctor(struct ehca_eq *eq);
+
+int ehca_eq_dtor(struct ehca_eq *eq);
+
+struct ehca_shca *ehca_shca_new(void);
+
+int ehca_shca_delete(struct ehca_shca *me);
+
+struct ehca_sport *ehca_sport_new(struct ehca_shca *anchor);
+
int ehca_init_pd_cache(void);
void ehca_cleanup_pd_cache(void);
int ehca_init_cq_cache(void);
@@ -324,6 +336,7 @@ extern int ehca_use_hp_mr;
extern int ehca_scaling_code;
struct ipzu_queue_resp {
+ u64 queue;/* points to first queue entry */
u32 qe_size; /* queue entry size */
u32 act_nr_of_sg;
u32 queue_length; /* queue length allocated in bytes */
@@ -336,6 +349,7 @@ struct ehca_create_cq_resp {
u32 cq_number;
u32 token;
struct ipzu_queue_resp ipz_queue;
+ struct h_galpas galpas;
};
struct ehca_create_qp_resp {
@@ -349,6 +363,7 @@ struct ehca_create_qp_resp {
u32 dummy; /* padding for 8 byte alignment */
struct ipzu_queue_resp ipz_squeue;
struct ipzu_queue_resp ipz_rqueue;
+ struct h_galpas galpas;
};
struct ehca_alloc_cq_parms {
diff --git a/drivers/infiniband/hw/ehca/ehca_cq.c
b/drivers/infiniband/hw/ehca/ehca_cq.c
index 9c7172b..ac0bb10 100644
--- a/drivers/infiniband/hw/ehca/ehca_cq.c
+++ b/drivers/infiniband/hw/ehca/ehca_cq.c
@@ -268,6 +268,7 @@ struct ib_cq *ehca_create_cq(struct ib_device *device, int
cqe, int comp_vector,
if (context) {
struct ipz_queue *ipz_queue = my_cq-ipz_queue;
struct ehca_create_cq_resp resp;
+ struct vm_area_struct *vma;
memset(resp, 0, sizeof(resp));
resp.cq_number = my_cq-cq_number;
resp.token = my_cq-token;
@@ -276,14 +277,40 @@ struct ib_cq *ehca_create_cq(struct ib_device *device,
int cqe, int comp_vector,
resp.ipz_queue.queue_length = ipz_queue-queue_length;
resp.ipz_queue.pagesize = ipz_queue-pagesize;
resp.ipz_queue.toggle_state = ipz_queue-toggle_state;
+ ret = ehca_mmap_nopage(((u64)(my_cq-token) 32) | 0x1200,
+ ipz_queue-queue_length,
+ (void**)resp.ipz_queue.queue,
+ vma);
+ if (ret) {
+ ehca_err(device, Could not mmap queue pages);
+ cq = ERR_PTR(ret);
+ goto create_cq_exit4;
+ }
+ my_cq-uspace_queue = resp.ipz_queue.queue;
+ resp.galpas = my_cq-galpas;
+ ret = ehca_mmap_register(my_cq-galpas.user.fw_handle,
[ewg] Re: [PATCH ofed-1.2-rc3 2/4] ehca: backport for rhel-4.5 - mmap functonality
Hi Michael,
Below is the version without conflicts. And it should compile.
As soon as the build scripts are ready, I'll test the whole backport.
Thanks
Nam
From 6fa28219914394064a49c34030a09e23d160231c Mon Sep 17 00:00:00 2001
From: [EMAIL PROTECTED]
Date: Wed, 25 Jul 2007 17:16:53 +0200
Subject: [PATCH ofed-1.3-alpha] ehca: backport_ehca_2_rhel45_umap.patch
---
drivers/infiniband/hw/ehca/ehca_classes.h | 29 ++-
drivers/infiniband/hw/ehca/ehca_cq.c | 66 -
drivers/infiniband/hw/ehca/ehca_iverbs.h |8 +
drivers/infiniband/hw/ehca/ehca_qp.c | 92 +--
drivers/infiniband/hw/ehca/ehca_uverbs.c | 423 +
5 files changed, 395 insertions(+), 223 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_classes.h
b/drivers/infiniband/hw/ehca/ehca_classes.h
index 3725aa8..49d6155 100644
--- a/drivers/infiniband/hw/ehca/ehca_classes.h
+++ b/drivers/infiniband/hw/ehca/ehca_classes.h
@@ -160,14 +160,13 @@ struct ehca_qp {
struct ipz_qp_handle ipz_qp_handle;
struct ehca_pfqp pf;
struct ib_qp_init_attr init_attr;
+ u64 uspace_squeue;
+ u64 uspace_rqueue;
+ u64 uspace_fwh;
struct ehca_cq *send_cq;
struct ehca_cq *recv_cq;
unsigned int sqerr_purgeflag;
struct hlist_node list_entries;
- /* mmap counter for resources mapped into user space */
- u32 mm_count_squeue;
- u32 mm_count_rqueue;
- u32 mm_count_galpa;
};
#define IS_SRQ(qp) (qp-ext_type == EQPT_SRQ)
@@ -188,6 +187,8 @@ struct ehca_cq {
struct ipz_cq_handle ipz_cq_handle;
struct ehca_pfcq pf;
spinlock_t cb_lock;
+ u64 uspace_queue;
+ u64 uspace_fwh;
struct hlist_head qp_hashtab[QP_HASHTAB_LEN];
struct list_head entry;
u32 nr_callbacks; /* #events assigned to cpu by scaling code */
@@ -195,9 +196,6 @@ struct ehca_cq {
wait_queue_head_t wait_completion;
spinlock_t task_lock;
u32 ownpid;
- /* mmap counter for resources mapped into user space */
- u32 mm_count_queue;
- u32 mm_count_galpa;
};
enum ehca_mr_flag {
@@ -300,6 +298,20 @@ struct ehca_ucontext {
struct ib_ucontext ib_ucontext;
};
+struct ehca_module *ehca_module_new(void);
+
+int ehca_module_delete(struct ehca_module *me);
+
+int ehca_eq_ctor(struct ehca_eq *eq);
+
+int ehca_eq_dtor(struct ehca_eq *eq);
+
+struct ehca_shca *ehca_shca_new(void);
+
+int ehca_shca_delete(struct ehca_shca *me);
+
+struct ehca_sport *ehca_sport_new(struct ehca_shca *anchor);
+
int ehca_init_pd_cache(void);
void ehca_cleanup_pd_cache(void);
int ehca_init_cq_cache(void);
@@ -324,6 +336,7 @@ extern int ehca_use_hp_mr;
extern int ehca_scaling_code;
struct ipzu_queue_resp {
+ u64 queue;/* points to first queue entry */
u32 qe_size; /* queue entry size */
u32 act_nr_of_sg;
u32 queue_length; /* queue length allocated in bytes */
@@ -336,6 +349,7 @@ struct ehca_create_cq_resp {
u32 cq_number;
u32 token;
struct ipzu_queue_resp ipz_queue;
+ struct h_galpas galpas;
};
struct ehca_create_qp_resp {
@@ -349,6 +363,7 @@ struct ehca_create_qp_resp {
u32 dummy; /* padding for 8 byte alignment */
struct ipzu_queue_resp ipz_squeue;
struct ipzu_queue_resp ipz_rqueue;
+ struct h_galpas galpas;
};
struct ehca_alloc_cq_parms {
diff --git a/drivers/infiniband/hw/ehca/ehca_cq.c
b/drivers/infiniband/hw/ehca/ehca_cq.c
index 9c7172b..ac0bb10 100644
--- a/drivers/infiniband/hw/ehca/ehca_cq.c
+++ b/drivers/infiniband/hw/ehca/ehca_cq.c
@@ -268,6 +268,7 @@ struct ib_cq *ehca_create_cq(struct ib_device *device, int
cqe, int comp_vector,
if (context) {
struct ipz_queue *ipz_queue = my_cq-ipz_queue;
struct ehca_create_cq_resp resp;
+ struct vm_area_struct *vma;
memset(resp, 0, sizeof(resp));
resp.cq_number = my_cq-cq_number;
resp.token = my_cq-token;
@@ -276,14 +277,40 @@ struct ib_cq *ehca_create_cq(struct ib_device *device,
int cqe, int comp_vector,
resp.ipz_queue.queue_length = ipz_queue-queue_length;
resp.ipz_queue.pagesize = ipz_queue-pagesize;
resp.ipz_queue.toggle_state = ipz_queue-toggle_state;
+ ret = ehca_mmap_nopage(((u64)(my_cq-token) 32) | 0x1200,
+ ipz_queue-queue_length,
+ (void**)resp.ipz_queue.queue,
+ vma);
+ if (ret) {
+ ehca_err(device, Could not mmap queue pages);
+ cq = ERR_PTR(ret);
+ goto create_cq_exit4;
+ }
+ my_cq-uspace_queue = resp.ipz_queue.queue;
+ resp.galpas = my_cq-galpas;
+ ret = ehca_mmap_register(my_cq-galpas.user.fw_handle,
Re: [ofa-general] [PATCH 1/5] ehca: Supports large page MRs
Hi Roland! > I applied this, but I agree with checkpatch.pl: > > > WARNING: externs should be avoided in .c files > > #227: FILE: drivers/infiniband/hw/ehca/ehca_mrmw.c:67: > > +extern int ehca_mr_largepage; > > > > WARNING: externs should be avoided in .c files > > #949: FILE: drivers/infiniband/hw/ehca/hcp_if.c:753: > > + extern int ehca_debug_level; > > if you need to use a variable in more than one .c file, put the extern > declaration in a common header that's included everywhere you use the > variable, including the .c file that it is defined in. That way the > compiler can see if you get confused about the type of the variable. That's true. > When you get a chance, please post a follow-on patch to fix this. Sure thing. Will do that for rc2. Thanks! Nam - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [ofa-general] [PATCH 1/5] ehca: Supports large page MRs
Hi Roland! I applied this, but I agree with checkpatch.pl: WARNING: externs should be avoided in .c files #227: FILE: drivers/infiniband/hw/ehca/ehca_mrmw.c:67: +extern int ehca_mr_largepage; WARNING: externs should be avoided in .c files #949: FILE: drivers/infiniband/hw/ehca/hcp_if.c:753: + extern int ehca_debug_level; if you need to use a variable in more than one .c file, put the extern declaration in a common header that's included everywhere you use the variable, including the .c file that it is defined in. That way the compiler can see if you get confused about the type of the variable. That's true. When you get a chance, please post a follow-on patch to fix this. Sure thing. Will do that for rc2. Thanks! Nam - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [ofa-general] [PATCH 1/5] ehca: Supports large page MRs
Hi Roland! I applied this, but I agree with checkpatch.pl: WARNING: externs should be avoided in .c files #227: FILE: drivers/infiniband/hw/ehca/ehca_mrmw.c:67: +extern int ehca_mr_largepage; WARNING: externs should be avoided in .c files #949: FILE: drivers/infiniband/hw/ehca/hcp_if.c:753: + extern int ehca_debug_level; if you need to use a variable in more than one .c file, put the extern declaration in a common header that's included everywhere you use the variable, including the .c file that it is defined in. That way the compiler can see if you get confused about the type of the variable. That's true. When you get a chance, please post a follow-on patch to fix this. Sure thing. Will do that for rc2. Thanks! Nam ___ Linuxppc-dev mailing list Linuxppc-dev@ozlabs.org https://ozlabs.org/mailman/listinfo/linuxppc-dev
[PATCH 5/5] ehca: Support small QP queues
From: Stefan Roscher
Date: Fri, 20 Jul 2007 13:59:14 +0200
Subject: [PATCH 5/5] IB/ehca: Small QP queues
eHCA2 supports QP queues that can be as small as 512 bytes. This greatly
reduces memory overhead for consumers that use lots of QPs with small queues
(e.g. RDMA-only QPs). Apart from dealing with firmware, this code needs to
manage bite-sized chunks of kernel pages, making sure that no kernel page is
shared between different protection domains.
Signed-off-by: Hoang-Nam Nguyen <[EMAIL PROTECTED]>
---
drivers/infiniband/hw/ehca/ehca_classes.h | 41 --
drivers/infiniband/hw/ehca/ehca_cq.c |8 +-
drivers/infiniband/hw/ehca/ehca_eq.c |8 +-
drivers/infiniband/hw/ehca/ehca_main.c| 14 ++-
drivers/infiniband/hw/ehca/ehca_pd.c | 25 +++-
drivers/infiniband/hw/ehca/ehca_qp.c | 163 +-
drivers/infiniband/hw/ehca/ehca_uverbs.c |2 +-
drivers/infiniband/hw/ehca/hcp_if.c | 30 +++--
drivers/infiniband/hw/ehca/ipz_pt_fn.c| 222 ++---
drivers/infiniband/hw/ehca/ipz_pt_fn.h| 26 +++-
10 files changed, 379 insertions(+), 160 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_classes.h
b/drivers/infiniband/hw/ehca/ehca_classes.h
index 63b8b9f..3725aa8 100644
--- a/drivers/infiniband/hw/ehca/ehca_classes.h
+++ b/drivers/infiniband/hw/ehca/ehca_classes.h
@@ -43,7 +43,6 @@
#ifndef __EHCA_CLASSES_H__
#define __EHCA_CLASSES_H__
-
struct ehca_module;
struct ehca_qp;
struct ehca_cq;
@@ -129,6 +128,10 @@ struct ehca_pd {
struct ib_pd ib_pd;
struct ipz_pd fw_pd;
u32 ownpid;
+ /* small queue mgmt */
+ struct mutex lock;
+ struct list_head free[2];
+ struct list_head full[2];
};
enum ehca_ext_qp_type {
@@ -307,6 +310,8 @@ int ehca_init_av_cache(void);
void ehca_cleanup_av_cache(void);
int ehca_init_mrmw_cache(void);
void ehca_cleanup_mrmw_cache(void);
+int ehca_init_small_qp_cache(void);
+void ehca_cleanup_small_qp_cache(void);
extern rwlock_t ehca_qp_idr_lock;
extern rwlock_t ehca_cq_idr_lock;
@@ -324,7 +329,7 @@ struct ipzu_queue_resp {
u32 queue_length; /* queue length allocated in bytes */
u32 pagesize;
u32 toggle_state;
- u32 dummy; /* padding for 8 byte alignment */
+ u32 offset; /* save offset within a page for small_qp */
};
struct ehca_create_cq_resp {
@@ -366,15 +371,29 @@ enum ehca_ll_comp_flags {
LLQP_COMP_MASK = 0x60,
};
+struct ehca_alloc_queue_parms {
+ /* input parameters */
+ int max_wr;
+ int max_sge;
+ int page_size;
+ int is_small;
+
+ /* output parameters */
+ u16 act_nr_wqes;
+ u8 act_nr_sges;
+ u32 queue_size; /* bytes for small queues, pages otherwise */
+};
+
struct ehca_alloc_qp_parms {
-/* input parameters */
+ struct ehca_alloc_queue_parms squeue;
+ struct ehca_alloc_queue_parms rqueue;
+
+ /* input parameters */
enum ehca_service_type servicetype;
+ int qp_storage;
int sigtype;
enum ehca_ext_qp_type ext_type;
enum ehca_ll_comp_flags ll_comp_flags;
-
- int max_send_wr, max_recv_wr;
- int max_send_sge, max_recv_sge;
int ud_av_l_key_ctl;
u32 token;
@@ -384,18 +403,10 @@ struct ehca_alloc_qp_parms {
u32 srq_qpn, srq_token, srq_limit;
-/* output parameters */
+ /* output parameters */
u32 real_qp_num;
struct ipz_qp_handle qp_handle;
struct h_galpas galpas;
-
- u16 act_nr_send_wqes;
- u16 act_nr_recv_wqes;
- u8 act_nr_recv_sges;
- u8 act_nr_send_sges;
-
- u32 nr_rq_pages;
- u32 nr_sq_pages;
};
int ehca_cq_assign_qp(struct ehca_cq *cq, struct ehca_qp *qp);
diff --git a/drivers/infiniband/hw/ehca/ehca_cq.c
b/drivers/infiniband/hw/ehca/ehca_cq.c
index 9e87883..5746787 100644
--- a/drivers/infiniband/hw/ehca/ehca_cq.c
+++ b/drivers/infiniband/hw/ehca/ehca_cq.c
@@ -190,8 +190,8 @@ struct ib_cq *ehca_create_cq(struct ib_device *device, int
cqe, int comp_vector,
goto create_cq_exit2;
}
- ipz_rc = ipz_queue_ctor(_cq->ipz_queue, param.act_pages,
- EHCA_PAGESIZE, sizeof(struct ehca_cqe), 0);
+ ipz_rc = ipz_queue_ctor(NULL, _cq->ipz_queue, param.act_pages,
+ EHCA_PAGESIZE, sizeof(struct ehca_cqe), 0, 0);
if (!ipz_rc) {
ehca_err(device, "ipz_queue_ctor() failed ipz_rc=%x device=%p",
ipz_rc, device);
@@ -285,7 +285,7 @@ struct ib_cq *ehca_create_cq(struct ib_device *device, int
cqe, int comp_vector,
return cq;
create_cq_exit4:
- ipz_queue_dtor(_cq->ipz_queue);
+ ipz_queue_dtor(NULL, _cq->ipz_queue);
create_cq_exit3:
h_ret = hipz_h_destroy_cq(adapter_handle, my_cq, 1);
@@ -359,7 +359,7 @@ int ehca_destroy_cq(struct ib_cq *cq)
&
[PATCH 2/5] ehca: Generate event when SRQ limit reached
From: Joachim Fenkes <[EMAIL PROTECTED]>
Date: Thu, 19 Jul 2007 20:51:43 +0200
Subject: [PATCH 2/5] IB/ehca: Generate event when SRQ limit reached
Signed-off-by: Joachim Fenkes <[EMAIL PROTECTED]>
---
drivers/infiniband/hw/ehca/ehca_irq.c | 42 ++---
1 files changed, 28 insertions(+), 14 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_irq.c
b/drivers/infiniband/hw/ehca/ehca_irq.c
index 4fb01fc..71c0799 100644
--- a/drivers/infiniband/hw/ehca/ehca_irq.c
+++ b/drivers/infiniband/hw/ehca/ehca_irq.c
@@ -175,9 +175,8 @@ error_data1:
}
-static void qp_event_callback(struct ehca_shca *shca,
- u64 eqe,
- enum ib_event_type event_type)
+static void qp_event_callback(struct ehca_shca *shca, u64 eqe,
+ enum ib_event_type event_type, int fatal)
{
struct ib_event event;
struct ehca_qp *qp;
@@ -191,16 +190,26 @@ static void qp_event_callback(struct ehca_shca *shca,
if (!qp)
return;
- ehca_error_data(shca, qp, qp->ipz_qp_handle.handle);
+ if (fatal)
+ ehca_error_data(shca, qp, qp->ipz_qp_handle.handle);
- if (!qp->ib_qp.event_handler)
- return;
+ event.device = >ib_device;
- event.device = >ib_device;
- event.event = event_type;
- event.element.qp = >ib_qp;
+ if (qp->ext_type == EQPT_SRQ) {
+ if (!qp->ib_srq.event_handler)
+ return;
- qp->ib_qp.event_handler(, qp->ib_qp.qp_context);
+ event.event = fatal ? IB_EVENT_SRQ_ERR : event_type;
+ event.element.srq = >ib_srq;
+ qp->ib_srq.event_handler(, qp->ib_srq.srq_context);
+ } else {
+ if (!qp->ib_qp.event_handler)
+ return;
+
+ event.event = event_type;
+ event.element.qp = >ib_qp;
+ qp->ib_qp.event_handler(, qp->ib_qp.qp_context);
+ }
return;
}
@@ -234,17 +243,17 @@ static void parse_identifier(struct ehca_shca *shca, u64
eqe)
switch (identifier) {
case 0x02: /* path migrated */
- qp_event_callback(shca, eqe, IB_EVENT_PATH_MIG);
+ qp_event_callback(shca, eqe, IB_EVENT_PATH_MIG, 0);
break;
case 0x03: /* communication established */
- qp_event_callback(shca, eqe, IB_EVENT_COMM_EST);
+ qp_event_callback(shca, eqe, IB_EVENT_COMM_EST, 0);
break;
case 0x04: /* send queue drained */
- qp_event_callback(shca, eqe, IB_EVENT_SQ_DRAINED);
+ qp_event_callback(shca, eqe, IB_EVENT_SQ_DRAINED, 0);
break;
case 0x05: /* QP error */
case 0x06: /* QP error */
- qp_event_callback(shca, eqe, IB_EVENT_QP_FATAL);
+ qp_event_callback(shca, eqe, IB_EVENT_QP_FATAL, 1);
break;
case 0x07: /* CQ error */
case 0x08: /* CQ error */
@@ -278,6 +287,11 @@ static void parse_identifier(struct ehca_shca *shca, u64
eqe)
ehca_err(>ib_device, "Interface trace stopped.");
break;
case 0x14: /* first error capture info available */
+ ehca_info(>ib_device, "First error capture available");
+ break;
+ case 0x15: /* SRQ limit reached */
+ qp_event_callback(shca, eqe, IB_EVENT_SRQ_LIMIT_REACHED, 0);
+ break;
default:
ehca_err(>ib_device, "Unknown identifier: %x on %s.",
identifier, shca->ib_device.name);
--
1.5.2
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[PATCH 4/5] ehca: Make internal_create/destroy_qp() static
From: Joachim Fenkes <[EMAIL PROTECTED]>
Date: Thu, 19 Jul 2007 21:40:00 +0200
Subject: [PATCH 4/5] IB/ehca: Make internal_{create,destroy}_qp() static
They're only used in ehca_qp.c
Signed-off-by: Joachim Fenkes <[EMAIL PROTECTED]>
---
drivers/infiniband/hw/ehca/ehca_qp.c | 17 +
1 files changed, 9 insertions(+), 8 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_qp.c
b/drivers/infiniband/hw/ehca/ehca_qp.c
index 48e9cea..b916d9c 100644
--- a/drivers/infiniband/hw/ehca/ehca_qp.c
+++ b/drivers/infiniband/hw/ehca/ehca_qp.c
@@ -363,10 +363,11 @@ init_qp_queue1:
* the value of the is_srq parameter. If init_attr and srq_init_attr share
* fields, the field out of init_attr is used.
*/
-struct ehca_qp *internal_create_qp(struct ib_pd *pd,
- struct ib_qp_init_attr *init_attr,
- struct ib_srq_init_attr *srq_init_attr,
- struct ib_udata *udata, int is_srq)
+static struct ehca_qp *internal_create_qp(
+ struct ib_pd *pd,
+ struct ib_qp_init_attr *init_attr,
+ struct ib_srq_init_attr *srq_init_attr,
+ struct ib_udata *udata, int is_srq)
{
struct ehca_qp *my_qp;
struct ehca_pd *my_pd = container_of(pd, struct ehca_pd, ib_pd);
@@ -752,8 +753,8 @@ struct ib_qp *ehca_create_qp(struct ib_pd *pd,
return IS_ERR(ret) ? (struct ib_qp *)ret : >ib_qp;
}
-int internal_destroy_qp(struct ib_device *dev, struct ehca_qp *my_qp,
- struct ib_uobject *uobject);
+static int internal_destroy_qp(struct ib_device *dev, struct ehca_qp *my_qp,
+ struct ib_uobject *uobject);
struct ib_srq *ehca_create_srq(struct ib_pd *pd,
struct ib_srq_init_attr *srq_init_attr,
@@ -1669,8 +1670,8 @@ query_srq_exit1:
return ret;
}
-int internal_destroy_qp(struct ib_device *dev, struct ehca_qp *my_qp,
- struct ib_uobject *uobject)
+static int internal_destroy_qp(struct ib_device *dev, struct ehca_qp *my_qp,
+ struct ib_uobject *uobject)
{
struct ehca_shca *shca = container_of(dev, struct ehca_shca, ib_device);
struct ehca_pd *my_pd = container_of(my_qp->ib_qp.pd, struct ehca_pd,
--
1.5.2
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[PATCH 3/5] ehca: Make ehca2ib_return_code() non-inline
From: Joachim Fenkes <[EMAIL PROTECTED]>
Date: Thu, 19 Jul 2007 21:13:57 +0200
Subject: [PATCH 3/5] IB/ehca: Make ehca2ib_return_code() non-inline
It's nowhere in the main path and making it non-inline saves ~1.5K of code.
Signed-off-by: Joachim Fenkes <[EMAIL PROTECTED]>
---
drivers/infiniband/hw/ehca/ehca_main.c | 17 +
drivers/infiniband/hw/ehca/ehca_tools.h | 19 +--
2 files changed, 18 insertions(+), 18 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_main.c
b/drivers/infiniband/hw/ehca/ehca_main.c
index 34661c3..3bd7afb 100644
--- a/drivers/infiniband/hw/ehca/ehca_main.c
+++ b/drivers/infiniband/hw/ehca/ehca_main.c
@@ -130,6 +130,23 @@ void ehca_free_fw_ctrlblock(void *ptr)
}
#endif
+int ehca2ib_return_code(u64 ehca_rc)
+{
+ switch (ehca_rc) {
+ case H_SUCCESS:
+ return 0;
+ case H_RESOURCE: /* Resource in use */
+ case H_BUSY:
+ return -EBUSY;
+ case H_NOT_ENOUGH_RESOURCES: /* insufficient resources */
+ case H_CONSTRAINED: /* resource constraint */
+ case H_NO_MEM:
+ return -ENOMEM;
+ default:
+ return -EINVAL;
+ }
+}
+
static int ehca_create_slab_caches(void)
{
int ret;
diff --git a/drivers/infiniband/hw/ehca/ehca_tools.h
b/drivers/infiniband/hw/ehca/ehca_tools.h
index 678b813..57c77a7 100644
--- a/drivers/infiniband/hw/ehca/ehca_tools.h
+++ b/drivers/infiniband/hw/ehca/ehca_tools.h
@@ -154,24 +154,7 @@ extern int ehca_debug_level;
#define EHCA_BMASK_GET(mask, value) \
(EHCA_BMASK_MASK(mask) & (((u64)(value)) >> EHCA_BMASK_SHIFTPOS(mask)))
-
/* Converts ehca to ib return code */
-static inline int ehca2ib_return_code(u64 ehca_rc)
-{
- switch (ehca_rc) {
- case H_SUCCESS:
- return 0;
- case H_RESOURCE: /* Resource in use */
- case H_BUSY:
- return -EBUSY;
- case H_NOT_ENOUGH_RESOURCES: /* insufficient resources */
- case H_CONSTRAINED: /* resource constraint */
- case H_NO_MEM:
- return -ENOMEM;
- default:
- return -EINVAL;
- }
-}
-
+int ehca2ib_return_code(u64 ehca_rc);
#endif /* EHCA_TOOLS_H */
--
1.5.2
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[PATCH 1/5] ehca: Supports large page MRs
From: Hoang-Nam Nguyen <[EMAIL PROTECTED]>
Date: Thu, 19 Jul 2007 20:48:04 +0200
Subject: [PATCH 1/5] IB/ehca: Support large page MRs
Add support for MR pages larger than 4K on eHCA2. This reduces firmware
memory consumption. If enabled via the mr_largepage module parameter, the MR
page size will be determined based on the MR length and the hardware
capabilities - if the MR is >= 16M, 16M pages are used, for example.
Signed-off-by: Joachim Fenkes <[EMAIL PROTECTED]>
---
drivers/infiniband/hw/ehca/ehca_classes.h |9 +
drivers/infiniband/hw/ehca/ehca_main.c| 18 ++-
drivers/infiniband/hw/ehca/ehca_mrmw.c| 371 -
drivers/infiniband/hw/ehca/ehca_mrmw.h|2 +-
drivers/infiniband/hw/ehca/hcp_if.c | 20 ++-
5 files changed, 357 insertions(+), 63 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_classes.h
b/drivers/infiniband/hw/ehca/ehca_classes.h
index 043e4fb..63b8b9f 100644
--- a/drivers/infiniband/hw/ehca/ehca_classes.h
+++ b/drivers/infiniband/hw/ehca/ehca_classes.h
@@ -100,6 +100,11 @@ struct ehca_sport {
struct ehca_sma_attr saved_attr;
};
+#define HCA_CAP_MR_PGSIZE_4K 1
+#define HCA_CAP_MR_PGSIZE_64K 2
+#define HCA_CAP_MR_PGSIZE_1M 4
+#define HCA_CAP_MR_PGSIZE_16M 8
+
struct ehca_shca {
struct ib_device ib_device;
struct ibmebus_dev *ibmebus_dev;
@@ -115,6 +120,8 @@ struct ehca_shca {
struct h_galpas galpas;
struct mutex modify_mutex;
u64 hca_cap;
+ /* MR pgsize: bit 0-3 means 4K, 64K, 1M, 16M respectively */
+ u32 hca_cap_mr_pgsize;
int max_mtu;
};
@@ -206,6 +213,7 @@ struct ehca_mr {
enum ehca_mr_flag flags;
u32 num_kpages; /* number of kernel pages */
u32 num_hwpages;/* number of hw pages to form MR */
+ u64 hwpage_size;/* hw page size used for this MR */
int acl;/* ACL (stored here for usage in reregister) */
u64 *start; /* virtual start address (stored here for */
/* usage in reregister) */
@@ -240,6 +248,7 @@ struct ehca_mr_pginfo {
enum ehca_mr_pgi_type type;
u64 num_kpages;
u64 kpage_cnt;
+ u64 hwpage_size; /* hw page size used for this MR */
u64 num_hwpages; /* number of hw pages */
u64 hwpage_cnt; /* counter for hw pages */
u64 next_hwpage; /* next hw page in buffer/chunk/listelem */
diff --git a/drivers/infiniband/hw/ehca/ehca_main.c
b/drivers/infiniband/hw/ehca/ehca_main.c
index 36377c6..34661c3 100644
--- a/drivers/infiniband/hw/ehca/ehca_main.c
+++ b/drivers/infiniband/hw/ehca/ehca_main.c
@@ -63,6 +63,7 @@ int ehca_port_act_time = 30;
int ehca_poll_all_eqs = 1;
int ehca_static_rate = -1;
int ehca_scaling_code = 0;
+int ehca_mr_largepage = 0;
module_param_named(open_aqp1, ehca_open_aqp1, int, 0);
module_param_named(debug_level, ehca_debug_level, int, 0);
@@ -72,7 +73,8 @@ module_param_named(use_hp_mr, ehca_use_hp_mr, int, 0);
module_param_named(port_act_time, ehca_port_act_time, int, 0);
module_param_named(poll_all_eqs, ehca_poll_all_eqs, int, 0);
module_param_named(static_rate, ehca_static_rate, int, 0);
-module_param_named(scaling_code, ehca_scaling_code, int, 0);
+module_param_named(scaling_code, ehca_scaling_code, int, 0);
+module_param_named(mr_largepage, ehca_mr_largepage, int, 0);
MODULE_PARM_DESC(open_aqp1,
"AQP1 on startup (0: no (default), 1: yes)");
@@ -95,6 +97,9 @@ MODULE_PARM_DESC(static_rate,
"set permanent static rate (default: disabled)");
MODULE_PARM_DESC(scaling_code,
"set scaling code (0: disabled/default, 1: enabled)");
+MODULE_PARM_DESC(mr_largepage,
+"use large page for MR (0: use PAGE_SIZE (default), "
+"1: use large page depending on MR size");
DEFINE_RWLOCK(ehca_qp_idr_lock);
DEFINE_RWLOCK(ehca_cq_idr_lock);
@@ -295,6 +300,8 @@ int ehca_sense_attributes(struct ehca_shca *shca)
if (EHCA_BMASK_GET(hca_cap_descr[i].mask, shca->hca_cap))
ehca_gen_dbg(" %s", hca_cap_descr[i].descr);
+ shca->hca_cap_mr_pgsize = rblock->memory_page_size_supported;
+
port = (struct hipz_query_port *)rblock;
h_ret = hipz_h_query_port(shca->ipz_hca_handle, 1, port);
if (h_ret != H_SUCCESS) {
@@ -590,6 +597,14 @@ static ssize_t ehca_show_adapter_handle(struct device *dev,
}
static DEVICE_ATTR(adapter_handle, S_IRUGO, ehca_show_adapter_handle, NULL);
+static ssize_t ehca_show_mr_largepage(struct device *dev,
+ struct device_attribute *attr,
+ char *buf)
+{
+ return sprintf(buf, "%d\n", ehca_mr_largepage);
+}
+static DEVICE_ATTR
[PATCH 0/5] ehca: MR large page, small queue and fixes
Here is a patch set against Roland's git, branch for-2.6.23 for ehca. It adds support for MR large page and small queues. In addition of that it also contains various small fixes from previous comments and what we found. They are in details: [1/5] adds support for MR large page [2/5] generates event when SRQ limit reached [3/5] makes ehca2ib_return_code() non inline [4/5] makes internal_create/destroy_qp() static [5/5] adds support for small queues The patches should apply cleanly, in order, against Roland's git. Please review the changes and apply the patches if they are okay. Regards, Nam & Stefan - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
[ofa-general] [PATCH 0/5] ehca: MR large page, small queue and fixes
Here is a patch set against Roland's git, branch for-2.6.23 for ehca. It adds support for MR large page and small queues. In addition of that it also contains various small fixes from previous comments and what we found. They are in details: [1/5] adds support for MR large page [2/5] generates event when SRQ limit reached [3/5] makes ehca2ib_return_code() non inline [4/5] makes internal_create/destroy_qp() static [5/5] adds support for small queues The patches should apply cleanly, in order, against Roland's git. Please review the changes and apply the patches if they are okay. Regards, Nam Stefan ___ general mailing list general@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
[ofa-general] [PATCH 1/5] ehca: Supports large page MRs
From: Hoang-Nam Nguyen [EMAIL PROTECTED]
Date: Thu, 19 Jul 2007 20:48:04 +0200
Subject: [PATCH 1/5] IB/ehca: Support large page MRs
Add support for MR pages larger than 4K on eHCA2. This reduces firmware
memory consumption. If enabled via the mr_largepage module parameter, the MR
page size will be determined based on the MR length and the hardware
capabilities - if the MR is = 16M, 16M pages are used, for example.
Signed-off-by: Joachim Fenkes [EMAIL PROTECTED]
---
drivers/infiniband/hw/ehca/ehca_classes.h |9 +
drivers/infiniband/hw/ehca/ehca_main.c| 18 ++-
drivers/infiniband/hw/ehca/ehca_mrmw.c| 371 -
drivers/infiniband/hw/ehca/ehca_mrmw.h|2 +-
drivers/infiniband/hw/ehca/hcp_if.c | 20 ++-
5 files changed, 357 insertions(+), 63 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_classes.h
b/drivers/infiniband/hw/ehca/ehca_classes.h
index 043e4fb..63b8b9f 100644
--- a/drivers/infiniband/hw/ehca/ehca_classes.h
+++ b/drivers/infiniband/hw/ehca/ehca_classes.h
@@ -100,6 +100,11 @@ struct ehca_sport {
struct ehca_sma_attr saved_attr;
};
+#define HCA_CAP_MR_PGSIZE_4K 1
+#define HCA_CAP_MR_PGSIZE_64K 2
+#define HCA_CAP_MR_PGSIZE_1M 4
+#define HCA_CAP_MR_PGSIZE_16M 8
+
struct ehca_shca {
struct ib_device ib_device;
struct ibmebus_dev *ibmebus_dev;
@@ -115,6 +120,8 @@ struct ehca_shca {
struct h_galpas galpas;
struct mutex modify_mutex;
u64 hca_cap;
+ /* MR pgsize: bit 0-3 means 4K, 64K, 1M, 16M respectively */
+ u32 hca_cap_mr_pgsize;
int max_mtu;
};
@@ -206,6 +213,7 @@ struct ehca_mr {
enum ehca_mr_flag flags;
u32 num_kpages; /* number of kernel pages */
u32 num_hwpages;/* number of hw pages to form MR */
+ u64 hwpage_size;/* hw page size used for this MR */
int acl;/* ACL (stored here for usage in reregister) */
u64 *start; /* virtual start address (stored here for */
/* usage in reregister) */
@@ -240,6 +248,7 @@ struct ehca_mr_pginfo {
enum ehca_mr_pgi_type type;
u64 num_kpages;
u64 kpage_cnt;
+ u64 hwpage_size; /* hw page size used for this MR */
u64 num_hwpages; /* number of hw pages */
u64 hwpage_cnt; /* counter for hw pages */
u64 next_hwpage; /* next hw page in buffer/chunk/listelem */
diff --git a/drivers/infiniband/hw/ehca/ehca_main.c
b/drivers/infiniband/hw/ehca/ehca_main.c
index 36377c6..34661c3 100644
--- a/drivers/infiniband/hw/ehca/ehca_main.c
+++ b/drivers/infiniband/hw/ehca/ehca_main.c
@@ -63,6 +63,7 @@ int ehca_port_act_time = 30;
int ehca_poll_all_eqs = 1;
int ehca_static_rate = -1;
int ehca_scaling_code = 0;
+int ehca_mr_largepage = 0;
module_param_named(open_aqp1, ehca_open_aqp1, int, 0);
module_param_named(debug_level, ehca_debug_level, int, 0);
@@ -72,7 +73,8 @@ module_param_named(use_hp_mr, ehca_use_hp_mr, int, 0);
module_param_named(port_act_time, ehca_port_act_time, int, 0);
module_param_named(poll_all_eqs, ehca_poll_all_eqs, int, 0);
module_param_named(static_rate, ehca_static_rate, int, 0);
-module_param_named(scaling_code, ehca_scaling_code, int, 0);
+module_param_named(scaling_code, ehca_scaling_code, int, 0);
+module_param_named(mr_largepage, ehca_mr_largepage, int, 0);
MODULE_PARM_DESC(open_aqp1,
AQP1 on startup (0: no (default), 1: yes));
@@ -95,6 +97,9 @@ MODULE_PARM_DESC(static_rate,
set permanent static rate (default: disabled));
MODULE_PARM_DESC(scaling_code,
set scaling code (0: disabled/default, 1: enabled));
+MODULE_PARM_DESC(mr_largepage,
+use large page for MR (0: use PAGE_SIZE (default),
+1: use large page depending on MR size);
DEFINE_RWLOCK(ehca_qp_idr_lock);
DEFINE_RWLOCK(ehca_cq_idr_lock);
@@ -295,6 +300,8 @@ int ehca_sense_attributes(struct ehca_shca *shca)
if (EHCA_BMASK_GET(hca_cap_descr[i].mask, shca-hca_cap))
ehca_gen_dbg( %s, hca_cap_descr[i].descr);
+ shca-hca_cap_mr_pgsize = rblock-memory_page_size_supported;
+
port = (struct hipz_query_port *)rblock;
h_ret = hipz_h_query_port(shca-ipz_hca_handle, 1, port);
if (h_ret != H_SUCCESS) {
@@ -590,6 +597,14 @@ static ssize_t ehca_show_adapter_handle(struct device *dev,
}
static DEVICE_ATTR(adapter_handle, S_IRUGO, ehca_show_adapter_handle, NULL);
+static ssize_t ehca_show_mr_largepage(struct device *dev,
+ struct device_attribute *attr,
+ char *buf)
+{
+ return sprintf(buf, %d\n, ehca_mr_largepage);
+}
+static DEVICE_ATTR(mr_largepage, S_IRUGO, ehca_show_mr_largepage, NULL);
+
static struct attribute *ehca_dev_attrs[] = {
dev_attr_adapter_handle.attr
[ofa-general] [PATCH 2/5] ehca: Generate event when SRQ limit reached
From: Joachim Fenkes [EMAIL PROTECTED]
Date: Thu, 19 Jul 2007 20:51:43 +0200
Subject: [PATCH 2/5] IB/ehca: Generate event when SRQ limit reached
Signed-off-by: Joachim Fenkes [EMAIL PROTECTED]
---
drivers/infiniband/hw/ehca/ehca_irq.c | 42 ++---
1 files changed, 28 insertions(+), 14 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_irq.c
b/drivers/infiniband/hw/ehca/ehca_irq.c
index 4fb01fc..71c0799 100644
--- a/drivers/infiniband/hw/ehca/ehca_irq.c
+++ b/drivers/infiniband/hw/ehca/ehca_irq.c
@@ -175,9 +175,8 @@ error_data1:
}
-static void qp_event_callback(struct ehca_shca *shca,
- u64 eqe,
- enum ib_event_type event_type)
+static void qp_event_callback(struct ehca_shca *shca, u64 eqe,
+ enum ib_event_type event_type, int fatal)
{
struct ib_event event;
struct ehca_qp *qp;
@@ -191,16 +190,26 @@ static void qp_event_callback(struct ehca_shca *shca,
if (!qp)
return;
- ehca_error_data(shca, qp, qp-ipz_qp_handle.handle);
+ if (fatal)
+ ehca_error_data(shca, qp, qp-ipz_qp_handle.handle);
- if (!qp-ib_qp.event_handler)
- return;
+ event.device = shca-ib_device;
- event.device = shca-ib_device;
- event.event = event_type;
- event.element.qp = qp-ib_qp;
+ if (qp-ext_type == EQPT_SRQ) {
+ if (!qp-ib_srq.event_handler)
+ return;
- qp-ib_qp.event_handler(event, qp-ib_qp.qp_context);
+ event.event = fatal ? IB_EVENT_SRQ_ERR : event_type;
+ event.element.srq = qp-ib_srq;
+ qp-ib_srq.event_handler(event, qp-ib_srq.srq_context);
+ } else {
+ if (!qp-ib_qp.event_handler)
+ return;
+
+ event.event = event_type;
+ event.element.qp = qp-ib_qp;
+ qp-ib_qp.event_handler(event, qp-ib_qp.qp_context);
+ }
return;
}
@@ -234,17 +243,17 @@ static void parse_identifier(struct ehca_shca *shca, u64
eqe)
switch (identifier) {
case 0x02: /* path migrated */
- qp_event_callback(shca, eqe, IB_EVENT_PATH_MIG);
+ qp_event_callback(shca, eqe, IB_EVENT_PATH_MIG, 0);
break;
case 0x03: /* communication established */
- qp_event_callback(shca, eqe, IB_EVENT_COMM_EST);
+ qp_event_callback(shca, eqe, IB_EVENT_COMM_EST, 0);
break;
case 0x04: /* send queue drained */
- qp_event_callback(shca, eqe, IB_EVENT_SQ_DRAINED);
+ qp_event_callback(shca, eqe, IB_EVENT_SQ_DRAINED, 0);
break;
case 0x05: /* QP error */
case 0x06: /* QP error */
- qp_event_callback(shca, eqe, IB_EVENT_QP_FATAL);
+ qp_event_callback(shca, eqe, IB_EVENT_QP_FATAL, 1);
break;
case 0x07: /* CQ error */
case 0x08: /* CQ error */
@@ -278,6 +287,11 @@ static void parse_identifier(struct ehca_shca *shca, u64
eqe)
ehca_err(shca-ib_device, Interface trace stopped.);
break;
case 0x14: /* first error capture info available */
+ ehca_info(shca-ib_device, First error capture available);
+ break;
+ case 0x15: /* SRQ limit reached */
+ qp_event_callback(shca, eqe, IB_EVENT_SRQ_LIMIT_REACHED, 0);
+ break;
default:
ehca_err(shca-ib_device, Unknown identifier: %x on %s.,
identifier, shca-ib_device.name);
--
1.5.2
___
general mailing list
general@lists.openfabrics.org
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
[ofa-general] [PATCH 5/5] ehca: Support small QP queues
From: Stefan Roscher stefan.roscher at de.ibm.com
Date: Fri, 20 Jul 2007 13:59:14 +0200
Subject: [PATCH 5/5] IB/ehca: Small QP queues
eHCA2 supports QP queues that can be as small as 512 bytes. This greatly
reduces memory overhead for consumers that use lots of QPs with small queues
(e.g. RDMA-only QPs). Apart from dealing with firmware, this code needs to
manage bite-sized chunks of kernel pages, making sure that no kernel page is
shared between different protection domains.
Signed-off-by: Hoang-Nam Nguyen [EMAIL PROTECTED]
---
drivers/infiniband/hw/ehca/ehca_classes.h | 41 --
drivers/infiniband/hw/ehca/ehca_cq.c |8 +-
drivers/infiniband/hw/ehca/ehca_eq.c |8 +-
drivers/infiniband/hw/ehca/ehca_main.c| 14 ++-
drivers/infiniband/hw/ehca/ehca_pd.c | 25 +++-
drivers/infiniband/hw/ehca/ehca_qp.c | 163 +-
drivers/infiniband/hw/ehca/ehca_uverbs.c |2 +-
drivers/infiniband/hw/ehca/hcp_if.c | 30 +++--
drivers/infiniband/hw/ehca/ipz_pt_fn.c| 222 ++---
drivers/infiniband/hw/ehca/ipz_pt_fn.h| 26 +++-
10 files changed, 379 insertions(+), 160 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_classes.h
b/drivers/infiniband/hw/ehca/ehca_classes.h
index 63b8b9f..3725aa8 100644
--- a/drivers/infiniband/hw/ehca/ehca_classes.h
+++ b/drivers/infiniband/hw/ehca/ehca_classes.h
@@ -43,7 +43,6 @@
#ifndef __EHCA_CLASSES_H__
#define __EHCA_CLASSES_H__
-
struct ehca_module;
struct ehca_qp;
struct ehca_cq;
@@ -129,6 +128,10 @@ struct ehca_pd {
struct ib_pd ib_pd;
struct ipz_pd fw_pd;
u32 ownpid;
+ /* small queue mgmt */
+ struct mutex lock;
+ struct list_head free[2];
+ struct list_head full[2];
};
enum ehca_ext_qp_type {
@@ -307,6 +310,8 @@ int ehca_init_av_cache(void);
void ehca_cleanup_av_cache(void);
int ehca_init_mrmw_cache(void);
void ehca_cleanup_mrmw_cache(void);
+int ehca_init_small_qp_cache(void);
+void ehca_cleanup_small_qp_cache(void);
extern rwlock_t ehca_qp_idr_lock;
extern rwlock_t ehca_cq_idr_lock;
@@ -324,7 +329,7 @@ struct ipzu_queue_resp {
u32 queue_length; /* queue length allocated in bytes */
u32 pagesize;
u32 toggle_state;
- u32 dummy; /* padding for 8 byte alignment */
+ u32 offset; /* save offset within a page for small_qp */
};
struct ehca_create_cq_resp {
@@ -366,15 +371,29 @@ enum ehca_ll_comp_flags {
LLQP_COMP_MASK = 0x60,
};
+struct ehca_alloc_queue_parms {
+ /* input parameters */
+ int max_wr;
+ int max_sge;
+ int page_size;
+ int is_small;
+
+ /* output parameters */
+ u16 act_nr_wqes;
+ u8 act_nr_sges;
+ u32 queue_size; /* bytes for small queues, pages otherwise */
+};
+
struct ehca_alloc_qp_parms {
-/* input parameters */
+ struct ehca_alloc_queue_parms squeue;
+ struct ehca_alloc_queue_parms rqueue;
+
+ /* input parameters */
enum ehca_service_type servicetype;
+ int qp_storage;
int sigtype;
enum ehca_ext_qp_type ext_type;
enum ehca_ll_comp_flags ll_comp_flags;
-
- int max_send_wr, max_recv_wr;
- int max_send_sge, max_recv_sge;
int ud_av_l_key_ctl;
u32 token;
@@ -384,18 +403,10 @@ struct ehca_alloc_qp_parms {
u32 srq_qpn, srq_token, srq_limit;
-/* output parameters */
+ /* output parameters */
u32 real_qp_num;
struct ipz_qp_handle qp_handle;
struct h_galpas galpas;
-
- u16 act_nr_send_wqes;
- u16 act_nr_recv_wqes;
- u8 act_nr_recv_sges;
- u8 act_nr_send_sges;
-
- u32 nr_rq_pages;
- u32 nr_sq_pages;
};
int ehca_cq_assign_qp(struct ehca_cq *cq, struct ehca_qp *qp);
diff --git a/drivers/infiniband/hw/ehca/ehca_cq.c
b/drivers/infiniband/hw/ehca/ehca_cq.c
index 9e87883..5746787 100644
--- a/drivers/infiniband/hw/ehca/ehca_cq.c
+++ b/drivers/infiniband/hw/ehca/ehca_cq.c
@@ -190,8 +190,8 @@ struct ib_cq *ehca_create_cq(struct ib_device *device, int
cqe, int comp_vector,
goto create_cq_exit2;
}
- ipz_rc = ipz_queue_ctor(my_cq-ipz_queue, param.act_pages,
- EHCA_PAGESIZE, sizeof(struct ehca_cqe), 0);
+ ipz_rc = ipz_queue_ctor(NULL, my_cq-ipz_queue, param.act_pages,
+ EHCA_PAGESIZE, sizeof(struct ehca_cqe), 0, 0);
if (!ipz_rc) {
ehca_err(device, ipz_queue_ctor() failed ipz_rc=%x device=%p,
ipz_rc, device);
@@ -285,7 +285,7 @@ struct ib_cq *ehca_create_cq(struct ib_device *device, int
cqe, int comp_vector,
return cq;
create_cq_exit4:
- ipz_queue_dtor(my_cq-ipz_queue);
+ ipz_queue_dtor(NULL, my_cq-ipz_queue);
create_cq_exit3:
h_ret = hipz_h_destroy_cq(adapter_handle, my_cq, 1);
@@ -359,7 +359,7 @@ int ehca_destroy_cq(struct ib_cq *cq
[PATCH 0/5] ehca: MR large page, small queue and fixes
Here is a patch set against Roland's git, branch for-2.6.23 for ehca. It adds support for MR large page and small queues. In addition of that it also contains various small fixes from previous comments and what we found. They are in details: [1/5] adds support for MR large page [2/5] generates event when SRQ limit reached [3/5] makes ehca2ib_return_code() non inline [4/5] makes internal_create/destroy_qp() static [5/5] adds support for small queues The patches should apply cleanly, in order, against Roland's git. Please review the changes and apply the patches if they are okay. Regards, Nam Stefan - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
[PATCH 1/5] ehca: Supports large page MRs
From: Hoang-Nam Nguyen [EMAIL PROTECTED]
Date: Thu, 19 Jul 2007 20:48:04 +0200
Subject: [PATCH 1/5] IB/ehca: Support large page MRs
Add support for MR pages larger than 4K on eHCA2. This reduces firmware
memory consumption. If enabled via the mr_largepage module parameter, the MR
page size will be determined based on the MR length and the hardware
capabilities - if the MR is = 16M, 16M pages are used, for example.
Signed-off-by: Joachim Fenkes [EMAIL PROTECTED]
---
drivers/infiniband/hw/ehca/ehca_classes.h |9 +
drivers/infiniband/hw/ehca/ehca_main.c| 18 ++-
drivers/infiniband/hw/ehca/ehca_mrmw.c| 371 -
drivers/infiniband/hw/ehca/ehca_mrmw.h|2 +-
drivers/infiniband/hw/ehca/hcp_if.c | 20 ++-
5 files changed, 357 insertions(+), 63 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_classes.h
b/drivers/infiniband/hw/ehca/ehca_classes.h
index 043e4fb..63b8b9f 100644
--- a/drivers/infiniband/hw/ehca/ehca_classes.h
+++ b/drivers/infiniband/hw/ehca/ehca_classes.h
@@ -100,6 +100,11 @@ struct ehca_sport {
struct ehca_sma_attr saved_attr;
};
+#define HCA_CAP_MR_PGSIZE_4K 1
+#define HCA_CAP_MR_PGSIZE_64K 2
+#define HCA_CAP_MR_PGSIZE_1M 4
+#define HCA_CAP_MR_PGSIZE_16M 8
+
struct ehca_shca {
struct ib_device ib_device;
struct ibmebus_dev *ibmebus_dev;
@@ -115,6 +120,8 @@ struct ehca_shca {
struct h_galpas galpas;
struct mutex modify_mutex;
u64 hca_cap;
+ /* MR pgsize: bit 0-3 means 4K, 64K, 1M, 16M respectively */
+ u32 hca_cap_mr_pgsize;
int max_mtu;
};
@@ -206,6 +213,7 @@ struct ehca_mr {
enum ehca_mr_flag flags;
u32 num_kpages; /* number of kernel pages */
u32 num_hwpages;/* number of hw pages to form MR */
+ u64 hwpage_size;/* hw page size used for this MR */
int acl;/* ACL (stored here for usage in reregister) */
u64 *start; /* virtual start address (stored here for */
/* usage in reregister) */
@@ -240,6 +248,7 @@ struct ehca_mr_pginfo {
enum ehca_mr_pgi_type type;
u64 num_kpages;
u64 kpage_cnt;
+ u64 hwpage_size; /* hw page size used for this MR */
u64 num_hwpages; /* number of hw pages */
u64 hwpage_cnt; /* counter for hw pages */
u64 next_hwpage; /* next hw page in buffer/chunk/listelem */
diff --git a/drivers/infiniband/hw/ehca/ehca_main.c
b/drivers/infiniband/hw/ehca/ehca_main.c
index 36377c6..34661c3 100644
--- a/drivers/infiniband/hw/ehca/ehca_main.c
+++ b/drivers/infiniband/hw/ehca/ehca_main.c
@@ -63,6 +63,7 @@ int ehca_port_act_time = 30;
int ehca_poll_all_eqs = 1;
int ehca_static_rate = -1;
int ehca_scaling_code = 0;
+int ehca_mr_largepage = 0;
module_param_named(open_aqp1, ehca_open_aqp1, int, 0);
module_param_named(debug_level, ehca_debug_level, int, 0);
@@ -72,7 +73,8 @@ module_param_named(use_hp_mr, ehca_use_hp_mr, int, 0);
module_param_named(port_act_time, ehca_port_act_time, int, 0);
module_param_named(poll_all_eqs, ehca_poll_all_eqs, int, 0);
module_param_named(static_rate, ehca_static_rate, int, 0);
-module_param_named(scaling_code, ehca_scaling_code, int, 0);
+module_param_named(scaling_code, ehca_scaling_code, int, 0);
+module_param_named(mr_largepage, ehca_mr_largepage, int, 0);
MODULE_PARM_DESC(open_aqp1,
AQP1 on startup (0: no (default), 1: yes));
@@ -95,6 +97,9 @@ MODULE_PARM_DESC(static_rate,
set permanent static rate (default: disabled));
MODULE_PARM_DESC(scaling_code,
set scaling code (0: disabled/default, 1: enabled));
+MODULE_PARM_DESC(mr_largepage,
+use large page for MR (0: use PAGE_SIZE (default),
+1: use large page depending on MR size);
DEFINE_RWLOCK(ehca_qp_idr_lock);
DEFINE_RWLOCK(ehca_cq_idr_lock);
@@ -295,6 +300,8 @@ int ehca_sense_attributes(struct ehca_shca *shca)
if (EHCA_BMASK_GET(hca_cap_descr[i].mask, shca-hca_cap))
ehca_gen_dbg( %s, hca_cap_descr[i].descr);
+ shca-hca_cap_mr_pgsize = rblock-memory_page_size_supported;
+
port = (struct hipz_query_port *)rblock;
h_ret = hipz_h_query_port(shca-ipz_hca_handle, 1, port);
if (h_ret != H_SUCCESS) {
@@ -590,6 +597,14 @@ static ssize_t ehca_show_adapter_handle(struct device *dev,
}
static DEVICE_ATTR(adapter_handle, S_IRUGO, ehca_show_adapter_handle, NULL);
+static ssize_t ehca_show_mr_largepage(struct device *dev,
+ struct device_attribute *attr,
+ char *buf)
+{
+ return sprintf(buf, %d\n, ehca_mr_largepage);
+}
+static DEVICE_ATTR(mr_largepage, S_IRUGO, ehca_show_mr_largepage, NULL);
+
static struct attribute *ehca_dev_attrs[] = {
dev_attr_adapter_handle.attr
[PATCH 3/5] ehca: Make ehca2ib_return_code() non-inline
From: Joachim Fenkes [EMAIL PROTECTED]
Date: Thu, 19 Jul 2007 21:13:57 +0200
Subject: [PATCH 3/5] IB/ehca: Make ehca2ib_return_code() non-inline
It's nowhere in the main path and making it non-inline saves ~1.5K of code.
Signed-off-by: Joachim Fenkes [EMAIL PROTECTED]
---
drivers/infiniband/hw/ehca/ehca_main.c | 17 +
drivers/infiniband/hw/ehca/ehca_tools.h | 19 +--
2 files changed, 18 insertions(+), 18 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_main.c
b/drivers/infiniband/hw/ehca/ehca_main.c
index 34661c3..3bd7afb 100644
--- a/drivers/infiniband/hw/ehca/ehca_main.c
+++ b/drivers/infiniband/hw/ehca/ehca_main.c
@@ -130,6 +130,23 @@ void ehca_free_fw_ctrlblock(void *ptr)
}
#endif
+int ehca2ib_return_code(u64 ehca_rc)
+{
+ switch (ehca_rc) {
+ case H_SUCCESS:
+ return 0;
+ case H_RESOURCE: /* Resource in use */
+ case H_BUSY:
+ return -EBUSY;
+ case H_NOT_ENOUGH_RESOURCES: /* insufficient resources */
+ case H_CONSTRAINED: /* resource constraint */
+ case H_NO_MEM:
+ return -ENOMEM;
+ default:
+ return -EINVAL;
+ }
+}
+
static int ehca_create_slab_caches(void)
{
int ret;
diff --git a/drivers/infiniband/hw/ehca/ehca_tools.h
b/drivers/infiniband/hw/ehca/ehca_tools.h
index 678b813..57c77a7 100644
--- a/drivers/infiniband/hw/ehca/ehca_tools.h
+++ b/drivers/infiniband/hw/ehca/ehca_tools.h
@@ -154,24 +154,7 @@ extern int ehca_debug_level;
#define EHCA_BMASK_GET(mask, value) \
(EHCA_BMASK_MASK(mask) (((u64)(value)) EHCA_BMASK_SHIFTPOS(mask)))
-
/* Converts ehca to ib return code */
-static inline int ehca2ib_return_code(u64 ehca_rc)
-{
- switch (ehca_rc) {
- case H_SUCCESS:
- return 0;
- case H_RESOURCE: /* Resource in use */
- case H_BUSY:
- return -EBUSY;
- case H_NOT_ENOUGH_RESOURCES: /* insufficient resources */
- case H_CONSTRAINED: /* resource constraint */
- case H_NO_MEM:
- return -ENOMEM;
- default:
- return -EINVAL;
- }
-}
-
+int ehca2ib_return_code(u64 ehca_rc);
#endif /* EHCA_TOOLS_H */
--
1.5.2
-
To unsubscribe from this list: send the line unsubscribe linux-kernel in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[PATCH 2/5] ehca: Generate event when SRQ limit reached
From: Joachim Fenkes [EMAIL PROTECTED]
Date: Thu, 19 Jul 2007 20:51:43 +0200
Subject: [PATCH 2/5] IB/ehca: Generate event when SRQ limit reached
Signed-off-by: Joachim Fenkes [EMAIL PROTECTED]
---
drivers/infiniband/hw/ehca/ehca_irq.c | 42 ++---
1 files changed, 28 insertions(+), 14 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_irq.c
b/drivers/infiniband/hw/ehca/ehca_irq.c
index 4fb01fc..71c0799 100644
--- a/drivers/infiniband/hw/ehca/ehca_irq.c
+++ b/drivers/infiniband/hw/ehca/ehca_irq.c
@@ -175,9 +175,8 @@ error_data1:
}
-static void qp_event_callback(struct ehca_shca *shca,
- u64 eqe,
- enum ib_event_type event_type)
+static void qp_event_callback(struct ehca_shca *shca, u64 eqe,
+ enum ib_event_type event_type, int fatal)
{
struct ib_event event;
struct ehca_qp *qp;
@@ -191,16 +190,26 @@ static void qp_event_callback(struct ehca_shca *shca,
if (!qp)
return;
- ehca_error_data(shca, qp, qp-ipz_qp_handle.handle);
+ if (fatal)
+ ehca_error_data(shca, qp, qp-ipz_qp_handle.handle);
- if (!qp-ib_qp.event_handler)
- return;
+ event.device = shca-ib_device;
- event.device = shca-ib_device;
- event.event = event_type;
- event.element.qp = qp-ib_qp;
+ if (qp-ext_type == EQPT_SRQ) {
+ if (!qp-ib_srq.event_handler)
+ return;
- qp-ib_qp.event_handler(event, qp-ib_qp.qp_context);
+ event.event = fatal ? IB_EVENT_SRQ_ERR : event_type;
+ event.element.srq = qp-ib_srq;
+ qp-ib_srq.event_handler(event, qp-ib_srq.srq_context);
+ } else {
+ if (!qp-ib_qp.event_handler)
+ return;
+
+ event.event = event_type;
+ event.element.qp = qp-ib_qp;
+ qp-ib_qp.event_handler(event, qp-ib_qp.qp_context);
+ }
return;
}
@@ -234,17 +243,17 @@ static void parse_identifier(struct ehca_shca *shca, u64
eqe)
switch (identifier) {
case 0x02: /* path migrated */
- qp_event_callback(shca, eqe, IB_EVENT_PATH_MIG);
+ qp_event_callback(shca, eqe, IB_EVENT_PATH_MIG, 0);
break;
case 0x03: /* communication established */
- qp_event_callback(shca, eqe, IB_EVENT_COMM_EST);
+ qp_event_callback(shca, eqe, IB_EVENT_COMM_EST, 0);
break;
case 0x04: /* send queue drained */
- qp_event_callback(shca, eqe, IB_EVENT_SQ_DRAINED);
+ qp_event_callback(shca, eqe, IB_EVENT_SQ_DRAINED, 0);
break;
case 0x05: /* QP error */
case 0x06: /* QP error */
- qp_event_callback(shca, eqe, IB_EVENT_QP_FATAL);
+ qp_event_callback(shca, eqe, IB_EVENT_QP_FATAL, 1);
break;
case 0x07: /* CQ error */
case 0x08: /* CQ error */
@@ -278,6 +287,11 @@ static void parse_identifier(struct ehca_shca *shca, u64
eqe)
ehca_err(shca-ib_device, Interface trace stopped.);
break;
case 0x14: /* first error capture info available */
+ ehca_info(shca-ib_device, First error capture available);
+ break;
+ case 0x15: /* SRQ limit reached */
+ qp_event_callback(shca, eqe, IB_EVENT_SRQ_LIMIT_REACHED, 0);
+ break;
default:
ehca_err(shca-ib_device, Unknown identifier: %x on %s.,
identifier, shca-ib_device.name);
--
1.5.2
-
To unsubscribe from this list: send the line unsubscribe linux-kernel in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[PATCH 4/5] ehca: Make internal_create/destroy_qp() static
From: Joachim Fenkes [EMAIL PROTECTED]
Date: Thu, 19 Jul 2007 21:40:00 +0200
Subject: [PATCH 4/5] IB/ehca: Make internal_{create,destroy}_qp() static
They're only used in ehca_qp.c
Signed-off-by: Joachim Fenkes [EMAIL PROTECTED]
---
drivers/infiniband/hw/ehca/ehca_qp.c | 17 +
1 files changed, 9 insertions(+), 8 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_qp.c
b/drivers/infiniband/hw/ehca/ehca_qp.c
index 48e9cea..b916d9c 100644
--- a/drivers/infiniband/hw/ehca/ehca_qp.c
+++ b/drivers/infiniband/hw/ehca/ehca_qp.c
@@ -363,10 +363,11 @@ init_qp_queue1:
* the value of the is_srq parameter. If init_attr and srq_init_attr share
* fields, the field out of init_attr is used.
*/
-struct ehca_qp *internal_create_qp(struct ib_pd *pd,
- struct ib_qp_init_attr *init_attr,
- struct ib_srq_init_attr *srq_init_attr,
- struct ib_udata *udata, int is_srq)
+static struct ehca_qp *internal_create_qp(
+ struct ib_pd *pd,
+ struct ib_qp_init_attr *init_attr,
+ struct ib_srq_init_attr *srq_init_attr,
+ struct ib_udata *udata, int is_srq)
{
struct ehca_qp *my_qp;
struct ehca_pd *my_pd = container_of(pd, struct ehca_pd, ib_pd);
@@ -752,8 +753,8 @@ struct ib_qp *ehca_create_qp(struct ib_pd *pd,
return IS_ERR(ret) ? (struct ib_qp *)ret : ret-ib_qp;
}
-int internal_destroy_qp(struct ib_device *dev, struct ehca_qp *my_qp,
- struct ib_uobject *uobject);
+static int internal_destroy_qp(struct ib_device *dev, struct ehca_qp *my_qp,
+ struct ib_uobject *uobject);
struct ib_srq *ehca_create_srq(struct ib_pd *pd,
struct ib_srq_init_attr *srq_init_attr,
@@ -1669,8 +1670,8 @@ query_srq_exit1:
return ret;
}
-int internal_destroy_qp(struct ib_device *dev, struct ehca_qp *my_qp,
- struct ib_uobject *uobject)
+static int internal_destroy_qp(struct ib_device *dev, struct ehca_qp *my_qp,
+ struct ib_uobject *uobject)
{
struct ehca_shca *shca = container_of(dev, struct ehca_shca, ib_device);
struct ehca_pd *my_pd = container_of(my_qp-ib_qp.pd, struct ehca_pd,
--
1.5.2
-
To unsubscribe from this list: send the line unsubscribe linux-kernel in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[PATCH 5/5] ehca: Support small QP queues
From: Stefan Roscher stefan.roscher at de.ibm.com
Date: Fri, 20 Jul 2007 13:59:14 +0200
Subject: [PATCH 5/5] IB/ehca: Small QP queues
eHCA2 supports QP queues that can be as small as 512 bytes. This greatly
reduces memory overhead for consumers that use lots of QPs with small queues
(e.g. RDMA-only QPs). Apart from dealing with firmware, this code needs to
manage bite-sized chunks of kernel pages, making sure that no kernel page is
shared between different protection domains.
Signed-off-by: Hoang-Nam Nguyen [EMAIL PROTECTED]
---
drivers/infiniband/hw/ehca/ehca_classes.h | 41 --
drivers/infiniband/hw/ehca/ehca_cq.c |8 +-
drivers/infiniband/hw/ehca/ehca_eq.c |8 +-
drivers/infiniband/hw/ehca/ehca_main.c| 14 ++-
drivers/infiniband/hw/ehca/ehca_pd.c | 25 +++-
drivers/infiniband/hw/ehca/ehca_qp.c | 163 +-
drivers/infiniband/hw/ehca/ehca_uverbs.c |2 +-
drivers/infiniband/hw/ehca/hcp_if.c | 30 +++--
drivers/infiniband/hw/ehca/ipz_pt_fn.c| 222 ++---
drivers/infiniband/hw/ehca/ipz_pt_fn.h| 26 +++-
10 files changed, 379 insertions(+), 160 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_classes.h
b/drivers/infiniband/hw/ehca/ehca_classes.h
index 63b8b9f..3725aa8 100644
--- a/drivers/infiniband/hw/ehca/ehca_classes.h
+++ b/drivers/infiniband/hw/ehca/ehca_classes.h
@@ -43,7 +43,6 @@
#ifndef __EHCA_CLASSES_H__
#define __EHCA_CLASSES_H__
-
struct ehca_module;
struct ehca_qp;
struct ehca_cq;
@@ -129,6 +128,10 @@ struct ehca_pd {
struct ib_pd ib_pd;
struct ipz_pd fw_pd;
u32 ownpid;
+ /* small queue mgmt */
+ struct mutex lock;
+ struct list_head free[2];
+ struct list_head full[2];
};
enum ehca_ext_qp_type {
@@ -307,6 +310,8 @@ int ehca_init_av_cache(void);
void ehca_cleanup_av_cache(void);
int ehca_init_mrmw_cache(void);
void ehca_cleanup_mrmw_cache(void);
+int ehca_init_small_qp_cache(void);
+void ehca_cleanup_small_qp_cache(void);
extern rwlock_t ehca_qp_idr_lock;
extern rwlock_t ehca_cq_idr_lock;
@@ -324,7 +329,7 @@ struct ipzu_queue_resp {
u32 queue_length; /* queue length allocated in bytes */
u32 pagesize;
u32 toggle_state;
- u32 dummy; /* padding for 8 byte alignment */
+ u32 offset; /* save offset within a page for small_qp */
};
struct ehca_create_cq_resp {
@@ -366,15 +371,29 @@ enum ehca_ll_comp_flags {
LLQP_COMP_MASK = 0x60,
};
+struct ehca_alloc_queue_parms {
+ /* input parameters */
+ int max_wr;
+ int max_sge;
+ int page_size;
+ int is_small;
+
+ /* output parameters */
+ u16 act_nr_wqes;
+ u8 act_nr_sges;
+ u32 queue_size; /* bytes for small queues, pages otherwise */
+};
+
struct ehca_alloc_qp_parms {
-/* input parameters */
+ struct ehca_alloc_queue_parms squeue;
+ struct ehca_alloc_queue_parms rqueue;
+
+ /* input parameters */
enum ehca_service_type servicetype;
+ int qp_storage;
int sigtype;
enum ehca_ext_qp_type ext_type;
enum ehca_ll_comp_flags ll_comp_flags;
-
- int max_send_wr, max_recv_wr;
- int max_send_sge, max_recv_sge;
int ud_av_l_key_ctl;
u32 token;
@@ -384,18 +403,10 @@ struct ehca_alloc_qp_parms {
u32 srq_qpn, srq_token, srq_limit;
-/* output parameters */
+ /* output parameters */
u32 real_qp_num;
struct ipz_qp_handle qp_handle;
struct h_galpas galpas;
-
- u16 act_nr_send_wqes;
- u16 act_nr_recv_wqes;
- u8 act_nr_recv_sges;
- u8 act_nr_send_sges;
-
- u32 nr_rq_pages;
- u32 nr_sq_pages;
};
int ehca_cq_assign_qp(struct ehca_cq *cq, struct ehca_qp *qp);
diff --git a/drivers/infiniband/hw/ehca/ehca_cq.c
b/drivers/infiniband/hw/ehca/ehca_cq.c
index 9e87883..5746787 100644
--- a/drivers/infiniband/hw/ehca/ehca_cq.c
+++ b/drivers/infiniband/hw/ehca/ehca_cq.c
@@ -190,8 +190,8 @@ struct ib_cq *ehca_create_cq(struct ib_device *device, int
cqe, int comp_vector,
goto create_cq_exit2;
}
- ipz_rc = ipz_queue_ctor(my_cq-ipz_queue, param.act_pages,
- EHCA_PAGESIZE, sizeof(struct ehca_cqe), 0);
+ ipz_rc = ipz_queue_ctor(NULL, my_cq-ipz_queue, param.act_pages,
+ EHCA_PAGESIZE, sizeof(struct ehca_cqe), 0, 0);
if (!ipz_rc) {
ehca_err(device, ipz_queue_ctor() failed ipz_rc=%x device=%p,
ipz_rc, device);
@@ -285,7 +285,7 @@ struct ib_cq *ehca_create_cq(struct ib_device *device, int
cqe, int comp_vector,
return cq;
create_cq_exit4:
- ipz_queue_dtor(my_cq-ipz_queue);
+ ipz_queue_dtor(NULL, my_cq-ipz_queue);
create_cq_exit3:
h_ret = hipz_h_destroy_cq(adapter_handle, my_cq, 1);
@@ -359,7 +359,7 @@ int ehca_destroy_cq(struct ib_cq *cq
[PATCH 1/5] ehca: Supports large page MRs
From: Hoang-Nam Nguyen [EMAIL PROTECTED]
Date: Thu, 19 Jul 2007 20:48:04 +0200
Subject: [PATCH 1/5] IB/ehca: Support large page MRs
Add support for MR pages larger than 4K on eHCA2. This reduces firmware
memory consumption. If enabled via the mr_largepage module parameter, the MR
page size will be determined based on the MR length and the hardware
capabilities - if the MR is = 16M, 16M pages are used, for example.
Signed-off-by: Joachim Fenkes [EMAIL PROTECTED]
---
drivers/infiniband/hw/ehca/ehca_classes.h |9 +
drivers/infiniband/hw/ehca/ehca_main.c| 18 ++-
drivers/infiniband/hw/ehca/ehca_mrmw.c| 371 -
drivers/infiniband/hw/ehca/ehca_mrmw.h|2 +-
drivers/infiniband/hw/ehca/hcp_if.c | 20 ++-
5 files changed, 357 insertions(+), 63 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_classes.h
b/drivers/infiniband/hw/ehca/ehca_classes.h
index 043e4fb..63b8b9f 100644
--- a/drivers/infiniband/hw/ehca/ehca_classes.h
+++ b/drivers/infiniband/hw/ehca/ehca_classes.h
@@ -100,6 +100,11 @@ struct ehca_sport {
struct ehca_sma_attr saved_attr;
};
+#define HCA_CAP_MR_PGSIZE_4K 1
+#define HCA_CAP_MR_PGSIZE_64K 2
+#define HCA_CAP_MR_PGSIZE_1M 4
+#define HCA_CAP_MR_PGSIZE_16M 8
+
struct ehca_shca {
struct ib_device ib_device;
struct ibmebus_dev *ibmebus_dev;
@@ -115,6 +120,8 @@ struct ehca_shca {
struct h_galpas galpas;
struct mutex modify_mutex;
u64 hca_cap;
+ /* MR pgsize: bit 0-3 means 4K, 64K, 1M, 16M respectively */
+ u32 hca_cap_mr_pgsize;
int max_mtu;
};
@@ -206,6 +213,7 @@ struct ehca_mr {
enum ehca_mr_flag flags;
u32 num_kpages; /* number of kernel pages */
u32 num_hwpages;/* number of hw pages to form MR */
+ u64 hwpage_size;/* hw page size used for this MR */
int acl;/* ACL (stored here for usage in reregister) */
u64 *start; /* virtual start address (stored here for */
/* usage in reregister) */
@@ -240,6 +248,7 @@ struct ehca_mr_pginfo {
enum ehca_mr_pgi_type type;
u64 num_kpages;
u64 kpage_cnt;
+ u64 hwpage_size; /* hw page size used for this MR */
u64 num_hwpages; /* number of hw pages */
u64 hwpage_cnt; /* counter for hw pages */
u64 next_hwpage; /* next hw page in buffer/chunk/listelem */
diff --git a/drivers/infiniband/hw/ehca/ehca_main.c
b/drivers/infiniband/hw/ehca/ehca_main.c
index 36377c6..34661c3 100644
--- a/drivers/infiniband/hw/ehca/ehca_main.c
+++ b/drivers/infiniband/hw/ehca/ehca_main.c
@@ -63,6 +63,7 @@ int ehca_port_act_time = 30;
int ehca_poll_all_eqs = 1;
int ehca_static_rate = -1;
int ehca_scaling_code = 0;
+int ehca_mr_largepage = 0;
module_param_named(open_aqp1, ehca_open_aqp1, int, 0);
module_param_named(debug_level, ehca_debug_level, int, 0);
@@ -72,7 +73,8 @@ module_param_named(use_hp_mr, ehca_use_hp_mr, int, 0);
module_param_named(port_act_time, ehca_port_act_time, int, 0);
module_param_named(poll_all_eqs, ehca_poll_all_eqs, int, 0);
module_param_named(static_rate, ehca_static_rate, int, 0);
-module_param_named(scaling_code, ehca_scaling_code, int, 0);
+module_param_named(scaling_code, ehca_scaling_code, int, 0);
+module_param_named(mr_largepage, ehca_mr_largepage, int, 0);
MODULE_PARM_DESC(open_aqp1,
AQP1 on startup (0: no (default), 1: yes));
@@ -95,6 +97,9 @@ MODULE_PARM_DESC(static_rate,
set permanent static rate (default: disabled));
MODULE_PARM_DESC(scaling_code,
set scaling code (0: disabled/default, 1: enabled));
+MODULE_PARM_DESC(mr_largepage,
+use large page for MR (0: use PAGE_SIZE (default),
+1: use large page depending on MR size);
DEFINE_RWLOCK(ehca_qp_idr_lock);
DEFINE_RWLOCK(ehca_cq_idr_lock);
@@ -295,6 +300,8 @@ int ehca_sense_attributes(struct ehca_shca *shca)
if (EHCA_BMASK_GET(hca_cap_descr[i].mask, shca-hca_cap))
ehca_gen_dbg( %s, hca_cap_descr[i].descr);
+ shca-hca_cap_mr_pgsize = rblock-memory_page_size_supported;
+
port = (struct hipz_query_port *)rblock;
h_ret = hipz_h_query_port(shca-ipz_hca_handle, 1, port);
if (h_ret != H_SUCCESS) {
@@ -590,6 +597,14 @@ static ssize_t ehca_show_adapter_handle(struct device *dev,
}
static DEVICE_ATTR(adapter_handle, S_IRUGO, ehca_show_adapter_handle, NULL);
+static ssize_t ehca_show_mr_largepage(struct device *dev,
+ struct device_attribute *attr,
+ char *buf)
+{
+ return sprintf(buf, %d\n, ehca_mr_largepage);
+}
+static DEVICE_ATTR(mr_largepage, S_IRUGO, ehca_show_mr_largepage, NULL);
+
static struct attribute *ehca_dev_attrs[] = {
dev_attr_adapter_handle.attr
[PATCH 2/5] ehca: Generate event when SRQ limit reached
From: Joachim Fenkes [EMAIL PROTECTED]
Date: Thu, 19 Jul 2007 20:51:43 +0200
Subject: [PATCH 2/5] IB/ehca: Generate event when SRQ limit reached
Signed-off-by: Joachim Fenkes [EMAIL PROTECTED]
---
drivers/infiniband/hw/ehca/ehca_irq.c | 42 ++---
1 files changed, 28 insertions(+), 14 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_irq.c
b/drivers/infiniband/hw/ehca/ehca_irq.c
index 4fb01fc..71c0799 100644
--- a/drivers/infiniband/hw/ehca/ehca_irq.c
+++ b/drivers/infiniband/hw/ehca/ehca_irq.c
@@ -175,9 +175,8 @@ error_data1:
}
-static void qp_event_callback(struct ehca_shca *shca,
- u64 eqe,
- enum ib_event_type event_type)
+static void qp_event_callback(struct ehca_shca *shca, u64 eqe,
+ enum ib_event_type event_type, int fatal)
{
struct ib_event event;
struct ehca_qp *qp;
@@ -191,16 +190,26 @@ static void qp_event_callback(struct ehca_shca *shca,
if (!qp)
return;
- ehca_error_data(shca, qp, qp-ipz_qp_handle.handle);
+ if (fatal)
+ ehca_error_data(shca, qp, qp-ipz_qp_handle.handle);
- if (!qp-ib_qp.event_handler)
- return;
+ event.device = shca-ib_device;
- event.device = shca-ib_device;
- event.event = event_type;
- event.element.qp = qp-ib_qp;
+ if (qp-ext_type == EQPT_SRQ) {
+ if (!qp-ib_srq.event_handler)
+ return;
- qp-ib_qp.event_handler(event, qp-ib_qp.qp_context);
+ event.event = fatal ? IB_EVENT_SRQ_ERR : event_type;
+ event.element.srq = qp-ib_srq;
+ qp-ib_srq.event_handler(event, qp-ib_srq.srq_context);
+ } else {
+ if (!qp-ib_qp.event_handler)
+ return;
+
+ event.event = event_type;
+ event.element.qp = qp-ib_qp;
+ qp-ib_qp.event_handler(event, qp-ib_qp.qp_context);
+ }
return;
}
@@ -234,17 +243,17 @@ static void parse_identifier(struct ehca_shca *shca, u64
eqe)
switch (identifier) {
case 0x02: /* path migrated */
- qp_event_callback(shca, eqe, IB_EVENT_PATH_MIG);
+ qp_event_callback(shca, eqe, IB_EVENT_PATH_MIG, 0);
break;
case 0x03: /* communication established */
- qp_event_callback(shca, eqe, IB_EVENT_COMM_EST);
+ qp_event_callback(shca, eqe, IB_EVENT_COMM_EST, 0);
break;
case 0x04: /* send queue drained */
- qp_event_callback(shca, eqe, IB_EVENT_SQ_DRAINED);
+ qp_event_callback(shca, eqe, IB_EVENT_SQ_DRAINED, 0);
break;
case 0x05: /* QP error */
case 0x06: /* QP error */
- qp_event_callback(shca, eqe, IB_EVENT_QP_FATAL);
+ qp_event_callback(shca, eqe, IB_EVENT_QP_FATAL, 1);
break;
case 0x07: /* CQ error */
case 0x08: /* CQ error */
@@ -278,6 +287,11 @@ static void parse_identifier(struct ehca_shca *shca, u64
eqe)
ehca_err(shca-ib_device, Interface trace stopped.);
break;
case 0x14: /* first error capture info available */
+ ehca_info(shca-ib_device, First error capture available);
+ break;
+ case 0x15: /* SRQ limit reached */
+ qp_event_callback(shca, eqe, IB_EVENT_SRQ_LIMIT_REACHED, 0);
+ break;
default:
ehca_err(shca-ib_device, Unknown identifier: %x on %s.,
identifier, shca-ib_device.name);
--
1.5.2
___
Linuxppc-dev mailing list
Linuxppc-dev@ozlabs.org
https://ozlabs.org/mailman/listinfo/linuxppc-dev
[PATCH 3/5] ehca: Make ehca2ib_return_code() non-inline
From: Joachim Fenkes [EMAIL PROTECTED]
Date: Thu, 19 Jul 2007 21:13:57 +0200
Subject: [PATCH 3/5] IB/ehca: Make ehca2ib_return_code() non-inline
It's nowhere in the main path and making it non-inline saves ~1.5K of code.
Signed-off-by: Joachim Fenkes [EMAIL PROTECTED]
---
drivers/infiniband/hw/ehca/ehca_main.c | 17 +
drivers/infiniband/hw/ehca/ehca_tools.h | 19 +--
2 files changed, 18 insertions(+), 18 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_main.c
b/drivers/infiniband/hw/ehca/ehca_main.c
index 34661c3..3bd7afb 100644
--- a/drivers/infiniband/hw/ehca/ehca_main.c
+++ b/drivers/infiniband/hw/ehca/ehca_main.c
@@ -130,6 +130,23 @@ void ehca_free_fw_ctrlblock(void *ptr)
}
#endif
+int ehca2ib_return_code(u64 ehca_rc)
+{
+ switch (ehca_rc) {
+ case H_SUCCESS:
+ return 0;
+ case H_RESOURCE: /* Resource in use */
+ case H_BUSY:
+ return -EBUSY;
+ case H_NOT_ENOUGH_RESOURCES: /* insufficient resources */
+ case H_CONSTRAINED: /* resource constraint */
+ case H_NO_MEM:
+ return -ENOMEM;
+ default:
+ return -EINVAL;
+ }
+}
+
static int ehca_create_slab_caches(void)
{
int ret;
diff --git a/drivers/infiniband/hw/ehca/ehca_tools.h
b/drivers/infiniband/hw/ehca/ehca_tools.h
index 678b813..57c77a7 100644
--- a/drivers/infiniband/hw/ehca/ehca_tools.h
+++ b/drivers/infiniband/hw/ehca/ehca_tools.h
@@ -154,24 +154,7 @@ extern int ehca_debug_level;
#define EHCA_BMASK_GET(mask, value) \
(EHCA_BMASK_MASK(mask) (((u64)(value)) EHCA_BMASK_SHIFTPOS(mask)))
-
/* Converts ehca to ib return code */
-static inline int ehca2ib_return_code(u64 ehca_rc)
-{
- switch (ehca_rc) {
- case H_SUCCESS:
- return 0;
- case H_RESOURCE: /* Resource in use */
- case H_BUSY:
- return -EBUSY;
- case H_NOT_ENOUGH_RESOURCES: /* insufficient resources */
- case H_CONSTRAINED: /* resource constraint */
- case H_NO_MEM:
- return -ENOMEM;
- default:
- return -EINVAL;
- }
-}
-
+int ehca2ib_return_code(u64 ehca_rc);
#endif /* EHCA_TOOLS_H */
--
1.5.2
___
Linuxppc-dev mailing list
Linuxppc-dev@ozlabs.org
https://ozlabs.org/mailman/listinfo/linuxppc-dev
[PATCH 5/5] ehca: Support small QP queues
From: Stefan Roscher stefan.roscher at de.ibm.com
Date: Fri, 20 Jul 2007 13:59:14 +0200
Subject: [PATCH 5/5] IB/ehca: Small QP queues
eHCA2 supports QP queues that can be as small as 512 bytes. This greatly
reduces memory overhead for consumers that use lots of QPs with small queues
(e.g. RDMA-only QPs). Apart from dealing with firmware, this code needs to
manage bite-sized chunks of kernel pages, making sure that no kernel page is
shared between different protection domains.
Signed-off-by: Hoang-Nam Nguyen [EMAIL PROTECTED]
---
drivers/infiniband/hw/ehca/ehca_classes.h | 41 --
drivers/infiniband/hw/ehca/ehca_cq.c |8 +-
drivers/infiniband/hw/ehca/ehca_eq.c |8 +-
drivers/infiniband/hw/ehca/ehca_main.c| 14 ++-
drivers/infiniband/hw/ehca/ehca_pd.c | 25 +++-
drivers/infiniband/hw/ehca/ehca_qp.c | 163 +-
drivers/infiniband/hw/ehca/ehca_uverbs.c |2 +-
drivers/infiniband/hw/ehca/hcp_if.c | 30 +++--
drivers/infiniband/hw/ehca/ipz_pt_fn.c| 222 ++---
drivers/infiniband/hw/ehca/ipz_pt_fn.h| 26 +++-
10 files changed, 379 insertions(+), 160 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_classes.h
b/drivers/infiniband/hw/ehca/ehca_classes.h
index 63b8b9f..3725aa8 100644
--- a/drivers/infiniband/hw/ehca/ehca_classes.h
+++ b/drivers/infiniband/hw/ehca/ehca_classes.h
@@ -43,7 +43,6 @@
#ifndef __EHCA_CLASSES_H__
#define __EHCA_CLASSES_H__
-
struct ehca_module;
struct ehca_qp;
struct ehca_cq;
@@ -129,6 +128,10 @@ struct ehca_pd {
struct ib_pd ib_pd;
struct ipz_pd fw_pd;
u32 ownpid;
+ /* small queue mgmt */
+ struct mutex lock;
+ struct list_head free[2];
+ struct list_head full[2];
};
enum ehca_ext_qp_type {
@@ -307,6 +310,8 @@ int ehca_init_av_cache(void);
void ehca_cleanup_av_cache(void);
int ehca_init_mrmw_cache(void);
void ehca_cleanup_mrmw_cache(void);
+int ehca_init_small_qp_cache(void);
+void ehca_cleanup_small_qp_cache(void);
extern rwlock_t ehca_qp_idr_lock;
extern rwlock_t ehca_cq_idr_lock;
@@ -324,7 +329,7 @@ struct ipzu_queue_resp {
u32 queue_length; /* queue length allocated in bytes */
u32 pagesize;
u32 toggle_state;
- u32 dummy; /* padding for 8 byte alignment */
+ u32 offset; /* save offset within a page for small_qp */
};
struct ehca_create_cq_resp {
@@ -366,15 +371,29 @@ enum ehca_ll_comp_flags {
LLQP_COMP_MASK = 0x60,
};
+struct ehca_alloc_queue_parms {
+ /* input parameters */
+ int max_wr;
+ int max_sge;
+ int page_size;
+ int is_small;
+
+ /* output parameters */
+ u16 act_nr_wqes;
+ u8 act_nr_sges;
+ u32 queue_size; /* bytes for small queues, pages otherwise */
+};
+
struct ehca_alloc_qp_parms {
-/* input parameters */
+ struct ehca_alloc_queue_parms squeue;
+ struct ehca_alloc_queue_parms rqueue;
+
+ /* input parameters */
enum ehca_service_type servicetype;
+ int qp_storage;
int sigtype;
enum ehca_ext_qp_type ext_type;
enum ehca_ll_comp_flags ll_comp_flags;
-
- int max_send_wr, max_recv_wr;
- int max_send_sge, max_recv_sge;
int ud_av_l_key_ctl;
u32 token;
@@ -384,18 +403,10 @@ struct ehca_alloc_qp_parms {
u32 srq_qpn, srq_token, srq_limit;
-/* output parameters */
+ /* output parameters */
u32 real_qp_num;
struct ipz_qp_handle qp_handle;
struct h_galpas galpas;
-
- u16 act_nr_send_wqes;
- u16 act_nr_recv_wqes;
- u8 act_nr_recv_sges;
- u8 act_nr_send_sges;
-
- u32 nr_rq_pages;
- u32 nr_sq_pages;
};
int ehca_cq_assign_qp(struct ehca_cq *cq, struct ehca_qp *qp);
diff --git a/drivers/infiniband/hw/ehca/ehca_cq.c
b/drivers/infiniband/hw/ehca/ehca_cq.c
index 9e87883..5746787 100644
--- a/drivers/infiniband/hw/ehca/ehca_cq.c
+++ b/drivers/infiniband/hw/ehca/ehca_cq.c
@@ -190,8 +190,8 @@ struct ib_cq *ehca_create_cq(struct ib_device *device, int
cqe, int comp_vector,
goto create_cq_exit2;
}
- ipz_rc = ipz_queue_ctor(my_cq-ipz_queue, param.act_pages,
- EHCA_PAGESIZE, sizeof(struct ehca_cqe), 0);
+ ipz_rc = ipz_queue_ctor(NULL, my_cq-ipz_queue, param.act_pages,
+ EHCA_PAGESIZE, sizeof(struct ehca_cqe), 0, 0);
if (!ipz_rc) {
ehca_err(device, ipz_queue_ctor() failed ipz_rc=%x device=%p,
ipz_rc, device);
@@ -285,7 +285,7 @@ struct ib_cq *ehca_create_cq(struct ib_device *device, int
cqe, int comp_vector,
return cq;
create_cq_exit4:
- ipz_queue_dtor(my_cq-ipz_queue);
+ ipz_queue_dtor(NULL, my_cq-ipz_queue);
create_cq_exit3:
h_ret = hipz_h_destroy_cq(adapter_handle, my_cq, 1);
@@ -359,7 +359,7 @@ int ehca_destroy_cq(struct ib_cq *cq
[PATCH 4/5] ehca: Make internal_create/destroy_qp() static
From: Joachim Fenkes [EMAIL PROTECTED]
Date: Thu, 19 Jul 2007 21:40:00 +0200
Subject: [PATCH 4/5] IB/ehca: Make internal_{create,destroy}_qp() static
They're only used in ehca_qp.c
Signed-off-by: Joachim Fenkes [EMAIL PROTECTED]
---
drivers/infiniband/hw/ehca/ehca_qp.c | 17 +
1 files changed, 9 insertions(+), 8 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_qp.c
b/drivers/infiniband/hw/ehca/ehca_qp.c
index 48e9cea..b916d9c 100644
--- a/drivers/infiniband/hw/ehca/ehca_qp.c
+++ b/drivers/infiniband/hw/ehca/ehca_qp.c
@@ -363,10 +363,11 @@ init_qp_queue1:
* the value of the is_srq parameter. If init_attr and srq_init_attr share
* fields, the field out of init_attr is used.
*/
-struct ehca_qp *internal_create_qp(struct ib_pd *pd,
- struct ib_qp_init_attr *init_attr,
- struct ib_srq_init_attr *srq_init_attr,
- struct ib_udata *udata, int is_srq)
+static struct ehca_qp *internal_create_qp(
+ struct ib_pd *pd,
+ struct ib_qp_init_attr *init_attr,
+ struct ib_srq_init_attr *srq_init_attr,
+ struct ib_udata *udata, int is_srq)
{
struct ehca_qp *my_qp;
struct ehca_pd *my_pd = container_of(pd, struct ehca_pd, ib_pd);
@@ -752,8 +753,8 @@ struct ib_qp *ehca_create_qp(struct ib_pd *pd,
return IS_ERR(ret) ? (struct ib_qp *)ret : ret-ib_qp;
}
-int internal_destroy_qp(struct ib_device *dev, struct ehca_qp *my_qp,
- struct ib_uobject *uobject);
+static int internal_destroy_qp(struct ib_device *dev, struct ehca_qp *my_qp,
+ struct ib_uobject *uobject);
struct ib_srq *ehca_create_srq(struct ib_pd *pd,
struct ib_srq_init_attr *srq_init_attr,
@@ -1669,8 +1670,8 @@ query_srq_exit1:
return ret;
}
-int internal_destroy_qp(struct ib_device *dev, struct ehca_qp *my_qp,
- struct ib_uobject *uobject)
+static int internal_destroy_qp(struct ib_device *dev, struct ehca_qp *my_qp,
+ struct ib_uobject *uobject)
{
struct ehca_shca *shca = container_of(dev, struct ehca_shca, ib_device);
struct ehca_pd *my_pd = container_of(my_qp-ib_qp.pd, struct ehca_pd,
--
1.5.2
___
Linuxppc-dev mailing list
Linuxppc-dev@ozlabs.org
https://ozlabs.org/mailman/listinfo/linuxppc-dev
Re: [PATCH 01/10] IB/ehca: Support for multiple event queues
Roland Dreier <[EMAIL PROTECTED]> wrote on 17.07.2007 19:52:55: > At a higher level, I'm left wondering why nobody talked about multiple > EQs during the last months of the 2.6.22 process and now all of a > sudden it becomes urgent in the last few days of the 2.6.23 merge > window. That's not really how I like to merge features OK, let keep multiple eqs for next release with a more stable verbs def. For the other patch to support MR with large pages we'll resend it (without deps on multiple eqs patch) to you soon. Regards Nam - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
[ofa-general] Re: [PATCH 01/10] IB/ehca: Support for multiple event queues
Roland Dreier [EMAIL PROTECTED] wrote on 17.07.2007 19:52:55: At a higher level, I'm left wondering why nobody talked about multiple EQs during the last months of the 2.6.22 process and now all of a sudden it becomes urgent in the last few days of the 2.6.23 merge window. That's not really how I like to merge features OK, let keep multiple eqs for next release with a more stable verbs def. For the other patch to support MR with large pages we'll resend it (without deps on multiple eqs patch) to you soon. Regards Nam ___ general mailing list general@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
Re: [PATCH 01/10] IB/ehca: Support for multiple event queues
Roland Dreier [EMAIL PROTECTED] wrote on 17.07.2007 19:52:55: At a higher level, I'm left wondering why nobody talked about multiple EQs during the last months of the 2.6.22 process and now all of a sudden it becomes urgent in the last few days of the 2.6.23 merge window. That's not really how I like to merge features OK, let keep multiple eqs for next release with a more stable verbs def. For the other patch to support MR with large pages we'll resend it (without deps on multiple eqs patch) to you soon. Regards Nam - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [PATCH 01/10] IB/ehca: Support for multiple event queues
Roland Dreier <[EMAIL PROTECTED]> wrote on 16.07.2007 18:04:26: > Do you have any data on how well this round-robin assignment works? > It seems not quite right to me for the driver to advertise nr_eqs > completion vectors, but then if round-robin is turned on to ignore the > consumer's decision about which vector to use. No, I've no figures to provide here. The background of this dist_eqs option is actually to allow us testing across all event queues without to change the testcases resp consumers to use certain event queue number. Thus, I should comment it as EXPERIMENTAL? > Maybe if round-robin is turned on you should report 0 as the number of > completion vectors? Or maybe we should allow well-known values for > the completion vector passed to ib_create_cq to allow consumers to > specify a policy (like round robin) instead of a particular vector? > Maybe the whole interface is broken and we should only be exposing > policies to consumers instead of the specific vector? Agree in that device driver should not overwrite consumer's policy of event queue assigment. Since dist_eqs is disabled as default, there's no issue, isn't it? Regarding ib_verbs: perhaps we should provide create/destroy_eq() and let upper level protocols or consumers dictate the assignment to cq by passing an event queue pointer to create_cq()... > I think I would rather hold off on multiple EQs for this merge window > and plan on having something really solid and thought-out for 2.6.24. Fair enough. However why don't let us gather experience with this feature now? Should we remove dist_eqs option for more consistency? Thanks Nam - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
[ofa-general] Re: [PATCH 01/10] IB/ehca: Support for multiple event queues
Roland Dreier [EMAIL PROTECTED] wrote on 16.07.2007 18:04:26: Do you have any data on how well this round-robin assignment works? It seems not quite right to me for the driver to advertise nr_eqs completion vectors, but then if round-robin is turned on to ignore the consumer's decision about which vector to use. No, I've no figures to provide here. The background of this dist_eqs option is actually to allow us testing across all event queues without to change the testcases resp consumers to use certain event queue number. Thus, I should comment it as EXPERIMENTAL? Maybe if round-robin is turned on you should report 0 as the number of completion vectors? Or maybe we should allow well-known values for the completion vector passed to ib_create_cq to allow consumers to specify a policy (like round robin) instead of a particular vector? Maybe the whole interface is broken and we should only be exposing policies to consumers instead of the specific vector? Agree in that device driver should not overwrite consumer's policy of event queue assigment. Since dist_eqs is disabled as default, there's no issue, isn't it? Regarding ib_verbs: perhaps we should provide create/destroy_eq() and let upper level protocols or consumers dictate the assignment to cq by passing an event queue pointer to create_cq()... I think I would rather hold off on multiple EQs for this merge window and plan on having something really solid and thought-out for 2.6.24. Fair enough. However why don't let us gather experience with this feature now? Should we remove dist_eqs option for more consistency? Thanks Nam ___ general mailing list general@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
Re: [PATCH 01/10] IB/ehca: Support for multiple event queues
Roland Dreier [EMAIL PROTECTED] wrote on 16.07.2007 18:04:26: Do you have any data on how well this round-robin assignment works? It seems not quite right to me for the driver to advertise nr_eqs completion vectors, but then if round-robin is turned on to ignore the consumer's decision about which vector to use. No, I've no figures to provide here. The background of this dist_eqs option is actually to allow us testing across all event queues without to change the testcases resp consumers to use certain event queue number. Thus, I should comment it as EXPERIMENTAL? Maybe if round-robin is turned on you should report 0 as the number of completion vectors? Or maybe we should allow well-known values for the completion vector passed to ib_create_cq to allow consumers to specify a policy (like round robin) instead of a particular vector? Maybe the whole interface is broken and we should only be exposing policies to consumers instead of the specific vector? Agree in that device driver should not overwrite consumer's policy of event queue assigment. Since dist_eqs is disabled as default, there's no issue, isn't it? Regarding ib_verbs: perhaps we should provide create/destroy_eq() and let upper level protocols or consumers dictate the assignment to cq by passing an event queue pointer to create_cq()... I think I would rather hold off on multiple EQs for this merge window and plan on having something really solid and thought-out for 2.6.24. Fair enough. However why don't let us gather experience with this feature now? Should we remove dist_eqs option for more consistency? Thanks Nam - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [PATCH] fix idr_get_new_above id alias bugs
> With this patch, idr.c should work as advertised allocating id > values in the range 0...0x7fff. Andrew had speculated that > it should allow the full range 0...0x to be used. I was > tempted to make changes to allow this, but it would require changes > to API, e.g. making the starting id value and the return value > unsigned. Hi Jim, thanks much for this patch. It should work fine as far as I can read. Will give it a try in next couple of days. Nam - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
[ofa-general] Re: [PATCH] fix idr_get_new_above id alias bugs
With this patch, idr.c should work as advertised allocating id values in the range 0...0x7fff. Andrew had speculated that it should allow the full range 0...0x to be used. I was tempted to make changes to allow this, but it would require changes to API, e.g. making the starting id value and the return value unsigned. Hi Jim, thanks much for this patch. It should work fine as far as I can read. Will give it a try in next couple of days. Nam ___ general mailing list general@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
Re: [PATCH] fix idr_get_new_above id alias bugs
With this patch, idr.c should work as advertised allocating id values in the range 0...0x7fff. Andrew had speculated that it should allow the full range 0...0x to be used. I was tempted to make changes to allow this, but it would require changes to API, e.g. making the starting id value and the return value unsigned. Hi Jim, thanks much for this patch. It should work fine as far as I can read. Will give it a try in next couple of days. Nam - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: idr_get_new_above() limitation?
On Tuesday 03 July 2007 02:31, Jim Houston wrote:
> The problem is in idr_get_new_above_int() in the loop which
> adds new layers to the top of the radix tree. It is failing
> the "layers < (MAX_LEVEL - 1)" test. It doesn't allocate the
> new layer but still calls sub_alloc() which relies on having
> the new layer properly constructed. I believe that it is
> allocating the slot which corresponds to id = 0.
Hi Jim,
Thanks for your quick reply.
Yes, I realized that while condition too and have tried with a tiny
change like (layers < MAX_LEVEL), but without success with idr_find(),
even though 6 layers were created and the object was added at proper
location. After several debug cycles I think to find the root cause
in the if-condition in idr_find():
void *idr_find(struct idr *idp, int id)
{
int n;
struct idr_layer *p;
n = idp->layers * IDR_BITS;
p = idp->top;
/* Mask off upper bits we don't use for the search. */
id &= MAX_ID_MASK;
if (id >= (1 << n))
return NULL;
...
}
Since idp->layers is now 6, n is equal 36, ie out of 32-bit-range,
and therefore
(1 << n) = (1 << 36) = 0
causing that if-cond to be true ie idr_find() fails.
Replacing that if-line by
if ((long)id >= (1L << n))
makes idr_find() working properly until MAX_ID_MASK.
Since there are other places to be changed like above as well eg.
idr_replace() and because you're creating a patch too, I'm waiting
first for your comment. Let me know if you prefer me to send a
patch.
Regards
Nam
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[ofa-general] Re: idr_get_new_above() limitation?
On Tuesday 03 July 2007 02:31, Jim Houston wrote:
The problem is in idr_get_new_above_int() in the loop which
adds new layers to the top of the radix tree. It is failing
the layers (MAX_LEVEL - 1) test. It doesn't allocate the
new layer but still calls sub_alloc() which relies on having
the new layer properly constructed. I believe that it is
allocating the slot which corresponds to id = 0.
Hi Jim,
Thanks for your quick reply.
Yes, I realized that while condition too and have tried with a tiny
change like (layers MAX_LEVEL), but without success with idr_find(),
even though 6 layers were created and the object was added at proper
location. After several debug cycles I think to find the root cause
in the if-condition in idr_find():
void *idr_find(struct idr *idp, int id)
{
int n;
struct idr_layer *p;
n = idp-layers * IDR_BITS;
p = idp-top;
/* Mask off upper bits we don't use for the search. */
id = MAX_ID_MASK;
if (id = (1 n))
return NULL;
...
}
Since idp-layers is now 6, n is equal 36, ie out of 32-bit-range,
and therefore
(1 n) = (1 36) = 0
causing that if-cond to be true ie idr_find() fails.
Replacing that if-line by
if ((long)id = (1L n))
makes idr_find() working properly until MAX_ID_MASK.
Since there are other places to be changed like above as well eg.
idr_replace() and because you're creating a patch too, I'm waiting
first for your comment. Let me know if you prefer me to send a
patch.
Regards
Nam
___
general mailing list
general@lists.openfabrics.org
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
Re: idr_get_new_above() limitation?
On Tuesday 03 July 2007 02:31, Jim Houston wrote:
The problem is in idr_get_new_above_int() in the loop which
adds new layers to the top of the radix tree. It is failing
the layers (MAX_LEVEL - 1) test. It doesn't allocate the
new layer but still calls sub_alloc() which relies on having
the new layer properly constructed. I believe that it is
allocating the slot which corresponds to id = 0.
Hi Jim,
Thanks for your quick reply.
Yes, I realized that while condition too and have tried with a tiny
change like (layers MAX_LEVEL), but without success with idr_find(),
even though 6 layers were created and the object was added at proper
location. After several debug cycles I think to find the root cause
in the if-condition in idr_find():
void *idr_find(struct idr *idp, int id)
{
int n;
struct idr_layer *p;
n = idp-layers * IDR_BITS;
p = idp-top;
/* Mask off upper bits we don't use for the search. */
id = MAX_ID_MASK;
if (id = (1 n))
return NULL;
...
}
Since idp-layers is now 6, n is equal 36, ie out of 32-bit-range,
and therefore
(1 n) = (1 36) = 0
causing that if-cond to be true ie idr_find() fails.
Replacing that if-line by
if ((long)id = (1L n))
makes idr_find() working properly until MAX_ID_MASK.
Since there are other places to be changed like above as well eg.
idr_replace() and because you're creating a patch too, I'm waiting
first for your comment. Let me know if you prefer me to send a
patch.
Regards
Nam
-
To unsubscribe from this list: send the line unsubscribe linux-kernel in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
idr_get_new_above() limitation?
Hello,
For ehca device driver we're intending to utilize
idr_get_new_above() and have written a test case, which I'm attaching
at the end. Basically it tries to get an idr token above a lower boundary
by calling idr_get_new_above() and then uses idr_find() to check if
the returned token can be found.
Here is our observation with 2.6.22-rc7 on ppc64:
Use lower boundary 0x3ffc
[EMAIL PROTECTED] idr_bug]# insmod idr_test_mod.ko start=1073741820
insmod: error inserting 'idr_test_mod.ko': -1 Unknown symbol in module
[EMAIL PROTECTED] idr_bug]# dmesg -c
i=3ffc token=3ffc t=3ffc
i=3ffd token=3ffd t=3ffd
i=3ffe token=3ffe t=3ffe
i=3fff token=3fff t=3fff
i=4000 token=4000 t=
Invalid object . Expected 4000
That means token 0x4000 seems to be the "upper boundary" of idr_find().
However the behaviour is not consistent in that it was returned by
idr_get_new_above().
Looking at void *idr_find(struct idr *idp, int id)
{
int n;
struct idr_layer *p;
n = idp->layers * IDR_BITS;
p = idp->top;
/* Mask off upper bits we don't use for the search. */
id &= MAX_ID_MASK;
if (id >= (1 << n))
return NULL;
while (n > 0 && p) {
n -= IDR_BITS;
p = p->ary[(id >> n) & IDR_MASK];
}
return((void *)p);
}
we found that the if-condition has failed:
layers = 5
IDR_BITS = 6
n = 30
(id >= (1 << n)) = (0x4000 >= 0x4000) = 1
Since MAX_ID_MASK=0x7fff, I'm wondering if 0x4000 is the actual
upper boundary. Any hints or suggestions are appreciated.
Thanks!
Nam
#include
#include
MODULE_LICENSE("GPL");
int start_opt = 0x7e00;
module_param_named(start, start_opt, int, 0);
MODULE_PARM_DESC(start,
"Start token for idr_get_new_above(). Default 0x7e00");
static int __init idr_test_init(void)
{
DEFINE_IDR(idr);
int token, ret;
unsigned long i;
for (i = start_opt; i <= MAX_ID_MASK; i++) {
void * t;
if (!idr_pre_get(, GFP_KERNEL)) {
printk(KERN_ERR "ERROR: Out of mem\n");
return -ENOENT;
}
ret = idr_get_new_above(, (void*)i, start_opt, );
switch (ret) {
case 0:
t = idr_find(, token);
printk(KERN_ERR "i=%lx token=%x t=%p\n", i, token, t);
if (t != (void*)i) {
printk(KERN_ERR "Invalid object %p. Expected
%lx\n",
t, i);
return -ENOENT;
}
break;
case -EAGAIN:
i--;
printk("idr_get_new_above() ret=-EAGAIN\n");
break;
default:
printk(KERN_ERR "ERROR: Out of mem\n");
break;
}
}
/*
* return an error in any case since we don't need the module
* loaded anyway.
*/
return -ENOENT;
}
static void __exit idr_test_exit(void)
{
printk(KERN_ERR "module exit\n");
}
module_init(idr_test_init);
module_exit(idr_test_exit);
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[ofa-general] idr_get_new_above() limitation?
Hello,
For ehca device driver we're intending to utilize
idr_get_new_above() and have written a test case, which I'm attaching
at the end. Basically it tries to get an idr token above a lower boundary
by calling idr_get_new_above() and then uses idr_find() to check if
the returned token can be found.
Here is our observation with 2.6.22-rc7 on ppc64:
Use lower boundary 0x3ffc
[EMAIL PROTECTED] idr_bug]# insmod idr_test_mod.ko start=1073741820
insmod: error inserting 'idr_test_mod.ko': -1 Unknown symbol in module
[EMAIL PROTECTED] idr_bug]# dmesg -c
i=3ffc token=3ffc t=3ffc
i=3ffd token=3ffd t=3ffd
i=3ffe token=3ffe t=3ffe
i=3fff token=3fff t=3fff
i=4000 token=4000 t=
Invalid object . Expected 4000
That means token 0x4000 seems to be the upper boundary of idr_find().
However the behaviour is not consistent in that it was returned by
idr_get_new_above().
Looking at void *idr_find(struct idr *idp, int id)
{
int n;
struct idr_layer *p;
n = idp-layers * IDR_BITS;
p = idp-top;
/* Mask off upper bits we don't use for the search. */
id = MAX_ID_MASK;
if (id = (1 n))
return NULL;
while (n 0 p) {
n -= IDR_BITS;
p = p-ary[(id n) IDR_MASK];
}
return((void *)p);
}
we found that the if-condition has failed:
layers = 5
IDR_BITS = 6
n = 30
(id = (1 n)) = (0x4000 = 0x4000) = 1
Since MAX_ID_MASK=0x7fff, I'm wondering if 0x4000 is the actual
upper boundary. Any hints or suggestions are appreciated.
Thanks!
Nam
#include linux/module.h
#include linux/idr.h
MODULE_LICENSE(GPL);
int start_opt = 0x7e00;
module_param_named(start, start_opt, int, 0);
MODULE_PARM_DESC(start,
Start token for idr_get_new_above(). Default 0x7e00);
static int __init idr_test_init(void)
{
DEFINE_IDR(idr);
int token, ret;
unsigned long i;
for (i = start_opt; i = MAX_ID_MASK; i++) {
void * t;
if (!idr_pre_get(idr, GFP_KERNEL)) {
printk(KERN_ERR ERROR: Out of mem\n);
return -ENOENT;
}
ret = idr_get_new_above(idr, (void*)i, start_opt, token);
switch (ret) {
case 0:
t = idr_find(idr, token);
printk(KERN_ERR i=%lx token=%x t=%p\n, i, token, t);
if (t != (void*)i) {
printk(KERN_ERR Invalid object %p. Expected
%lx\n,
t, i);
return -ENOENT;
}
break;
case -EAGAIN:
i--;
printk(idr_get_new_above() ret=-EAGAIN\n);
break;
default:
printk(KERN_ERR ERROR: Out of mem\n);
break;
}
}
/*
* return an error in any case since we don't need the module
* loaded anyway.
*/
return -ENOENT;
}
static void __exit idr_test_exit(void)
{
printk(KERN_ERR module exit\n);
}
module_init(idr_test_init);
module_exit(idr_test_exit);
___
general mailing list
general@lists.openfabrics.org
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
idr_get_new_above() limitation?
Hello,
For ehca device driver we're intending to utilize
idr_get_new_above() and have written a test case, which I'm attaching
at the end. Basically it tries to get an idr token above a lower boundary
by calling idr_get_new_above() and then uses idr_find() to check if
the returned token can be found.
Here is our observation with 2.6.22-rc7 on ppc64:
Use lower boundary 0x3ffc
[EMAIL PROTECTED] idr_bug]# insmod idr_test_mod.ko start=1073741820
insmod: error inserting 'idr_test_mod.ko': -1 Unknown symbol in module
[EMAIL PROTECTED] idr_bug]# dmesg -c
i=3ffc token=3ffc t=3ffc
i=3ffd token=3ffd t=3ffd
i=3ffe token=3ffe t=3ffe
i=3fff token=3fff t=3fff
i=4000 token=4000 t=
Invalid object . Expected 4000
That means token 0x4000 seems to be the upper boundary of idr_find().
However the behaviour is not consistent in that it was returned by
idr_get_new_above().
Looking at void *idr_find(struct idr *idp, int id)
{
int n;
struct idr_layer *p;
n = idp-layers * IDR_BITS;
p = idp-top;
/* Mask off upper bits we don't use for the search. */
id = MAX_ID_MASK;
if (id = (1 n))
return NULL;
while (n 0 p) {
n -= IDR_BITS;
p = p-ary[(id n) IDR_MASK];
}
return((void *)p);
}
we found that the if-condition has failed:
layers = 5
IDR_BITS = 6
n = 30
(id = (1 n)) = (0x4000 = 0x4000) = 1
Since MAX_ID_MASK=0x7fff, I'm wondering if 0x4000 is the actual
upper boundary. Any hints or suggestions are appreciated.
Thanks!
Nam
#include linux/module.h
#include linux/idr.h
MODULE_LICENSE(GPL);
int start_opt = 0x7e00;
module_param_named(start, start_opt, int, 0);
MODULE_PARM_DESC(start,
Start token for idr_get_new_above(). Default 0x7e00);
static int __init idr_test_init(void)
{
DEFINE_IDR(idr);
int token, ret;
unsigned long i;
for (i = start_opt; i = MAX_ID_MASK; i++) {
void * t;
if (!idr_pre_get(idr, GFP_KERNEL)) {
printk(KERN_ERR ERROR: Out of mem\n);
return -ENOENT;
}
ret = idr_get_new_above(idr, (void*)i, start_opt, token);
switch (ret) {
case 0:
t = idr_find(idr, token);
printk(KERN_ERR i=%lx token=%x t=%p\n, i, token, t);
if (t != (void*)i) {
printk(KERN_ERR Invalid object %p. Expected
%lx\n,
t, i);
return -ENOENT;
}
break;
case -EAGAIN:
i--;
printk(idr_get_new_above() ret=-EAGAIN\n);
break;
default:
printk(KERN_ERR ERROR: Out of mem\n);
break;
}
}
/*
* return an error in any case since we don't need the module
* loaded anyway.
*/
return -ENOENT;
}
static void __exit idr_test_exit(void)
{
printk(KERN_ERR module exit\n);
}
module_init(idr_test_init);
module_exit(idr_test_exit);
-
To unsubscribe from this list: send the line unsubscribe linux-kernel in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
idr_get_new_above() limitation?
Hello,
For ehca device driver we're intending to utilize
idr_get_new_above() and have written a test case, which I'm attaching
at the end. Basically it tries to get an idr token above a lower boundary
by calling idr_get_new_above() and then uses idr_find() to check if
the returned token can be found.
Here is our observation with 2.6.22-rc7 on ppc64:
Use lower boundary 0x3ffc
[EMAIL PROTECTED] idr_bug]# insmod idr_test_mod.ko start=1073741820
insmod: error inserting 'idr_test_mod.ko': -1 Unknown symbol in module
[EMAIL PROTECTED] idr_bug]# dmesg -c
i=3ffc token=3ffc t=3ffc
i=3ffd token=3ffd t=3ffd
i=3ffe token=3ffe t=3ffe
i=3fff token=3fff t=3fff
i=4000 token=4000 t=
Invalid object . Expected 4000
That means token 0x4000 seems to be the upper boundary of idr_find().
However the behaviour is not consistent in that it was returned by
idr_get_new_above().
Looking at void *idr_find(struct idr *idp, int id)
{
int n;
struct idr_layer *p;
n = idp-layers * IDR_BITS;
p = idp-top;
/* Mask off upper bits we don't use for the search. */
id = MAX_ID_MASK;
if (id = (1 n))
return NULL;
while (n 0 p) {
n -= IDR_BITS;
p = p-ary[(id n) IDR_MASK];
}
return((void *)p);
}
we found that the if-condition has failed:
layers = 5
IDR_BITS = 6
n = 30
(id = (1 n)) = (0x4000 = 0x4000) = 1
Since MAX_ID_MASK=0x7fff, I'm wondering if 0x4000 is the actual
upper boundary. Any hints or suggestions are appreciated.
Thanks!
Nam
#include linux/module.h
#include linux/idr.h
MODULE_LICENSE(GPL);
int start_opt = 0x7e00;
module_param_named(start, start_opt, int, 0);
MODULE_PARM_DESC(start,
Start token for idr_get_new_above(). Default 0x7e00);
static int __init idr_test_init(void)
{
DEFINE_IDR(idr);
int token, ret;
unsigned long i;
for (i = start_opt; i = MAX_ID_MASK; i++) {
void * t;
if (!idr_pre_get(idr, GFP_KERNEL)) {
printk(KERN_ERR ERROR: Out of mem\n);
return -ENOENT;
}
ret = idr_get_new_above(idr, (void*)i, start_opt, token);
switch (ret) {
case 0:
t = idr_find(idr, token);
printk(KERN_ERR i=%lx token=%x t=%p\n, i, token, t);
if (t != (void*)i) {
printk(KERN_ERR Invalid object %p. Expected
%lx\n,
t, i);
return -ENOENT;
}
break;
case -EAGAIN:
i--;
printk(idr_get_new_above() ret=-EAGAIN\n);
break;
default:
printk(KERN_ERR ERROR: Out of mem\n);
break;
}
}
/*
* return an error in any case since we don't need the module
* loaded anyway.
*/
return -ENOENT;
}
static void __exit idr_test_exit(void)
{
printk(KERN_ERR module exit\n);
}
module_init(idr_test_init);
module_exit(idr_test_exit);
___
Linuxppc-dev mailing list
Linuxppc-dev@ozlabs.org
https://ozlabs.org/mailman/listinfo/linuxppc-dev
Use of Commission
Hi list Sorry to post this question again. I've tried to use admin and DemoRepAll users to place orders to DemoRepAll and other users but I did not see any Commission Invoice generated at all. Is there a step-by-step usage example which I can take a look? Thanks Nam
[jira] Commented: (OFBIZ-1087) Order Ship to Another Address does not take in that address
[ https://issues.apache.org/jira/browse/OFBIZ-1087?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12507741 ] Nam Nguyen commented on OFBIZ-1087: --- I suppose it was a Sales Order. I visited Order Manager, and click on Order Entry. Order Ship to Another Address does not take in that address --- Key: OFBIZ-1087 URL: https://issues.apache.org/jira/browse/OFBIZ-1087 Project: OFBiz Issue Type: Bug Components: order Affects Versions: SVN trunk Environment: Windows XP SP2, JDK 1.6.0, Browser Opera Reporter: Nam Nguyen In the shipping stage of order entry, I picked ship to different address, selected a party, and chose one of the addresses of that party. Upon clicking Continue, the page goes back to say Please enter a shipping address. -- This message is automatically generated by JIRA. - You can reply to this email to add a comment to the issue online.
[ofa-general] Re: [ewg] OFED 1.2 rc5 release
Hi, I'm having troubles to reach www.openfabrics.org resp to download ofed-1.2-rc5. Do I need to consider something else? Thanks! Mit freundlichen Gruessen/Kind Regards Hoang-Nam Nguyen Tel. +49-7031-16-3570, email: [EMAIL PROTECTED] IBM Deutschland Entwicklung GmbH Vorsitzender des Aufsichtsrats: Martin Jetter Geschaeftsfuehrung: Herbert Kircher Sitz der Gesellschaft: Boeblingen Registergericht: Amtsgericht Stuttgart, HRB 243294 Tziporet Koren [EMAIL PROTECTED] x.co.il To Sent by: [EMAIL PROTECTED] [EMAIL PROTECTED] cc .openfabrics.org general@lists.openfabrics.org Subject [ewg] OFED 1.2 rc5 release 13.06.2007 16:25 Hi, OFED 1.2-RC5 is available on http://www.openfabrics.org/builds/ofed-1.2/ File: OFED-1.2-rc5.tgz To get BUILD_ID run ofed_info Please report any issues in bugzilla https://bugs.openfabrics.org/ The GA release is expected next Wed (June 20) based on RC5 tests Tziporet Vlad Release information: OS support: Novell: - SLES 9.0 SP3 - SLES10 - SLES10 SP1 RC5 Redhat: - Redhat EL4 up3, up4 and up5 - Redhat EL5 kernel.org: - 2.6.20 - 2.6.19 Note: Fedora C6 and SuSE Pro 10 are not part of the official list. We keep the backport patches for these OSes and make sure OFED compile and loaded properly but will not do full QA cycle. Systems: * x86_64 * x86 * ia64 * ppc64 Main changes from OFED-1.1-rc4: === 1. Fixed 8 bugs (see attached for fixed issues) 2. Added support for SLES10 SP1 RC5 (tvflash is disabled for now) 3. Added support for iSER on RHEL 4 4. Updated documents - all owners please review to make sure docs of your component is updated. See bugzilla for all open issues. Tasks that should be completed for the GA release: 1. Complete all documentation (release notes, README, etc.) 2. Run all QA tests on all platforms (See attached file: rc5_fixed_bugs.csv) ___ ewg mailing list [EMAIL PROTECTED] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg rc5_fixed_bugs.csv Description: Binary data ___ general mailing list general@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
Re: [ewg] OFED 1.2 rc5 release
Hi, I'm having troubles to reach www.openfabrics.org resp to download ofed-1.2-rc5. Do I need to consider something else? Thanks! Mit freundlichen Gruessen/Kind Regards Hoang-Nam Nguyen Tel. +49-7031-16-3570, email: [EMAIL PROTECTED] IBM Deutschland Entwicklung GmbH Vorsitzender des Aufsichtsrats: Martin Jetter Geschaeftsfuehrung: Herbert Kircher Sitz der Gesellschaft: Boeblingen Registergericht: Amtsgericht Stuttgart, HRB 243294 Tziporet Koren [EMAIL PROTECTED] x.co.il To Sent by: ewg@lists.openfabrics.org [EMAIL PROTECTED] cc .openfabrics.org [EMAIL PROTECTED] Subject [ewg] OFED 1.2 rc5 release 13.06.2007 16:25 Hi, OFED 1.2-RC5 is available on http://www.openfabrics.org/builds/ofed-1.2/ File: OFED-1.2-rc5.tgz To get BUILD_ID run ofed_info Please report any issues in bugzilla https://bugs.openfabrics.org/ The GA release is expected next Wed (June 20) based on RC5 tests Tziporet Vlad Release information: OS support: Novell: - SLES 9.0 SP3 - SLES10 - SLES10 SP1 RC5 Redhat: - Redhat EL4 up3, up4 and up5 - Redhat EL5 kernel.org: - 2.6.20 - 2.6.19 Note: Fedora C6 and SuSE Pro 10 are not part of the official list. We keep the backport patches for these OSes and make sure OFED compile and loaded properly but will not do full QA cycle. Systems: * x86_64 * x86 * ia64 * ppc64 Main changes from OFED-1.1-rc4: === 1. Fixed 8 bugs (see attached for fixed issues) 2. Added support for SLES10 SP1 RC5 (tvflash is disabled for now) 3. Added support for iSER on RHEL 4 4. Updated documents - all owners please review to make sure docs of your component is updated. See bugzilla for all open issues. Tasks that should be completed for the GA release: 1. Complete all documentation (release notes, README, etc.) 2. Run all QA tests on all platforms (See attached file: rc5_fixed_bugs.csv) ___ ewg mailing list ewg@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg rc5_fixed_bugs.csv Description: Binary data ___ ewg mailing list ewg@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg
Re: [ewg] ehca build fails on SLES 10 sp1 ppc
Hm, I could build rc6 on several ppc64 nodes with ehca without any problems. My uname -r says this: 2.6.16.21-0.8-ppc64 which should be sles10sp1, while yours seems to be newer according to your log, ie. /usr/src/linux-2.6.16.46-0.16-obj. Can you send me your ofed.conf, uname -r and a link to linux-2.6.16.46-0.16-obj dir (a copy of include/asm/hvcall.h would be sufficient)? Thanks Nam [EMAIL PROTECTED] wrote on 21.06.2007 19:23:34: I'm not sure if it is intended to be supported at this time but it fails to build. See attached log. [attachment ehca-build.log deleted by Hoang-Nam Nguyen/Germany/IBM] ___ ewg mailing list ewg@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg ___ ewg mailing list ewg@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg
How to Use Commission?
Hi list Is there a step-by-step example to make use of Commission? Cheers Nam
[jira] Created: (OFBIZ-1087) Order Ship to Another Address does not take in that address
Order Ship to Another Address does not take in that address --- Key: OFBIZ-1087 URL: https://issues.apache.org/jira/browse/OFBIZ-1087 Project: OFBiz Issue Type: Bug Components: order Affects Versions: SVN trunk Environment: Windows XP SP2, JDK 1.6.0, Browser Opera Reporter: Nam Nguyen In the shipping stage of order entry, I picked ship to different address, selected a party, and chose one of the addresses of that party. Upon clicking Continue, the page goes back to say Please enter a shipping address. -- This message is automatically generated by JIRA. - You can reply to this email to add a comment to the issue online.
[laptop-discuss] Protege R100 Wireless not working
brian.xu wrote: Prameet Chhabra wrote: Wirelessers, I have tried all the instructions on the http://www.sun.com/bigadmin/hcl/data/components/details/1206.html and http://www.opensolaris.org/os/community/laptop/wireless/iwi/ to enable wireless on my laptopb but all in vain... any help will be appricated. Which Solaris build does your laptop run? The iwi was integrated into snv_62, so suppose it should work by default if you run snv_62+. As to your failure, pls check: 1. The wifi chip is Intel 2200/2915 2. You get the correct version of driver from opensolaris.org if you run solaris other than snv_62+ 3. One thing that is always easy to neglect is the wifi button on the panel. Make sure it is open. Just curious, which panel you are mentioning here? Thanks, -Nam If it still doesn't work, then pls be more specific on where it fails and be better to provide the output of 'dmesg' '/usr/X11/bin/scanpci -v' 'modinfo |grep iwi' --Brian thanks Prameet ___ laptop-discuss mailing list laptop-discuss at opensolaris.org ___ laptop-discuss mailing list laptop-discuss at opensolaris.org
[PATCH 2.6.22] ehca: return proper error code if register_mr fails
This patch sets the return code of ehca_register_mr() to ENOMEM
if corresponding firmware call fails due to out of resources.
Some of error codes were mapped to EINVAL. They are now mapped
to default case, which already returns EINVAL anyway.
Signed-off-by: Hoang-Nam Nguyen <[EMAIL PROTECTED]>
---
ehca_mrmw.c |7 ++-
1 file changed, 2 insertions(+), 5 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_mrmw.c
b/drivers/infiniband/hw/ehca/ehca_mrmw.c
index 84c5bb4..add79bd 100644
--- a/drivers/infiniband/hw/ehca/ehca_mrmw.c
+++ b/drivers/infiniband/hw/ehca/ehca_mrmw.c
@@ -2050,13 +2050,10 @@ int ehca_mrmw_map_hrc_alloc(const u64 hi
switch (hipz_rc) {
case H_SUCCESS: /* successful completion */
return 0;
- case H_ADAPTER_PARM: /* invalid adapter handle */
- case H_RT_PARM: /* invalid resource type */
case H_NOT_ENOUGH_RESOURCES: /* insufficient resources */
- case H_MLENGTH_PARM: /* invalid memory length */
- case H_MEM_ACCESS_PARM: /* invalid access controls */
case H_CONSTRAINED: /* resource constraint */
- return -EINVAL;
+ case H_NO_MEM:
+ return -ENOMEM;
case H_BUSY: /* long busy */
return -EBUSY;
default:
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[ofa-general] [PATCH 2.6.22] ehca: return proper error code if register_mr fails
This patch sets the return code of ehca_register_mr() to ENOMEM
if corresponding firmware call fails due to out of resources.
Some of error codes were mapped to EINVAL. They are now mapped
to default case, which already returns EINVAL anyway.
Signed-off-by: Hoang-Nam Nguyen [EMAIL PROTECTED]
---
ehca_mrmw.c |7 ++-
1 file changed, 2 insertions(+), 5 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_mrmw.c
b/drivers/infiniband/hw/ehca/ehca_mrmw.c
index 84c5bb4..add79bd 100644
--- a/drivers/infiniband/hw/ehca/ehca_mrmw.c
+++ b/drivers/infiniband/hw/ehca/ehca_mrmw.c
@@ -2050,13 +2050,10 @@ int ehca_mrmw_map_hrc_alloc(const u64 hi
switch (hipz_rc) {
case H_SUCCESS: /* successful completion */
return 0;
- case H_ADAPTER_PARM: /* invalid adapter handle */
- case H_RT_PARM: /* invalid resource type */
case H_NOT_ENOUGH_RESOURCES: /* insufficient resources */
- case H_MLENGTH_PARM: /* invalid memory length */
- case H_MEM_ACCESS_PARM: /* invalid access controls */
case H_CONSTRAINED: /* resource constraint */
- return -EINVAL;
+ case H_NO_MEM:
+ return -ENOMEM;
case H_BUSY: /* long busy */
return -EBUSY;
default:
___
general mailing list
general@lists.openfabrics.org
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
[ofa-general] [PATCH ofed-1.2] ehca (kernel space): return proper error code if register_mr fails
Hello Tziporet!
Please accept below patch for ofed-1.2, because it fixes a mr resources
limitation problem reported by Troy and Kyle on this mailing list. Only
with this patch their application is able to release no longer used mrs
properly.
Thanks!
Regards
Nam
This patch sets the return code of ehca_register_mr() to ENOMEM
if corresponding firmware call fails due to out of resources.
Some of error codes were mapped to EINVAL. They are now mapped
to default case, which already returns EINVAL anyway.
Signed-off-by: Hoang-Nam Nguyen [EMAIL PROTECTED]
---
ehca_mrmw.c |7 ++-
1 file changed, 2 insertions(+), 5 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_mrmw.c
b/drivers/infiniband/hw/ehca/ehca_mrmw.c
index cfb362a..b3bbe3b 100644
--- a/drivers/infiniband/hw/ehca/ehca_mrmw.c
+++ b/drivers/infiniband/hw/ehca/ehca_mrmw.c
@@ -2045,13 +2045,10 @@ int ehca_mrmw_map_hrc_alloc(const u64 hi
switch (hipz_rc) {
case H_SUCCESS: /* successful completion */
return 0;
- case H_ADAPTER_PARM: /* invalid adapter handle */
- case H_RT_PARM: /* invalid resource type */
case H_NOT_ENOUGH_RESOURCES: /* insufficient resources */
- case H_MLENGTH_PARM: /* invalid memory length */
- case H_MEM_ACCESS_PARM: /* invalid access controls */
case H_CONSTRAINED: /* resource constraint */
- return -EINVAL;
+ case H_NO_MEM:
+ return -ENOMEM;
case H_BUSY: /* long busy */
return -EBUSY;
default:
___
general mailing list
general@lists.openfabrics.org
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
[PATCH 2.6.22] ehca: return proper error code if register_mr fails
This patch sets the return code of ehca_register_mr() to ENOMEM
if corresponding firmware call fails due to out of resources.
Some of error codes were mapped to EINVAL. They are now mapped
to default case, which already returns EINVAL anyway.
Signed-off-by: Hoang-Nam Nguyen [EMAIL PROTECTED]
---
ehca_mrmw.c |7 ++-
1 file changed, 2 insertions(+), 5 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_mrmw.c
b/drivers/infiniband/hw/ehca/ehca_mrmw.c
index 84c5bb4..add79bd 100644
--- a/drivers/infiniband/hw/ehca/ehca_mrmw.c
+++ b/drivers/infiniband/hw/ehca/ehca_mrmw.c
@@ -2050,13 +2050,10 @@ int ehca_mrmw_map_hrc_alloc(const u64 hi
switch (hipz_rc) {
case H_SUCCESS: /* successful completion */
return 0;
- case H_ADAPTER_PARM: /* invalid adapter handle */
- case H_RT_PARM: /* invalid resource type */
case H_NOT_ENOUGH_RESOURCES: /* insufficient resources */
- case H_MLENGTH_PARM: /* invalid memory length */
- case H_MEM_ACCESS_PARM: /* invalid access controls */
case H_CONSTRAINED: /* resource constraint */
- return -EINVAL;
+ case H_NO_MEM:
+ return -ENOMEM;
case H_BUSY: /* long busy */
return -EBUSY;
default:
-
To unsubscribe from this list: send the line unsubscribe linux-kernel in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Re: [ofa-general] [PATCH 1 of 3] IB/verbs: add cq comp_vector support in core
Hello Michael and Roland! How about a new verb delivering number of cqs associated with a comp_vector like this /** * Returns number of cqs assigned to comp_vector * @return 0 in error case eg invalid comp_vector */ int ib_query_comp_vector(struct ib_device *dev, int comp_vector); A consumer or ULP would be able to pick an empty comp_vector. Surely that does not prevent that a certain comp_vector resp. IRQ can be overloaded, and that's another topic. Thanks Nam PS: I'm waiting for your comments first, a patch will come later. I'm not convinced it's an interesting metric. A CQ which has multiple QPs assigned to it might get more traffic than a CQ which only has a single QP. That's true for association between CQ and QPs. My gut feeling would be that once ULPs learn to use multiple vectors, each ULPs will spread across them evenly, without help from provider. Enabling multiple vectors is to be done by a provider. ULP can utilize it by checking num_comp_vector. Above simple metric could be provided in ib_core. Anyway, as Shirley stated in her email, using comp_vector per port will help a lot, at least on ppc64 and with ehca - for other HCAs I haven't benchmarked. And that metric allows ULP to implement such one approach. Nam ___ general mailing list general@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
Re: [ofa-general] [PATCH 1 of 3] IB/verbs: add cq comp_vector support in core
Michael S. Tsirkin [EMAIL PROTECTED] wrote on 04.05.2007 11:13:06: using comp_vector per port will help a lot, at least on ppc64 and with ehca - for other HCAs I haven't benchmarked. And that metric allows ULP to implement such one approach. Looks like a bit of overdesign. I think you can just set comp_vector = port_num - 1 % num_comp_vectors without any special metrics Right, looks much simpler for that per-port-purpose. Nam ___ general mailing list general@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
Test
Hello, Could somebody tell me whether you receive my email ? My collegue in this mailing list, does not receive my email so I 'd like check it. Thanks ! Nam This message and any attachments (the message) is intended solely for the addressees and is confidential. If you receive this message in error, please delete it and immediately notify the sender. Any use not in accord with its purpose, any dissemination or disclosure, either whole or partial, is prohibited except formal approval. The internet can not guarantee the integrity of this message. BNP PARIBAS (and its subsidiaries) shall (will) not therefore be liable for the message if modified. - Ce message et toutes les pieces jointes (ci-apres le message) sont etablis a l'intention exclusive de ses destinataires et sont confidentiels. Si vous recevez ce message par erreur, merci de le detruire et d'en avertir immediatement l'expediteur. Toute utilisation de ce message non conforme a sa destination, toute diffusion ou toute publication, totale ou partielle, est interdite, sauf autorisation expresse. L'internet ne permettant pas d'assurer l'integrite de ce message, BNP PARIBAS (et ses filiales) decline(nt) toute responsabilite au titre de ce message, dans l'hypothese ou il aurait ete modifie.
Re: [ofa-general] [PATCH][RFC] IB: Return "maybe missed event" hint from ib_req_notify_cq()
Hi Roland!
As far as this concerns ehca this looks great.
Thanks
Nam
[EMAIL PROTECTED] wrote on 27.04.2007 00:43:19:
> > - "IB: Return "maybe missed event" hint from ib_req_notify_cq()"
> > This extends the API in a way that lets us implement NAPI, but may
> > be useful for other things too. It touches all the drivers, and I
> > still need to finish updating cxgb3 to work correctly. I haven't
> > heard anything negative about this, so I'll fix it up, post it one
> > more time for review, and plan on merging it.
>
> As promised, here is that patch for review, with a cxgb3
> implementation included.
>
> ---
>
> The semantics defined by the InfiniBand specification say that
> completion events are only generated when a completions is added to a
> completion queue (CQ) after completion notification is requested. In
> other words, this means that the following race is possible:
>
>while (CQ is not empty)
> ib_poll_cq(CQ);
>// new completion is added after while loop is exited
>ib_req_notify_cq(CQ);
>// no event is generated for the existing completion
>
> To close this race, the IB spec recommends doing another poll of the
> CQ after requesting notification.
>
> However, it is not always possible to arrange code this way (for
> example, we have found that NAPI for IPoIB cannot poll after
> requesting notification). Also, some hardware (eg Mellanox HCAs)
> actually will generate an event for completions added before the call
> to ib_req_notify_cq() -- which is allowed by the spec, since there's
> no way for any upper-layer consumer to know exactly when a completion
> was really added -- so the extra poll of the CQ is just a waste.
>
> Motivated by this, we add a new flag "IB_CQ_REPORT_MISSED_EVENTS" for
> ib_req_notify_cq() so that it can return a hint about whether the a
> completion may have been added before the request for notification.
> The return value of ib_req_notify_cq() is extended so:
>
> < 0 means an error occurred while requesting notification
>== 0 means notification was requested successfully, and if
> IB_CQ_REPORT_MISSED_EVENTS was passed in, then no
> events were missed and it is safe to wait for another
> event.
> > 0 is only returned if IB_CQ_REPORT_MISSED_EVENTS was
> passed in. It means that the consumer must poll the
> CQ again to make sure it is empty to avoid the race
> described above.
>
> We add a flag to enable this behavior rather than turning it on
> unconditionally, because checking for missed events may incur
> significant overhead for some low-level drivers, and consumers that
> don't care about the results of this test shouldn't be forced to pay
> for the test.
>
> Signed-off-by: Roland Dreier <[EMAIL PROTECTED]>
> ---
> drivers/infiniband/hw/amso1100/c2.h |2 +-
> drivers/infiniband/hw/amso1100/c2_cq.c | 16 ---
> drivers/infiniband/hw/cxgb3/cxio_hal.c |3 ++
> drivers/infiniband/hw/cxgb3/iwch_provider.c |8 +++--
> drivers/infiniband/hw/ehca/ehca_iverbs.h|2 +-
> drivers/infiniband/hw/ehca/ehca_reqs.c | 14 +++--
> drivers/infiniband/hw/ehca/ipz_pt_fn.h |8 +
> drivers/infiniband/hw/ipath/ipath_cq.c | 15 +++---
> drivers/infiniband/hw/ipath/ipath_verbs.h |2 +-
> drivers/infiniband/hw/mthca/mthca_cq.c | 12 +---
> drivers/infiniband/hw/mthca/mthca_dev.h |4 +-
> include/rdma/ib_verbs.h | 40
> +--
> 12 files changed, 93 insertions(+), 33 deletions(-)
>
> diff --git a/drivers/infiniband/hw/amso1100/c2.h
> b/drivers/infiniband/hw/amso1100/c2.h
> index 04a9db5..fa58200 100644
> --- a/drivers/infiniband/hw/amso1100/c2.h
> +++ b/drivers/infiniband/hw/amso1100/c2.h
> @@ -519,7 +519,7 @@ extern void c2_free_cq(struct c2_dev *c2dev,
> struct c2_cq *cq);
> extern void c2_cq_event(struct c2_dev *c2dev, u32 mq_index);
> extern void c2_cq_clean(struct c2_dev *c2dev, struct c2_qp *qp,
u32mq_index);
> extern int c2_poll_cq(struct ib_cq *ibcq, int num_entries, struct
> ib_wc *entry);
> -extern int c2_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify notify);
> +extern int c2_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags flags);
>
> /* CM */
> extern int c2_llp_connect(struct iw_cm_id *cm_id,
> diff --git a/drivers/infiniband/hw/amso1100/c2_cq.c
> b/drivers/infiniband/hw/amso1100/c2_cq.c
> index 5175c99..d2b3366 100644
> --- a/drivers/infiniband/hw/amso1100/c2_cq.c
> +++ b/drivers/infiniband/hw/amso1100/c2_cq.c
> @@ -217,17 +217,19 @@ int c2_poll_cq(struct ib_cq *ibcq, int
> num_entries, struct ib_wc *entry)
> return npolled;
> }
>
> -int c2_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify notify)
> +int c2_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags notify_flags)
> {
> struct c2_mq_shared __iomem *shared;
> struct c2_cq *cq;
> + unsigned long flags;
> + int ret = 0;
>
> cq = to_c2cq(ibcq);
>
Re: [ofa-general] [PATCH][RFC] IB: Return maybe missed event hint from ib_req_notify_cq()
Hi Roland!
As far as this concerns ehca this looks great.
Thanks
Nam
[EMAIL PROTECTED] wrote on 27.04.2007 00:43:19:
- IB: Return maybe missed event hint from ib_req_notify_cq()
This extends the API in a way that lets us implement NAPI, but may
be useful for other things too. It touches all the drivers, and I
still need to finish updating cxgb3 to work correctly. I haven't
heard anything negative about this, so I'll fix it up, post it one
more time for review, and plan on merging it.
As promised, here is that patch for review, with a cxgb3
implementation included.
---
The semantics defined by the InfiniBand specification say that
completion events are only generated when a completions is added to a
completion queue (CQ) after completion notification is requested. In
other words, this means that the following race is possible:
while (CQ is not empty)
ib_poll_cq(CQ);
// new completion is added after while loop is exited
ib_req_notify_cq(CQ);
// no event is generated for the existing completion
To close this race, the IB spec recommends doing another poll of the
CQ after requesting notification.
However, it is not always possible to arrange code this way (for
example, we have found that NAPI for IPoIB cannot poll after
requesting notification). Also, some hardware (eg Mellanox HCAs)
actually will generate an event for completions added before the call
to ib_req_notify_cq() -- which is allowed by the spec, since there's
no way for any upper-layer consumer to know exactly when a completion
was really added -- so the extra poll of the CQ is just a waste.
Motivated by this, we add a new flag IB_CQ_REPORT_MISSED_EVENTS for
ib_req_notify_cq() so that it can return a hint about whether the a
completion may have been added before the request for notification.
The return value of ib_req_notify_cq() is extended so:
0 means an error occurred while requesting notification
== 0 means notification was requested successfully, and if
IB_CQ_REPORT_MISSED_EVENTS was passed in, then no
events were missed and it is safe to wait for another
event.
0 is only returned if IB_CQ_REPORT_MISSED_EVENTS was
passed in. It means that the consumer must poll the
CQ again to make sure it is empty to avoid the race
described above.
We add a flag to enable this behavior rather than turning it on
unconditionally, because checking for missed events may incur
significant overhead for some low-level drivers, and consumers that
don't care about the results of this test shouldn't be forced to pay
for the test.
Signed-off-by: Roland Dreier [EMAIL PROTECTED]
---
drivers/infiniband/hw/amso1100/c2.h |2 +-
drivers/infiniband/hw/amso1100/c2_cq.c | 16 ---
drivers/infiniband/hw/cxgb3/cxio_hal.c |3 ++
drivers/infiniband/hw/cxgb3/iwch_provider.c |8 +++--
drivers/infiniband/hw/ehca/ehca_iverbs.h|2 +-
drivers/infiniband/hw/ehca/ehca_reqs.c | 14 +++--
drivers/infiniband/hw/ehca/ipz_pt_fn.h |8 +
drivers/infiniband/hw/ipath/ipath_cq.c | 15 +++---
drivers/infiniband/hw/ipath/ipath_verbs.h |2 +-
drivers/infiniband/hw/mthca/mthca_cq.c | 12 +---
drivers/infiniband/hw/mthca/mthca_dev.h |4 +-
include/rdma/ib_verbs.h | 40
+--
12 files changed, 93 insertions(+), 33 deletions(-)
diff --git a/drivers/infiniband/hw/amso1100/c2.h
b/drivers/infiniband/hw/amso1100/c2.h
index 04a9db5..fa58200 100644
--- a/drivers/infiniband/hw/amso1100/c2.h
+++ b/drivers/infiniband/hw/amso1100/c2.h
@@ -519,7 +519,7 @@ extern void c2_free_cq(struct c2_dev *c2dev,
struct c2_cq *cq);
extern void c2_cq_event(struct c2_dev *c2dev, u32 mq_index);
extern void c2_cq_clean(struct c2_dev *c2dev, struct c2_qp *qp,
u32mq_index);
extern int c2_poll_cq(struct ib_cq *ibcq, int num_entries, struct
ib_wc *entry);
-extern int c2_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify notify);
+extern int c2_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags flags);
/* CM */
extern int c2_llp_connect(struct iw_cm_id *cm_id,
diff --git a/drivers/infiniband/hw/amso1100/c2_cq.c
b/drivers/infiniband/hw/amso1100/c2_cq.c
index 5175c99..d2b3366 100644
--- a/drivers/infiniband/hw/amso1100/c2_cq.c
+++ b/drivers/infiniband/hw/amso1100/c2_cq.c
@@ -217,17 +217,19 @@ int c2_poll_cq(struct ib_cq *ibcq, int
num_entries, struct ib_wc *entry)
return npolled;
}
-int c2_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify notify)
+int c2_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags notify_flags)
{
struct c2_mq_shared __iomem *shared;
struct c2_cq *cq;
+ unsigned long flags;
+ int ret = 0;
cq = to_c2cq(ibcq);
shared = cq-mq.peer;
- if (notify == IB_CQ_NEXT_COMP)
+ if ((notify_flags IB_CQ_SOLICITED_MASK) == IB_CQ_NEXT_COMP)
Re: [PATCH 2.6.21-rc2] ehca: fix mismatched sync between completion handler and destroy cq
> > +#include > This can just be , because you're only using > wait_queue_head_t and not struct completion, right? > I fixed this up before merging. Yes, right. Thanks for your help! Regards Nam - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [PATCH 2.6.21-rc2] ehca: fix mismatched sync between completion handler and destroy cq
+#include linux/completion.h This can just be linux/wait.h, because you're only using wait_queue_head_t and not struct completion, right? I fixed this up before merging. Yes, right. Thanks for your help! Regards Nam - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
[PATCH 2.6.21-rc2] ehca: fix mismatched sync between completion handler and destroy cq
This patch fixes two issues reported by Roland and Christoph H.:
- Mismatched sync/locking between completion handler and destroy cq
We introduced a counter nr_events per cq to track number of irq
events seen. This counter is incremented when an event queue
entry is seen and decremented after completion handler has been
called regardless if scaling code is active or not. Note that
nr_callbacks tracks number of events assigned to a cpu and
both counters can potentially diverge.
The sync between running completion handler and destroy cq
is done by using the global spin lock ehca_cq_idr_lock.
- Replace yield by wait_event on the counter above to become zero
Signed-off-by: Hoang-Nam Nguyen <[EMAIL PROTECTED]>
---
ehca_classes.h |6 -
ehca_cq.c | 16 +--
ehca_irq.c | 59 +
ehca_main.c|4 +--
4 files changed, 60 insertions(+), 25 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_classes.h
b/drivers/infiniband/hw/ehca/ehca_classes.h
index 40404c9..85fe741 100644
--- a/drivers/infiniband/hw/ehca/ehca_classes.h
+++ b/drivers/infiniband/hw/ehca/ehca_classes.h
@@ -52,6 +52,8 @@ struct ehca_mw;
struct ehca_pd;
struct ehca_av;
+#include
+
#include
#include
@@ -153,7 +155,9 @@ struct ehca_cq {
spinlock_t cb_lock;
struct hlist_head qp_hashtab[QP_HASHTAB_LEN];
struct list_head entry;
- u32 nr_callbacks;
+ u32 nr_callbacks; /* #events assigned to cpu by scaling code */
+ u32 nr_events;/* #events seen */
+ wait_queue_head_t wait_completion;
spinlock_t task_lock;
u32 ownpid;
/* mmap counter for resources mapped into user space */
diff --git a/drivers/infiniband/hw/ehca/ehca_cq.c
b/drivers/infiniband/hw/ehca/ehca_cq.c
index 6ebfa27..e2cdc1a 100644
--- a/drivers/infiniband/hw/ehca/ehca_cq.c
+++ b/drivers/infiniband/hw/ehca/ehca_cq.c
@@ -146,6 +146,7 @@ struct ib_cq *ehca_create_cq(struct ib_d
spin_lock_init(_cq->spinlock);
spin_lock_init(_cq->cb_lock);
spin_lock_init(_cq->task_lock);
+ init_waitqueue_head(_cq->wait_completion);
my_cq->ownpid = current->tgid;
cq = _cq->ib_cq;
@@ -302,6 +303,16 @@ create_cq_exit1:
return cq;
}
+static int get_cq_nr_events(struct ehca_cq *my_cq)
+{
+ int ret;
+ unsigned long flags;
+ spin_lock_irqsave(_cq_idr_lock, flags);
+ ret = my_cq->nr_events;
+ spin_unlock_irqrestore(_cq_idr_lock, flags);
+ return ret;
+}
+
int ehca_destroy_cq(struct ib_cq *cq)
{
u64 h_ret;
@@ -329,10 +340,11 @@ int ehca_destroy_cq(struct ib_cq *cq)
}
spin_lock_irqsave(_cq_idr_lock, flags);
- while (my_cq->nr_callbacks) {
+ while (my_cq->nr_events) {
spin_unlock_irqrestore(_cq_idr_lock, flags);
- yield();
+ wait_event(my_cq->wait_completion, !get_cq_nr_events(my_cq));
spin_lock_irqsave(_cq_idr_lock, flags);
+ /* recheck nr_events to assure no cqe has just arrived */
}
idr_remove(_cq_idr, my_cq->token);
diff --git a/drivers/infiniband/hw/ehca/ehca_irq.c
b/drivers/infiniband/hw/ehca/ehca_irq.c
index 3ec53c6..7d8b795 100644
--- a/drivers/infiniband/hw/ehca/ehca_irq.c
+++ b/drivers/infiniband/hw/ehca/ehca_irq.c
@@ -404,10 +403,11 @@ static inline void process_eqe(struct eh
u32 token;
unsigned long flags;
struct ehca_cq *cq;
+
eqe_value = eqe->entry;
ehca_dbg(>ib_device, "eqe_value=%lx", eqe_value);
if (EHCA_BMASK_GET(EQE_COMPLETION_EVENT, eqe_value)) {
- ehca_dbg(>ib_device, "... completion event");
+ ehca_dbg(>ib_device, "Got completion event");
token = EHCA_BMASK_GET(EQE_CQ_TOKEN, eqe_value);
spin_lock_irqsave(_cq_idr_lock, flags);
cq = idr_find(_cq_idr, token);
@@ -419,16 +419,20 @@ static inline void process_eqe(struct eh
return;
}
reset_eq_pending(cq);
- if (ehca_scaling_code) {
+ cq->nr_events++;
+ spin_unlock_irqrestore(_cq_idr_lock, flags);
+ if (ehca_scaling_code)
queue_comp_task(cq);
- spin_unlock_irqrestore(_cq_idr_lock, flags);
- } else {
- spin_unlock_irqrestore(_cq_idr_lock, flags);
+ else {
comp_event_callback(cq);
+ spin_lock_irqsave(_cq_idr_lock, flags);
+ cq->nr_events--;
+ if (!cq->nr_events)
+ wake_up(>wait_completion);
+ spin_unlock_irqrestore(_cq_idr_lock, flags);
}
[PATCH 2.6.21-rc2] ehca: fix mismatched sync between completion handler and destroy cq
This patch fixes two issues reported by Roland and Christoph H.:
- Mismatched sync/locking between completion handler and destroy cq
We introduced a counter nr_events per cq to track number of irq
events seen. This counter is incremented when an event queue
entry is seen and decremented after completion handler has been
called regardless if scaling code is active or not. Note that
nr_callbacks tracks number of events assigned to a cpu and
both counters can potentially diverge.
The sync between running completion handler and destroy cq
is done by using the global spin lock ehca_cq_idr_lock.
- Replace yield by wait_event on the counter above to become zero
Signed-off-by: Hoang-Nam Nguyen [EMAIL PROTECTED]
---
ehca_classes.h |6 -
ehca_cq.c | 16 +--
ehca_irq.c | 59 +
ehca_main.c|4 +--
4 files changed, 60 insertions(+), 25 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_classes.h
b/drivers/infiniband/hw/ehca/ehca_classes.h
index 40404c9..85fe741 100644
--- a/drivers/infiniband/hw/ehca/ehca_classes.h
+++ b/drivers/infiniband/hw/ehca/ehca_classes.h
@@ -52,6 +52,8 @@ struct ehca_mw;
struct ehca_pd;
struct ehca_av;
+#include linux/completion.h
+
#include rdma/ib_verbs.h
#include rdma/ib_user_verbs.h
@@ -153,7 +155,9 @@ struct ehca_cq {
spinlock_t cb_lock;
struct hlist_head qp_hashtab[QP_HASHTAB_LEN];
struct list_head entry;
- u32 nr_callbacks;
+ u32 nr_callbacks; /* #events assigned to cpu by scaling code */
+ u32 nr_events;/* #events seen */
+ wait_queue_head_t wait_completion;
spinlock_t task_lock;
u32 ownpid;
/* mmap counter for resources mapped into user space */
diff --git a/drivers/infiniband/hw/ehca/ehca_cq.c
b/drivers/infiniband/hw/ehca/ehca_cq.c
index 6ebfa27..e2cdc1a 100644
--- a/drivers/infiniband/hw/ehca/ehca_cq.c
+++ b/drivers/infiniband/hw/ehca/ehca_cq.c
@@ -146,6 +146,7 @@ struct ib_cq *ehca_create_cq(struct ib_d
spin_lock_init(my_cq-spinlock);
spin_lock_init(my_cq-cb_lock);
spin_lock_init(my_cq-task_lock);
+ init_waitqueue_head(my_cq-wait_completion);
my_cq-ownpid = current-tgid;
cq = my_cq-ib_cq;
@@ -302,6 +303,16 @@ create_cq_exit1:
return cq;
}
+static int get_cq_nr_events(struct ehca_cq *my_cq)
+{
+ int ret;
+ unsigned long flags;
+ spin_lock_irqsave(ehca_cq_idr_lock, flags);
+ ret = my_cq-nr_events;
+ spin_unlock_irqrestore(ehca_cq_idr_lock, flags);
+ return ret;
+}
+
int ehca_destroy_cq(struct ib_cq *cq)
{
u64 h_ret;
@@ -329,10 +340,11 @@ int ehca_destroy_cq(struct ib_cq *cq)
}
spin_lock_irqsave(ehca_cq_idr_lock, flags);
- while (my_cq-nr_callbacks) {
+ while (my_cq-nr_events) {
spin_unlock_irqrestore(ehca_cq_idr_lock, flags);
- yield();
+ wait_event(my_cq-wait_completion, !get_cq_nr_events(my_cq));
spin_lock_irqsave(ehca_cq_idr_lock, flags);
+ /* recheck nr_events to assure no cqe has just arrived */
}
idr_remove(ehca_cq_idr, my_cq-token);
diff --git a/drivers/infiniband/hw/ehca/ehca_irq.c
b/drivers/infiniband/hw/ehca/ehca_irq.c
index 3ec53c6..7d8b795 100644
--- a/drivers/infiniband/hw/ehca/ehca_irq.c
+++ b/drivers/infiniband/hw/ehca/ehca_irq.c
@@ -404,10 +403,11 @@ static inline void process_eqe(struct eh
u32 token;
unsigned long flags;
struct ehca_cq *cq;
+
eqe_value = eqe-entry;
ehca_dbg(shca-ib_device, eqe_value=%lx, eqe_value);
if (EHCA_BMASK_GET(EQE_COMPLETION_EVENT, eqe_value)) {
- ehca_dbg(shca-ib_device, ... completion event);
+ ehca_dbg(shca-ib_device, Got completion event);
token = EHCA_BMASK_GET(EQE_CQ_TOKEN, eqe_value);
spin_lock_irqsave(ehca_cq_idr_lock, flags);
cq = idr_find(ehca_cq_idr, token);
@@ -419,16 +419,20 @@ static inline void process_eqe(struct eh
return;
}
reset_eq_pending(cq);
- if (ehca_scaling_code) {
+ cq-nr_events++;
+ spin_unlock_irqrestore(ehca_cq_idr_lock, flags);
+ if (ehca_scaling_code)
queue_comp_task(cq);
- spin_unlock_irqrestore(ehca_cq_idr_lock, flags);
- } else {
- spin_unlock_irqrestore(ehca_cq_idr_lock, flags);
+ else {
comp_event_callback(cq);
+ spin_lock_irqsave(ehca_cq_idr_lock, flags);
+ cq-nr_events--;
+ if (!cq-nr_events)
+ wake_up(cq-wait_completion);
+ spin_unlock_irqrestore(ehca_cq_idr_lock, flags);
}
} else
[PATCH 1/2] ibmebus: dynamic addiiton/removal of adapters, uevent, root device based on struct device
From: Joachim Fenkes <[EMAIL PROTECTED]>
This adds two sysfs attributes to /sys/bus/ibmebus which can
be used to notify the ebus driver of added / removed ebus
devices in the OF device tree.
Echoing the device's location code (as found in the OFDT
"ibm,loc-code" property) into the "probe" attribute will
notify ebus of addition of the device and cause the appropriate
device driver's probe function to be called on the device.
Likewise, echoing the location code into the "remove" attribute
will cause the device to be removed from the system.
The writes will block until the respective operation has
finished and return an error code if the operation failed.
Additionally, uevent is now supported by ibmebus through the
generic of_device_uevent function.
The fake root device used to provide a common parent for all
ebus devices is now based on device instead of of_device - it
had no associated devtree node. This saves several checks
throughout the ebus driver.
Signed-off-by: Joachim Fenkes <[EMAIL PROTECTED]>
---
I repost this patch because a kind fellow (wave to Paul =)
reminded me of the finer points of patch descriptions - Thanks!
As said before, I deem this ready for inclusion.
arch/powerpc/kernel/ibmebus.c | 129 ++
include/asm-powerpc/ibmebus.h |2
diff -wurp a/arch/powerpc/kernel/ibmebus.c b/arch/powerpc/kernel/ibmebus.c
--- a/arch/powerpc/kernel/ibmebus.c 2007-02-22 05:26:24.971939672 +0100
+++ b/arch/powerpc/kernel/ibmebus.c 2007-02-20 23:31:39.0 +0100
@@ -2,6 +2,7 @@
* IBM PowerPC IBM eBus Infrastructure Support.
*
* Copyright (c) 2005 IBM Corporation
+ * Joachim Fenkes <[EMAIL PROTECTED]>
* Heiko J Schick <[EMAIL PROTECTED]>
*
* All rights reserved.
@@ -43,10 +44,8 @@
#include
#include
-static struct ibmebus_dev ibmebus_bus_device = { /* fake "parent" device */
- .name = ibmebus_bus_device.ofdev.dev.bus_id,
- .ofdev.dev.bus_id = "ibmebus",
- .ofdev.dev.bus= _bus_type,
+static struct device ibmebus_bus_device = { /* fake "parent" device */
+ .bus_id = "ibmebus",
};
static void *ibmebus_alloc_coherent(struct device *dev,
@@ -161,18 +160,19 @@ static void __devinit ibmebus_dev_releas
static ssize_t ibmebusdev_show_name(struct device *dev,
struct device_attribute *attr, char *buf)
{
- return sprintf(buf, "%s\n", to_ibmebus_dev(dev)->name);
+ struct ibmebus_dev *ebus_dev = to_ibmebus_dev(dev);
+ char *name = (char*)get_property(ebus_dev->ofdev.node, "name", NULL);
+ return sprintf(buf, "%s\n", name);
}
static DEVICE_ATTR(name, S_IRUSR | S_IRGRP | S_IROTH, ibmebusdev_show_name,
NULL);
-static struct ibmebus_dev* __devinit ibmebus_register_device_common(
+static int __devinit ibmebus_register_device_common(
struct ibmebus_dev *dev, const char *name)
{
int err = 0;
- dev->name = name;
- dev->ofdev.dev.parent = _bus_device.ofdev.dev;
+ dev->ofdev.dev.parent = _bus_device;
dev->ofdev.dev.bus = _bus_type;
dev->ofdev.dev.release = ibmebus_dev_release;
@@ -186,12 +186,12 @@ static struct ibmebus_dev* __devinit ibm
if ((err = of_device_register(>ofdev)) != 0) {
printk(KERN_ERR "%s: failed to register device (%d).\n",
__FUNCTION__, err);
- return NULL;
+ return -ENODEV;
}
device_create_file(>ofdev.dev, _attr_name);
- return dev;
+ return 0;
}
static struct ibmebus_dev* __devinit ibmebus_register_device_node(
@@ -205,18 +205,18 @@ static struct ibmebus_dev* __devinit ibm
if (!loc_code) {
printk(KERN_WARNING "%s: node %s missing 'ibm,loc-code'\n",
__FUNCTION__, dn->name ? dn->name : "");
- return NULL;
+ return ERR_PTR(-EINVAL);
}
if (strlen(loc_code) == 0) {
printk(KERN_WARNING "%s: 'ibm,loc-code' is invalid\n",
__FUNCTION__);
- return NULL;
+ return ERR_PTR(-EINVAL);
}
dev = kzalloc(sizeof(struct ibmebus_dev), GFP_KERNEL);
if (!dev) {
- return NULL;
+ return ERR_PTR(-ENOMEM);
}
dev->ofdev.node = of_node_get(dn);
@@ -227,9 +227,9 @@ static struct ibmebus_dev* __devinit ibm
min(length, BUS_ID_SIZE - 1));
/* Register with generic device framework. */
- if (ibmebus_register_device_common(dev, dn->name) == NULL) {
+ if (ibmebus_register_device_common(dev, dn->name) != 0) {
kfree(dev);
- return NULL;
+ return ERR_PTR(-ENODEV);
}
return dev;
@@ -240,9 +240,8 @@ static void ibmebus_probe_of_nodes(char*
struct device_node *dn = NULL;
while ((dn = of_find_node_by_name(dn,
[PATCH 2/2] ibmebus: whitespace fixes
From: Joachim Fenkes <[EMAIL PROTECTED]>
This fixes whitespacing in ibmebus.[ch] - no functionality changed.
Signed-off-by: Joachim Fenkes <[EMAIL PROTECTED]>
---
arch/powerpc/kernel/ibmebus.c | 126 +-
include/asm-powerpc/ibmebus.h | 42 +++---
diff -urp b/arch/powerpc/kernel/ibmebus.c c/arch/powerpc/kernel/ibmebus.c
--- b/arch/powerpc/kernel/ibmebus.c 2007-02-22 05:43:32.133934656 +0100
+++ c/arch/powerpc/kernel/ibmebus.c 2007-02-20 23:31:39.0 +0100
@@ -4,35 +4,35 @@
* Copyright (c) 2005 IBM Corporation
* Joachim Fenkes <[EMAIL PROTECTED]>
* Heiko J Schick <[EMAIL PROTECTED]>
- *
+ *
* All rights reserved.
*
- * This source code is distributed under a dual license of GPL v2.0 and OpenIB
- * BSD.
+ * This source code is distributed under a dual license of GPL v2.0 and OpenIB
+ * BSD.
*
* OpenIB BSD License
*
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions are met:
+ * 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 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
+ * 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.
+ * provided with the distribution.
*
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
- * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
- * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
- * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
- * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * 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
+ * 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.
*/
@@ -54,7 +54,7 @@ static void *ibmebus_alloc_coherent(stru
gfp_t flag)
{
void *mem;
-
+
mem = kmalloc(size, flag);
*dma_handle = (dma_addr_t)mem;
@@ -62,7 +62,7 @@ static void *ibmebus_alloc_coherent(stru
}
static void ibmebus_free_coherent(struct device *dev,
- size_t size, void *vaddr,
+ size_t size, void *vaddr,
dma_addr_t dma_handle)
{
kfree(vaddr);
@@ -78,7 +78,7 @@ static dma_addr_t ibmebus_map_single(str
static void ibmebus_unmap_single(struct device *dev,
dma_addr_t dma_addr,
-size_t size,
+size_t size,
enum dma_data_direction direction)
{
return;
@@ -89,13 +89,13 @@ static int ibmebus_map_sg(struct device
int nents, enum dma_data_direction direction)
{
int i;
-
+
for (i = 0; i < nents; i++) {
- sg[i].dma_address = (dma_addr_t)page_address(sg[i].page)
+ sg[i].dma_address = (dma_addr_t)page_address(sg[i].page)
+ sg[i].offset;
sg[i].dma_length = sg[i].length;
}
-
+
return nents;
}
@@ -127,15 +127,15 @@ static int ibmebus_bus_probe(struct devi
struct ibmebus_driver *ibmebusdrv = to_ibmebus_driver(dev->driver);
const struct of_device_id *id;
int error = -ENODEV;
-
+
if (!ibmebusdrv->probe)
return
[PATCH 1/2] ibmebus: dynamic addiiton/removal of adapters, uevent, root device based on struct device
From: Joachim Fenkes [EMAIL PROTECTED]
This adds two sysfs attributes to /sys/bus/ibmebus which can
be used to notify the ebus driver of added / removed ebus
devices in the OF device tree.
Echoing the device's location code (as found in the OFDT
ibm,loc-code property) into the probe attribute will
notify ebus of addition of the device and cause the appropriate
device driver's probe function to be called on the device.
Likewise, echoing the location code into the remove attribute
will cause the device to be removed from the system.
The writes will block until the respective operation has
finished and return an error code if the operation failed.
Additionally, uevent is now supported by ibmebus through the
generic of_device_uevent function.
The fake root device used to provide a common parent for all
ebus devices is now based on device instead of of_device - it
had no associated devtree node. This saves several checks
throughout the ebus driver.
Signed-off-by: Joachim Fenkes [EMAIL PROTECTED]
---
I repost this patch because a kind fellow (wave to Paul =)
reminded me of the finer points of patch descriptions - Thanks!
As said before, I deem this ready for inclusion.
arch/powerpc/kernel/ibmebus.c | 129 ++
include/asm-powerpc/ibmebus.h |2
diff -wurp a/arch/powerpc/kernel/ibmebus.c b/arch/powerpc/kernel/ibmebus.c
--- a/arch/powerpc/kernel/ibmebus.c 2007-02-22 05:26:24.971939672 +0100
+++ b/arch/powerpc/kernel/ibmebus.c 2007-02-20 23:31:39.0 +0100
@@ -2,6 +2,7 @@
* IBM PowerPC IBM eBus Infrastructure Support.
*
* Copyright (c) 2005 IBM Corporation
+ * Joachim Fenkes [EMAIL PROTECTED]
* Heiko J Schick [EMAIL PROTECTED]
*
* All rights reserved.
@@ -43,10 +44,8 @@
#include asm/ibmebus.h
#include asm/abs_addr.h
-static struct ibmebus_dev ibmebus_bus_device = { /* fake parent device */
- .name = ibmebus_bus_device.ofdev.dev.bus_id,
- .ofdev.dev.bus_id = ibmebus,
- .ofdev.dev.bus= ibmebus_bus_type,
+static struct device ibmebus_bus_device = { /* fake parent device */
+ .bus_id = ibmebus,
};
static void *ibmebus_alloc_coherent(struct device *dev,
@@ -161,18 +160,19 @@ static void __devinit ibmebus_dev_releas
static ssize_t ibmebusdev_show_name(struct device *dev,
struct device_attribute *attr, char *buf)
{
- return sprintf(buf, %s\n, to_ibmebus_dev(dev)-name);
+ struct ibmebus_dev *ebus_dev = to_ibmebus_dev(dev);
+ char *name = (char*)get_property(ebus_dev-ofdev.node, name, NULL);
+ return sprintf(buf, %s\n, name);
}
static DEVICE_ATTR(name, S_IRUSR | S_IRGRP | S_IROTH, ibmebusdev_show_name,
NULL);
-static struct ibmebus_dev* __devinit ibmebus_register_device_common(
+static int __devinit ibmebus_register_device_common(
struct ibmebus_dev *dev, const char *name)
{
int err = 0;
- dev-name = name;
- dev-ofdev.dev.parent = ibmebus_bus_device.ofdev.dev;
+ dev-ofdev.dev.parent = ibmebus_bus_device;
dev-ofdev.dev.bus = ibmebus_bus_type;
dev-ofdev.dev.release = ibmebus_dev_release;
@@ -186,12 +186,12 @@ static struct ibmebus_dev* __devinit ibm
if ((err = of_device_register(dev-ofdev)) != 0) {
printk(KERN_ERR %s: failed to register device (%d).\n,
__FUNCTION__, err);
- return NULL;
+ return -ENODEV;
}
device_create_file(dev-ofdev.dev, dev_attr_name);
- return dev;
+ return 0;
}
static struct ibmebus_dev* __devinit ibmebus_register_device_node(
@@ -205,18 +205,18 @@ static struct ibmebus_dev* __devinit ibm
if (!loc_code) {
printk(KERN_WARNING %s: node %s missing 'ibm,loc-code'\n,
__FUNCTION__, dn-name ? dn-name : unknown);
- return NULL;
+ return ERR_PTR(-EINVAL);
}
if (strlen(loc_code) == 0) {
printk(KERN_WARNING %s: 'ibm,loc-code' is invalid\n,
__FUNCTION__);
- return NULL;
+ return ERR_PTR(-EINVAL);
}
dev = kzalloc(sizeof(struct ibmebus_dev), GFP_KERNEL);
if (!dev) {
- return NULL;
+ return ERR_PTR(-ENOMEM);
}
dev-ofdev.node = of_node_get(dn);
@@ -227,9 +227,9 @@ static struct ibmebus_dev* __devinit ibm
min(length, BUS_ID_SIZE - 1));
/* Register with generic device framework. */
- if (ibmebus_register_device_common(dev, dn-name) == NULL) {
+ if (ibmebus_register_device_common(dev, dn-name) != 0) {
kfree(dev);
- return NULL;
+ return ERR_PTR(-ENODEV);
}
return dev;
@@ -240,9 +240,8 @@ static void ibmebus_probe_of_nodes(char*
struct device_node *dn = NULL;
while ((dn =
[PATCH 2/2] ibmebus: dynamic add/remove, uevent, root device and whitespace
This is the aforementioned whitespace fix which applies on top of
part 1/2.
Signed-off-by: Joachim Fenkes <[EMAIL PROTECTED]>
---
arch/powerpc/kernel/ibmebus.c | 126 +-
include/asm-powerpc/ibmebus.h | 42 +++---
diff -urp b/arch/powerpc/kernel/ibmebus.c c/arch/powerpc/kernel/ibmebus.c
--- b/arch/powerpc/kernel/ibmebus.c 2007-02-22 05:43:32.133934656 +0100
+++ c/arch/powerpc/kernel/ibmebus.c 2007-02-20 23:31:39.0 +0100
@@ -4,35 +4,35 @@
* Copyright (c) 2005 IBM Corporation
* Joachim Fenkes <[EMAIL PROTECTED]>
* Heiko J Schick <[EMAIL PROTECTED]>
- *
+ *
* All rights reserved.
*
- * This source code is distributed under a dual license of GPL v2.0 and OpenIB
- * BSD.
+ * This source code is distributed under a dual license of GPL v2.0 and OpenIB
+ * BSD.
*
* OpenIB BSD License
*
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions are met:
+ * 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 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
+ * 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.
+ * provided with the distribution.
*
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
- * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
- * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
- * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
- * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * 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
+ * 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.
*/
@@ -54,7 +54,7 @@ static void *ibmebus_alloc_coherent(stru
gfp_t flag)
{
void *mem;
-
+
mem = kmalloc(size, flag);
*dma_handle = (dma_addr_t)mem;
@@ -62,7 +62,7 @@ static void *ibmebus_alloc_coherent(stru
}
static void ibmebus_free_coherent(struct device *dev,
- size_t size, void *vaddr,
+ size_t size, void *vaddr,
dma_addr_t dma_handle)
{
kfree(vaddr);
@@ -78,7 +78,7 @@ static dma_addr_t ibmebus_map_single(str
static void ibmebus_unmap_single(struct device *dev,
dma_addr_t dma_addr,
-size_t size,
+size_t size,
enum dma_data_direction direction)
{
return;
@@ -89,13 +89,13 @@ static int ibmebus_map_sg(struct device
int nents, enum dma_data_direction direction)
{
int i;
-
+
for (i = 0; i < nents; i++) {
- sg[i].dma_address = (dma_addr_t)page_address(sg[i].page)
+ sg[i].dma_address = (dma_addr_t)page_address(sg[i].page)
+ sg[i].offset;
sg[i].dma_length = sg[i].length;
}
-
+
return nents;
}
@@ -127,15 +127,15 @@ static int ibmebus_bus_probe(struct devi
struct ibmebus_driver *ibmebusdrv = to_ibmebus_driver(dev->driver);
const struct of_device_id *id;
int error = -ENODEV;
-
+
if (!ibmebusdrv->probe)
return error;
-
+
id =
[PATCH 1/2] ibmebus: dynamic add/remove, uevent, root device and whitespace
The first part of this patch summarizes the patches of the
previous days, namely:
- Add dynamic addition/removal of adapters
(with spiffy error reporting)
- Implement the uevent interface using Sylvain's generic function
- Base fake root device on device instead of of_device
The first part will apply against the vanilla 2.6.20 source.
The second part is just a whitespace fix and applies on top of the first.
If nobody objects, I deem these patches ready for inclusion.
Signed-off-by: Joachim Fenkes <[EMAIL PROTECTED]>
---
arch/powerpc/kernel/ibmebus.c | 129 ++
include/asm-powerpc/ibmebus.h |2
diff -wurp a/arch/powerpc/kernel/ibmebus.c b/arch/powerpc/kernel/ibmebus.c
--- a/arch/powerpc/kernel/ibmebus.c 2007-02-22 05:26:24.971939672 +0100
+++ b/arch/powerpc/kernel/ibmebus.c 2007-02-20 23:31:39.0 +0100
@@ -2,6 +2,7 @@
* IBM PowerPC IBM eBus Infrastructure Support.
*
* Copyright (c) 2005 IBM Corporation
+ * Joachim Fenkes <[EMAIL PROTECTED]>
* Heiko J Schick <[EMAIL PROTECTED]>
*
* All rights reserved.
@@ -43,10 +44,8 @@
#include
#include
-static struct ibmebus_dev ibmebus_bus_device = { /* fake "parent" device */
- .name = ibmebus_bus_device.ofdev.dev.bus_id,
- .ofdev.dev.bus_id = "ibmebus",
- .ofdev.dev.bus= _bus_type,
+static struct device ibmebus_bus_device = { /* fake "parent" device */
+ .bus_id = "ibmebus",
};
static void *ibmebus_alloc_coherent(struct device *dev,
@@ -161,18 +160,19 @@ static void __devinit ibmebus_dev_releas
static ssize_t ibmebusdev_show_name(struct device *dev,
struct device_attribute *attr, char *buf)
{
- return sprintf(buf, "%s\n", to_ibmebus_dev(dev)->name);
+ struct ibmebus_dev *ebus_dev = to_ibmebus_dev(dev);
+ char *name = (char*)get_property(ebus_dev->ofdev.node, "name", NULL);
+ return sprintf(buf, "%s\n", name);
}
static DEVICE_ATTR(name, S_IRUSR | S_IRGRP | S_IROTH, ibmebusdev_show_name,
NULL);
-static struct ibmebus_dev* __devinit ibmebus_register_device_common(
+static int __devinit ibmebus_register_device_common(
struct ibmebus_dev *dev, const char *name)
{
int err = 0;
- dev->name = name;
- dev->ofdev.dev.parent = _bus_device.ofdev.dev;
+ dev->ofdev.dev.parent = _bus_device;
dev->ofdev.dev.bus = _bus_type;
dev->ofdev.dev.release = ibmebus_dev_release;
@@ -186,12 +186,12 @@ static struct ibmebus_dev* __devinit ibm
if ((err = of_device_register(>ofdev)) != 0) {
printk(KERN_ERR "%s: failed to register device (%d).\n",
__FUNCTION__, err);
- return NULL;
+ return -ENODEV;
}
device_create_file(>ofdev.dev, _attr_name);
- return dev;
+ return 0;
}
static struct ibmebus_dev* __devinit ibmebus_register_device_node(
@@ -205,18 +205,18 @@ static struct ibmebus_dev* __devinit ibm
if (!loc_code) {
printk(KERN_WARNING "%s: node %s missing 'ibm,loc-code'\n",
__FUNCTION__, dn->name ? dn->name : "");
- return NULL;
+ return ERR_PTR(-EINVAL);
}
if (strlen(loc_code) == 0) {
printk(KERN_WARNING "%s: 'ibm,loc-code' is invalid\n",
__FUNCTION__);
- return NULL;
+ return ERR_PTR(-EINVAL);
}
dev = kzalloc(sizeof(struct ibmebus_dev), GFP_KERNEL);
if (!dev) {
- return NULL;
+ return ERR_PTR(-ENOMEM);
}
dev->ofdev.node = of_node_get(dn);
@@ -227,9 +227,9 @@ static struct ibmebus_dev* __devinit ibm
min(length, BUS_ID_SIZE - 1));
/* Register with generic device framework. */
- if (ibmebus_register_device_common(dev, dn->name) == NULL) {
+ if (ibmebus_register_device_common(dev, dn->name) != 0) {
kfree(dev);
- return NULL;
+ return ERR_PTR(-ENODEV);
}
return dev;
@@ -240,9 +240,8 @@ static void ibmebus_probe_of_nodes(char*
struct device_node *dn = NULL;
while ((dn = of_find_node_by_name(dn, name))) {
- if (ibmebus_register_device_node(dn) == NULL) {
+ if (IS_ERR(ibmebus_register_device_node(dn))) {
of_node_put(dn);
-
return;
}
}
@@ -262,9 +261,15 @@ static void ibmebus_add_devices_by_id(st
return;
}
-static int ibmebus_match_helper(struct device *dev, void *data)
+static int ibmebus_match_helper_name(struct device *dev, void *data)
{
- if (strcmp((char*)data, to_ibmebus_dev(dev)->name) == 0)
+ const struct ibmebus_dev *ebus_dev = to_ibmebus_dev(dev);
+ char *name;
+
+ name =
[PATCH 1/2] ibmebus: dynamic add/remove, uevent, root device and whitespace
The first part of this patch summarizes the patches of the
previous days, namely:
- Add dynamic addition/removal of adapters
(with spiffy error reporting)
- Implement the uevent interface using Sylvain's generic function
- Base fake root device on device instead of of_device
The first part will apply against the vanilla 2.6.20 source.
The second part is just a whitespace fix and applies on top of the first.
If nobody objects, I deem these patches ready for inclusion.
Signed-off-by: Joachim Fenkes [EMAIL PROTECTED]
---
arch/powerpc/kernel/ibmebus.c | 129 ++
include/asm-powerpc/ibmebus.h |2
diff -wurp a/arch/powerpc/kernel/ibmebus.c b/arch/powerpc/kernel/ibmebus.c
--- a/arch/powerpc/kernel/ibmebus.c 2007-02-22 05:26:24.971939672 +0100
+++ b/arch/powerpc/kernel/ibmebus.c 2007-02-20 23:31:39.0 +0100
@@ -2,6 +2,7 @@
* IBM PowerPC IBM eBus Infrastructure Support.
*
* Copyright (c) 2005 IBM Corporation
+ * Joachim Fenkes [EMAIL PROTECTED]
* Heiko J Schick [EMAIL PROTECTED]
*
* All rights reserved.
@@ -43,10 +44,8 @@
#include asm/ibmebus.h
#include asm/abs_addr.h
-static struct ibmebus_dev ibmebus_bus_device = { /* fake parent device */
- .name = ibmebus_bus_device.ofdev.dev.bus_id,
- .ofdev.dev.bus_id = ibmebus,
- .ofdev.dev.bus= ibmebus_bus_type,
+static struct device ibmebus_bus_device = { /* fake parent device */
+ .bus_id = ibmebus,
};
static void *ibmebus_alloc_coherent(struct device *dev,
@@ -161,18 +160,19 @@ static void __devinit ibmebus_dev_releas
static ssize_t ibmebusdev_show_name(struct device *dev,
struct device_attribute *attr, char *buf)
{
- return sprintf(buf, %s\n, to_ibmebus_dev(dev)-name);
+ struct ibmebus_dev *ebus_dev = to_ibmebus_dev(dev);
+ char *name = (char*)get_property(ebus_dev-ofdev.node, name, NULL);
+ return sprintf(buf, %s\n, name);
}
static DEVICE_ATTR(name, S_IRUSR | S_IRGRP | S_IROTH, ibmebusdev_show_name,
NULL);
-static struct ibmebus_dev* __devinit ibmebus_register_device_common(
+static int __devinit ibmebus_register_device_common(
struct ibmebus_dev *dev, const char *name)
{
int err = 0;
- dev-name = name;
- dev-ofdev.dev.parent = ibmebus_bus_device.ofdev.dev;
+ dev-ofdev.dev.parent = ibmebus_bus_device;
dev-ofdev.dev.bus = ibmebus_bus_type;
dev-ofdev.dev.release = ibmebus_dev_release;
@@ -186,12 +186,12 @@ static struct ibmebus_dev* __devinit ibm
if ((err = of_device_register(dev-ofdev)) != 0) {
printk(KERN_ERR %s: failed to register device (%d).\n,
__FUNCTION__, err);
- return NULL;
+ return -ENODEV;
}
device_create_file(dev-ofdev.dev, dev_attr_name);
- return dev;
+ return 0;
}
static struct ibmebus_dev* __devinit ibmebus_register_device_node(
@@ -205,18 +205,18 @@ static struct ibmebus_dev* __devinit ibm
if (!loc_code) {
printk(KERN_WARNING %s: node %s missing 'ibm,loc-code'\n,
__FUNCTION__, dn-name ? dn-name : unknown);
- return NULL;
+ return ERR_PTR(-EINVAL);
}
if (strlen(loc_code) == 0) {
printk(KERN_WARNING %s: 'ibm,loc-code' is invalid\n,
__FUNCTION__);
- return NULL;
+ return ERR_PTR(-EINVAL);
}
dev = kzalloc(sizeof(struct ibmebus_dev), GFP_KERNEL);
if (!dev) {
- return NULL;
+ return ERR_PTR(-ENOMEM);
}
dev-ofdev.node = of_node_get(dn);
@@ -227,9 +227,9 @@ static struct ibmebus_dev* __devinit ibm
min(length, BUS_ID_SIZE - 1));
/* Register with generic device framework. */
- if (ibmebus_register_device_common(dev, dn-name) == NULL) {
+ if (ibmebus_register_device_common(dev, dn-name) != 0) {
kfree(dev);
- return NULL;
+ return ERR_PTR(-ENODEV);
}
return dev;
@@ -240,9 +240,8 @@ static void ibmebus_probe_of_nodes(char*
struct device_node *dn = NULL;
while ((dn = of_find_node_by_name(dn, name))) {
- if (ibmebus_register_device_node(dn) == NULL) {
+ if (IS_ERR(ibmebus_register_device_node(dn))) {
of_node_put(dn);
-
return;
}
}
@@ -262,9 +261,15 @@ static void ibmebus_add_devices_by_id(st
return;
}
-static int ibmebus_match_helper(struct device *dev, void *data)
+static int ibmebus_match_helper_name(struct device *dev, void *data)
{
- if (strcmp((char*)data, to_ibmebus_dev(dev)-name) == 0)
+ const struct ibmebus_dev *ebus_dev = to_ibmebus_dev(dev);
+ char *name;
+
[PATCH 2/2] ibmebus: dynamic add/remove, uevent, root device and whitespace
This is the aforementioned whitespace fix which applies on top of
part 1/2.
Signed-off-by: Joachim Fenkes [EMAIL PROTECTED]
---
arch/powerpc/kernel/ibmebus.c | 126 +-
include/asm-powerpc/ibmebus.h | 42 +++---
diff -urp b/arch/powerpc/kernel/ibmebus.c c/arch/powerpc/kernel/ibmebus.c
--- b/arch/powerpc/kernel/ibmebus.c 2007-02-22 05:43:32.133934656 +0100
+++ c/arch/powerpc/kernel/ibmebus.c 2007-02-20 23:31:39.0 +0100
@@ -4,35 +4,35 @@
* Copyright (c) 2005 IBM Corporation
* Joachim Fenkes [EMAIL PROTECTED]
* Heiko J Schick [EMAIL PROTECTED]
- *
+ *
* All rights reserved.
*
- * This source code is distributed under a dual license of GPL v2.0 and OpenIB
- * BSD.
+ * This source code is distributed under a dual license of GPL v2.0 and OpenIB
+ * BSD.
*
* OpenIB BSD License
*
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions are met:
+ * 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 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
+ * 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.
+ * provided with the distribution.
*
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS AS IS
- * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
- * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
- * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
- * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * 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
+ * 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.
*/
@@ -54,7 +54,7 @@ static void *ibmebus_alloc_coherent(stru
gfp_t flag)
{
void *mem;
-
+
mem = kmalloc(size, flag);
*dma_handle = (dma_addr_t)mem;
@@ -62,7 +62,7 @@ static void *ibmebus_alloc_coherent(stru
}
static void ibmebus_free_coherent(struct device *dev,
- size_t size, void *vaddr,
+ size_t size, void *vaddr,
dma_addr_t dma_handle)
{
kfree(vaddr);
@@ -78,7 +78,7 @@ static dma_addr_t ibmebus_map_single(str
static void ibmebus_unmap_single(struct device *dev,
dma_addr_t dma_addr,
-size_t size,
+size_t size,
enum dma_data_direction direction)
{
return;
@@ -89,13 +89,13 @@ static int ibmebus_map_sg(struct device
int nents, enum dma_data_direction direction)
{
int i;
-
+
for (i = 0; i nents; i++) {
- sg[i].dma_address = (dma_addr_t)page_address(sg[i].page)
+ sg[i].dma_address = (dma_addr_t)page_address(sg[i].page)
+ sg[i].offset;
sg[i].dma_length = sg[i].length;
}
-
+
return nents;
}
@@ -127,15 +127,15 @@ static int ibmebus_bus_probe(struct devi
struct ibmebus_driver *ibmebusdrv = to_ibmebus_driver(dev-driver);
const struct of_device_id *id;
int error = -ENODEV;
-
+
if (!ibmebusdrv-probe)
return error;
-
+
id =
[PATCH 2.6.21-rc1] ibmebus: more error reporting in dynamic add/remove code
Writing the ibmebus probe and remove attributes now throws an appropriate
error if something goes wrong. This way, userspace tools can check for
success or failure of an addition or removal. The write will block until
the probe/remove operation completes, so, when the write operation returns
without an error, you can be sure the probe was successful and the device
is present in the system.
As an added bonus, an eventual trailing newline is now removed from the
written string.
Signed-off-by: Joachim Fenkes <[EMAIL PROTECTED]>
---
ibmebus.c | 54 --
1 files changed, 36 insertions(+), 18 deletions(-)
diff -urp a/arch/powerpc/kernel/ibmebus.c b/arch/powerpc/kernel/ibmebus.c
--- a/arch/powerpc/kernel/ibmebus.c 2007-02-20 21:54:22.366443424 +0100
+++ b/arch/powerpc/kernel/ibmebus.c 2007-02-20 23:31:39.950486616 +0100
@@ -167,7 +167,7 @@ static ssize_t ibmebusdev_show_name(stru
static DEVICE_ATTR(name, S_IRUSR | S_IRGRP | S_IROTH, ibmebusdev_show_name,
NULL);
-static struct ibmebus_dev* __devinit ibmebus_register_device_common(
+static int __devinit ibmebus_register_device_common(
struct ibmebus_dev *dev, const char *name)
{
int err = 0;
@@ -186,12 +186,12 @@ static struct ibmebus_dev* __devinit ibm
if ((err = of_device_register(>ofdev)) != 0) {
printk(KERN_ERR "%s: failed to register device (%d).\n",
__FUNCTION__, err);
- return NULL;
+ return -ENODEV;
}
device_create_file(>ofdev.dev, _attr_name);
- return dev;
+ return 0;
}
static struct ibmebus_dev* __devinit ibmebus_register_device_node(
@@ -205,18 +205,18 @@ static struct ibmebus_dev* __devinit ibm
if (!loc_code) {
printk(KERN_WARNING "%s: node %s missing 'ibm,loc-code'\n",
__FUNCTION__, dn->name ? dn->name : "");
- return NULL;
+ return ERR_PTR(-EINVAL);
}
if (strlen(loc_code) == 0) {
printk(KERN_WARNING "%s: 'ibm,loc-code' is invalid\n",
__FUNCTION__);
- return NULL;
+ return ERR_PTR(-EINVAL);
}
dev = kzalloc(sizeof(struct ibmebus_dev), GFP_KERNEL);
if (!dev) {
- return NULL;
+ return ERR_PTR(-ENOMEM);
}
dev->ofdev.node = of_node_get(dn);
@@ -227,9 +227,9 @@ static struct ibmebus_dev* __devinit ibm
min(length, BUS_ID_SIZE - 1));
/* Register with generic device framework. */
- if (ibmebus_register_device_common(dev, dn->name) == NULL) {
+ if (ibmebus_register_device_common(dev, dn->name) != 0) {
kfree(dev);
- return NULL;
+ return ERR_PTR(-ENODEV);
}
return dev;
@@ -240,9 +240,8 @@ static void ibmebus_probe_of_nodes(char*
struct device_node *dn = NULL;
while ((dn = of_find_node_by_name(dn, name))) {
- if (ibmebus_register_device_node(dn) == NULL) {
+ if (IS_ERR(ibmebus_register_device_node(dn))) {
of_node_put(dn);
-
return;
}
}
@@ -390,30 +389,40 @@ static int ibmebus_match_helper_loc_code
return 0;
}
-static ssize_t ibmebus_store_probe(struct bus_type *dev,
+static ssize_t ibmebus_store_probe(struct bus_type *bus,
const char *buf, size_t count)
{
struct device_node *dn = NULL;
+ struct ibmebus_dev *dev;
char *loc_code;
+ buf[count] = '\0';
+ if (buf[count-1] == '\n')
+ buf[count-1] = '\0';
+
if (bus_find_device(_bus_type, NULL, (char*)buf,
ibmebus_match_helper_loc_code)) {
printk(KERN_WARNING "%s: loc_code %s has already been probed\n",
__FUNCTION__, buf);
- return count;
+ return -EINVAL;
}
while ((dn = of_find_all_nodes(dn))) {
loc_code = (char *)get_property(dn, "ibm,loc-code", NULL);
if (loc_code && (strncmp(loc_code, buf, count) == 0)) {
- if (ibmebus_register_device_node(dn) == NULL) {
+ dev = ibmebus_register_device_node(dn);
+ if (IS_ERR(dev)) {
of_node_put(dn);
- break;
- }
+ return PTR_ERR(dev);
+ } else
+ return count; /* success */
}
}
- return count;
+ /* if we drop out of the loop, the loc code was invalid */
+ printk(KERN_WARNING "%s: no device with loc_code %s found\n",
+ __FUNCTION__, buf);
+ return -ENODEV;
}
static BUS_ATTR(probe, S_IWUSR,
[PATCH 2.6.21-rc1] ibmebus: more error reporting in dynamic add/remove code
Writing the ibmebus probe and remove attributes now throws an appropriate
error if something goes wrong. This way, userspace tools can check for
success or failure of an addition or removal. The write will block until
the probe/remove operation completes, so, when the write operation returns
without an error, you can be sure the probe was successful and the device
is present in the system.
As an added bonus, an eventual trailing newline is now removed from the
written string.
Signed-off-by: Joachim Fenkes [EMAIL PROTECTED]
---
ibmebus.c | 54 --
1 files changed, 36 insertions(+), 18 deletions(-)
diff -urp a/arch/powerpc/kernel/ibmebus.c b/arch/powerpc/kernel/ibmebus.c
--- a/arch/powerpc/kernel/ibmebus.c 2007-02-20 21:54:22.366443424 +0100
+++ b/arch/powerpc/kernel/ibmebus.c 2007-02-20 23:31:39.950486616 +0100
@@ -167,7 +167,7 @@ static ssize_t ibmebusdev_show_name(stru
static DEVICE_ATTR(name, S_IRUSR | S_IRGRP | S_IROTH, ibmebusdev_show_name,
NULL);
-static struct ibmebus_dev* __devinit ibmebus_register_device_common(
+static int __devinit ibmebus_register_device_common(
struct ibmebus_dev *dev, const char *name)
{
int err = 0;
@@ -186,12 +186,12 @@ static struct ibmebus_dev* __devinit ibm
if ((err = of_device_register(dev-ofdev)) != 0) {
printk(KERN_ERR %s: failed to register device (%d).\n,
__FUNCTION__, err);
- return NULL;
+ return -ENODEV;
}
device_create_file(dev-ofdev.dev, dev_attr_name);
- return dev;
+ return 0;
}
static struct ibmebus_dev* __devinit ibmebus_register_device_node(
@@ -205,18 +205,18 @@ static struct ibmebus_dev* __devinit ibm
if (!loc_code) {
printk(KERN_WARNING %s: node %s missing 'ibm,loc-code'\n,
__FUNCTION__, dn-name ? dn-name : unknown);
- return NULL;
+ return ERR_PTR(-EINVAL);
}
if (strlen(loc_code) == 0) {
printk(KERN_WARNING %s: 'ibm,loc-code' is invalid\n,
__FUNCTION__);
- return NULL;
+ return ERR_PTR(-EINVAL);
}
dev = kzalloc(sizeof(struct ibmebus_dev), GFP_KERNEL);
if (!dev) {
- return NULL;
+ return ERR_PTR(-ENOMEM);
}
dev-ofdev.node = of_node_get(dn);
@@ -227,9 +227,9 @@ static struct ibmebus_dev* __devinit ibm
min(length, BUS_ID_SIZE - 1));
/* Register with generic device framework. */
- if (ibmebus_register_device_common(dev, dn-name) == NULL) {
+ if (ibmebus_register_device_common(dev, dn-name) != 0) {
kfree(dev);
- return NULL;
+ return ERR_PTR(-ENODEV);
}
return dev;
@@ -240,9 +240,8 @@ static void ibmebus_probe_of_nodes(char*
struct device_node *dn = NULL;
while ((dn = of_find_node_by_name(dn, name))) {
- if (ibmebus_register_device_node(dn) == NULL) {
+ if (IS_ERR(ibmebus_register_device_node(dn))) {
of_node_put(dn);
-
return;
}
}
@@ -390,30 +389,40 @@ static int ibmebus_match_helper_loc_code
return 0;
}
-static ssize_t ibmebus_store_probe(struct bus_type *dev,
+static ssize_t ibmebus_store_probe(struct bus_type *bus,
const char *buf, size_t count)
{
struct device_node *dn = NULL;
+ struct ibmebus_dev *dev;
char *loc_code;
+ buf[count] = '\0';
+ if (buf[count-1] == '\n')
+ buf[count-1] = '\0';
+
if (bus_find_device(ibmebus_bus_type, NULL, (char*)buf,
ibmebus_match_helper_loc_code)) {
printk(KERN_WARNING %s: loc_code %s has already been probed\n,
__FUNCTION__, buf);
- return count;
+ return -EINVAL;
}
while ((dn = of_find_all_nodes(dn))) {
loc_code = (char *)get_property(dn, ibm,loc-code, NULL);
if (loc_code (strncmp(loc_code, buf, count) == 0)) {
- if (ibmebus_register_device_node(dn) == NULL) {
+ dev = ibmebus_register_device_node(dn);
+ if (IS_ERR(dev)) {
of_node_put(dn);
- break;
- }
+ return PTR_ERR(dev);
+ } else
+ return count; /* success */
}
}
- return count;
+ /* if we drop out of the loop, the loc code was invalid */
+ printk(KERN_WARNING %s: no device with loc_code %s found\n,
+ __FUNCTION__, buf);
+ return -ENODEV;
}
static BUS_ATTR(probe, S_IWUSR,
[PATCH 2.6.21-rc1] ibmebus: make root device not of_device-based
The fake root device doesn't have an associated device tree node,
so it should not be an of_device. This patch makes it a normal,
bus-less device and thus saves a lot of checks later on.
Signed-off-by: Joachim Fenkes <[EMAIL PROTECTED]>
---
ibmebus.c | 33 ++---
1 files changed, 14 insertions(+), 19 deletions(-)
diff -urp a/arch/powerpc/kernel/ibmebus.c b/arch/powerpc/kernel/ibmebus.c
--- a/arch/powerpc/kernel/ibmebus.c 2007-02-20 01:35:54.0 +0100
+++ b/arch/powerpc/kernel/ibmebus.c 2007-02-20 01:39:24.0 +0100
@@ -44,9 +44,8 @@
#include
#include
-static struct ibmebus_dev ibmebus_bus_device = { /* fake "parent" device */
- .ofdev.dev.bus_id = "ibmebus",
- .ofdev.dev.bus= _bus_type,
+static struct device ibmebus_bus_device = { /* fake "parent" device */
+ .bus_id = "ibmebus",
};
static void *ibmebus_alloc_coherent(struct device *dev,
@@ -173,7 +172,7 @@ static struct ibmebus_dev* __devinit ibm
{
int err = 0;
- dev->ofdev.dev.parent = _bus_device.ofdev.dev;
+ dev->ofdev.dev.parent = _bus_device;
dev->ofdev.dev.bus = _bus_type;
dev->ofdev.dev.release = ibmebus_dev_release;
@@ -268,14 +267,12 @@ static int ibmebus_match_helper_name(str
const struct ibmebus_dev *ebus_dev = to_ibmebus_dev(dev);
char *name;
- /* parent device has no of_device node, so skip it */
- if (ebus_dev != _bus_device) {
- name = (char*)get_property(
- ebus_dev->ofdev.node, "name", NULL);
+ name = (char*)get_property(
+ ebus_dev->ofdev.node, "name", NULL);
+
+ if (name && (strcmp((char*)data, name) == 0))
+ return 1;
- if (name && (strcmp((char*)data, name) == 0))
- return 1;
- }
return 0;
}
@@ -384,14 +381,12 @@ static int ibmebus_match_helper_loc_code
const struct ibmebus_dev *ebus_dev = to_ibmebus_dev(dev);
char *loc_code;
- /* parent device has no of_device node, so skip it */
- if (ebus_dev != _bus_device) {
- loc_code = (char*)get_property(
- ebus_dev->ofdev.node, "ibm,loc-code", NULL);
+ loc_code = (char*)get_property(
+ ebus_dev->ofdev.node, "ibm,loc-code", NULL);
+
+ if (loc_code && (strcmp((char*)data, loc_code) == 0))
+ return 1;
- if (loc_code && (strcmp((char*)data, loc_code) == 0))
- return 1;
- }
return 0;
}
@@ -449,7 +444,7 @@ static int __init ibmebus_bus_init(void)
return err;
}
- err = device_register(_bus_device.ofdev.dev);
+ err = device_register(_bus_device);
if (err) {
printk(KERN_WARNING "%s: device_register returned %i\n",
__FUNCTION__, err);
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[PATCH 2.6.21-rc1] ibmebus: make root device not of_device-based
The fake root device doesn't have an associated device tree node,
so it should not be an of_device. This patch makes it a normal,
bus-less device and thus saves a lot of checks later on.
Signed-off-by: Joachim Fenkes [EMAIL PROTECTED]
---
ibmebus.c | 33 ++---
1 files changed, 14 insertions(+), 19 deletions(-)
diff -urp a/arch/powerpc/kernel/ibmebus.c b/arch/powerpc/kernel/ibmebus.c
--- a/arch/powerpc/kernel/ibmebus.c 2007-02-20 01:35:54.0 +0100
+++ b/arch/powerpc/kernel/ibmebus.c 2007-02-20 01:39:24.0 +0100
@@ -44,9 +44,8 @@
#include asm/ibmebus.h
#include asm/abs_addr.h
-static struct ibmebus_dev ibmebus_bus_device = { /* fake parent device */
- .ofdev.dev.bus_id = ibmebus,
- .ofdev.dev.bus= ibmebus_bus_type,
+static struct device ibmebus_bus_device = { /* fake parent device */
+ .bus_id = ibmebus,
};
static void *ibmebus_alloc_coherent(struct device *dev,
@@ -173,7 +172,7 @@ static struct ibmebus_dev* __devinit ibm
{
int err = 0;
- dev-ofdev.dev.parent = ibmebus_bus_device.ofdev.dev;
+ dev-ofdev.dev.parent = ibmebus_bus_device;
dev-ofdev.dev.bus = ibmebus_bus_type;
dev-ofdev.dev.release = ibmebus_dev_release;
@@ -268,14 +267,12 @@ static int ibmebus_match_helper_name(str
const struct ibmebus_dev *ebus_dev = to_ibmebus_dev(dev);
char *name;
- /* parent device has no of_device node, so skip it */
- if (ebus_dev != ibmebus_bus_device) {
- name = (char*)get_property(
- ebus_dev-ofdev.node, name, NULL);
+ name = (char*)get_property(
+ ebus_dev-ofdev.node, name, NULL);
+
+ if (name (strcmp((char*)data, name) == 0))
+ return 1;
- if (name (strcmp((char*)data, name) == 0))
- return 1;
- }
return 0;
}
@@ -384,14 +381,12 @@ static int ibmebus_match_helper_loc_code
const struct ibmebus_dev *ebus_dev = to_ibmebus_dev(dev);
char *loc_code;
- /* parent device has no of_device node, so skip it */
- if (ebus_dev != ibmebus_bus_device) {
- loc_code = (char*)get_property(
- ebus_dev-ofdev.node, ibm,loc-code, NULL);
+ loc_code = (char*)get_property(
+ ebus_dev-ofdev.node, ibm,loc-code, NULL);
+
+ if (loc_code (strcmp((char*)data, loc_code) == 0))
+ return 1;
- if (loc_code (strcmp((char*)data, loc_code) == 0))
- return 1;
- }
return 0;
}
@@ -449,7 +444,7 @@ static int __init ibmebus_bus_init(void)
return err;
}
- err = device_register(ibmebus_bus_device.ofdev.dev);
+ err = device_register(ibmebus_bus_device);
if (err) {
printk(KERN_WARNING %s: device_register returned %i\n,
__FUNCTION__, err);
-
To unsubscribe from this list: send the line unsubscribe linux-kernel in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[PATCH 2.6.21-rc1] powerpc: Make of_device_uevent() compatible with ibmebus
ibmebus has a fake root device that's not associated with an ofdt node. Filter out any such devices in of_device_uevent(). Signed-off-by: Joachim Fenkes <[EMAIL PROTECTED]> --- of_device.c |4 1 files changed, 4 insertions(+) diff -urp a/arch/powerpc/kernel/of_device.c b/arch/powerpc/kernel/of_device.c --- a/arch/powerpc/kernel/of_device.c 2007-02-17 16:36:32.116368480 +0100 +++ b/arch/powerpc/kernel/of_device.c 2007-02-17 16:44:01.319366352 +0100 @@ -180,6 +180,10 @@ int of_device_uevent(struct device *dev, ofdev = to_of_device(dev); + /* e.g. ibmebus has a fake root device w/o ofdt node -- filter that */ + if (!ofdev->node) + return -ENODEV; + if (add_uevent_var(envp, num_envp, , buffer, buffer_size, , "OF_NAME=%s", ofdev->node->name)) - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
[PATCH 2.6.21-rc1] powerpc: Make of_device_uevent() compatible with ibmebus
ibmebus has a fake root device that's not associated with an ofdt node. Filter out any such devices in of_device_uevent(). Signed-off-by: Joachim Fenkes [EMAIL PROTECTED] --- of_device.c |4 1 files changed, 4 insertions(+) diff -urp a/arch/powerpc/kernel/of_device.c b/arch/powerpc/kernel/of_device.c --- a/arch/powerpc/kernel/of_device.c 2007-02-17 16:36:32.116368480 +0100 +++ b/arch/powerpc/kernel/of_device.c 2007-02-17 16:44:01.319366352 +0100 @@ -180,6 +180,10 @@ int of_device_uevent(struct device *dev, ofdev = to_of_device(dev); + /* e.g. ibmebus has a fake root device w/o ofdt node -- filter that */ + if (!ofdev-node) + return -ENODEV; + if (add_uevent_var(envp, num_envp, i, buffer, buffer_size, length, OF_NAME=%s, ofdev-node-name)) - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
[PATCH 2.6.21-rc1] ibmebus: Use of_device_uevent()
This patch replaces ibmebus_uevent() by Sylvain's generic function.
Signed-off-by: Joachim Fenkes <[EMAIL PROTECTED]>
---
ibmebus.c | 37 +++--
1 file changed, 3 insertions(+), 34 deletions(-)
diff -urp a/arch/powerpc/kernel/ibmebus.c b/arch/powerpc/kernel/ibmebus.c
--- a/arch/powerpc/kernel/ibmebus.c 2007-02-17 03:25:53.403375784 +0100
+++ b/arch/powerpc/kernel/ibmebus.c 2007-02-17 03:27:28.230336112 +0100
@@ -372,41 +372,10 @@ static int ibmebus_bus_match(struct devi
return 0;
}
-static int ibmebus_uevent(struct device *dev, char **envp, int num_envp,
- char *buffer, int buffer_size)
-{
- const struct ibmebus_dev *ebus_dev = to_ibmebus_dev(dev);
- char *name, *cp, *loc_code;
- int length;
-
- if (!num_envp)
- return -ENOMEM;
-
- if (!ebus_dev->ofdev.node)
- return -ENODEV;
-
- name = (char *)get_property(ebus_dev->ofdev.node, "name", NULL);
- cp = (char *)get_property(ebus_dev->ofdev.node, "compatible", NULL);
- loc_code = (char *)get_property(ebus_dev->ofdev.node,
- "ibm,loc-code", NULL);
- if (!(name && cp && loc_code))
- return -ENODEV;
-
- envp[0] = buffer;
- length = scnprintf(buffer, buffer_size,
- "MODALIAS=ibmebus:T%s:S%s:L%s",
- name, cp, loc_code);
- if (buffer_size - length <= 0)
- return -ENOMEM;
- envp[1] = NULL;
-
- return 0;
-}
-
struct bus_type ibmebus_bus_type = {
- .name = "ibmebus",
- .uevent = ibmebus_uevent,
- .match = ibmebus_bus_match,
+ .name = "ibmebus",
+ .uevent = of_device_uevent,
+ .match = ibmebus_bus_match,
};
EXPORT_SYMBOL(ibmebus_bus_type);
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[PATCH 2.6.21-rc1] ibmebus: Use of_device_uevent()
This patch replaces ibmebus_uevent() by Sylvain's generic function.
Signed-off-by: Joachim Fenkes [EMAIL PROTECTED]
---
ibmebus.c | 37 +++--
1 file changed, 3 insertions(+), 34 deletions(-)
diff -urp a/arch/powerpc/kernel/ibmebus.c b/arch/powerpc/kernel/ibmebus.c
--- a/arch/powerpc/kernel/ibmebus.c 2007-02-17 03:25:53.403375784 +0100
+++ b/arch/powerpc/kernel/ibmebus.c 2007-02-17 03:27:28.230336112 +0100
@@ -372,41 +372,10 @@ static int ibmebus_bus_match(struct devi
return 0;
}
-static int ibmebus_uevent(struct device *dev, char **envp, int num_envp,
- char *buffer, int buffer_size)
-{
- const struct ibmebus_dev *ebus_dev = to_ibmebus_dev(dev);
- char *name, *cp, *loc_code;
- int length;
-
- if (!num_envp)
- return -ENOMEM;
-
- if (!ebus_dev-ofdev.node)
- return -ENODEV;
-
- name = (char *)get_property(ebus_dev-ofdev.node, name, NULL);
- cp = (char *)get_property(ebus_dev-ofdev.node, compatible, NULL);
- loc_code = (char *)get_property(ebus_dev-ofdev.node,
- ibm,loc-code, NULL);
- if (!(name cp loc_code))
- return -ENODEV;
-
- envp[0] = buffer;
- length = scnprintf(buffer, buffer_size,
- MODALIAS=ibmebus:T%s:S%s:L%s,
- name, cp, loc_code);
- if (buffer_size - length = 0)
- return -ENOMEM;
- envp[1] = NULL;
-
- return 0;
-}
-
struct bus_type ibmebus_bus_type = {
- .name = ibmebus,
- .uevent = ibmebus_uevent,
- .match = ibmebus_bus_match,
+ .name = ibmebus,
+ .uevent = of_device_uevent,
+ .match = ibmebus_bus_match,
};
EXPORT_SYMBOL(ibmebus_bus_type);
-
To unsubscribe from this list: send the line unsubscribe linux-kernel in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Re: [PATCH 2.6.21-rc1 4/5] ehca: replace yield() by wait_for_completion()
Hi, > So I'm holding off on applying this for now. Please think it over and > either tell me the current patch is OK, or fix it up. There's not > really too much urgency because a change like this is something I > would be comfortable merging between 2.6.21-rc1 and -rc2. You're absolutely right. Let's target for rc2. Thanks for this good catch! Nam - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
[PATCH 2.6.21-rc1] ibmebus: Support dynamic addition and removal of adapters
This patch will add two sysfs attributes to /sys/bus/ibmebus which can be used
to notify the ebus driver of added / removed ebus devices in the OF device
tree.
Echoing the device's location code (as found in the OFDT "ibm,loc-code"
property) into the "probe" attribute will notify ebus of addition of the device
and cause the appropriate device driver's probe function to be called on the
device.
Likewise, echoing the location code into the "remove" attribute will cause
the device to be removed from the system.
Additionally, the uevent interface is now implemented in the driver.
Signed-off-by: Joachim Fenkes <[EMAIL PROTECTED]>
---
arch/powerpc/kernel/ibmebus.c | 121 +++---
include/asm-powerpc/ibmebus.h |2
2 files changed, 114 insertions(+), 9 deletions(-)
diff -wurp linux-2.6.20/arch/powerpc/kernel/ibmebus.c
linux-2.6.20-ebus/arch/powerpc/kernel/ibmebus.c
--- linux-2.6.20/arch/powerpc/kernel/ibmebus.c 2007-02-04 19:44:54.0
+0100
+++ linux-2.6.20-ebus/arch/powerpc/kernel/ibmebus.c 2007-02-14
17:58:00.0 +0100
@@ -2,6 +2,7 @@
* IBM PowerPC IBM eBus Infrastructure Support.
*
* Copyright (c) 2005 IBM Corporation
+ * Joachim Fenkes <[EMAIL PROTECTED]>
* Heiko J Schick <[EMAIL PROTECTED]>
*
* All rights reserved.
@@ -44,7 +45,6 @@
#include
static struct ibmebus_dev ibmebus_bus_device = { /* fake "parent" device */
- .name = ibmebus_bus_device.ofdev.dev.bus_id,
.ofdev.dev.bus_id = "ibmebus",
.ofdev.dev.bus= _bus_type,
};
@@ -161,7 +161,9 @@ static void __devinit ibmebus_dev_releas
static ssize_t ibmebusdev_show_name(struct device *dev,
struct device_attribute *attr, char *buf)
{
- return sprintf(buf, "%s\n", to_ibmebus_dev(dev)->name);
+ struct ibmebus_dev *ebus_dev = to_ibmebus_dev(dev);
+ char *name = (char*)get_property(ebus_dev->ofdev.node, "name", NULL);
+ return sprintf(buf, "%s\n", name);
}
static DEVICE_ATTR(name, S_IRUSR | S_IRGRP | S_IROTH, ibmebusdev_show_name,
NULL);
@@ -171,7 +173,6 @@ static struct ibmebus_dev* __devinit ibm
{
int err = 0;
- dev->name = name;
dev->ofdev.dev.parent = _bus_device.ofdev.dev;
dev->ofdev.dev.bus = _bus_type;
dev->ofdev.dev.release = ibmebus_dev_release;
@@ -262,11 +263,19 @@ static void ibmebus_add_devices_by_id(st
return;
}
-static int ibmebus_match_helper(struct device *dev, void *data)
+static int ibmebus_match_helper_name(struct device *dev, void *data)
{
- if (strcmp((char*)data, to_ibmebus_dev(dev)->name) == 0)
- return 1;
+ const struct ibmebus_dev *ebus_dev = to_ibmebus_dev(dev);
+ char *name;
+ /* parent device has no of_device node, so skip it */
+ if (ebus_dev != _bus_device) {
+ name = (char*)get_property(
+ ebus_dev->ofdev.node, "name", NULL);
+
+ if (name && (strcmp((char*)data, name) == 0))
+ return 1;
+ }
return 0;
}
@@ -285,11 +294,10 @@ static void ibmebus_remove_devices_by_id
while (strlen(idt->name) > 0) {
while ((dev = bus_find_device(_bus_type, NULL,
(void*)idt->name,
- ibmebus_match_helper))) {
+ ibmebus_match_helper_name))) {
ibmebus_unregister_device(dev);
}
idt++;
-
}
return;
@@ -307,6 +315,9 @@ int ibmebus_register_driver(struct ibmeb
if ((err = driver_register(>driver) != 0))
return err;
+ /* remove all supported devices first, in case someone
+* probed them manually before registering the driver */
+ ibmebus_remove_devices_by_id(drv->id_table);
ibmebus_add_devices_by_id(drv->id_table);
return 0;
@@ -361,12 +372,101 @@ static int ibmebus_bus_match(struct devi
return 0;
}
+static int ibmebus_uevent(struct device *dev, char **envp, int num_envp,
+ char *buffer, int buffer_size)
+{
+ const struct ibmebus_dev *ebus_dev = to_ibmebus_dev(dev);
+ char *name, *cp, *loc_code;
+ int length;
+
+ if (!num_envp)
+ return -ENOMEM;
+
+ if (!ebus_dev->ofdev.node)
+ return -ENODEV;
+
+ name = (char *)get_property(ebus_dev->ofdev.node, "name", NULL);
+ cp = (char *)get_property(ebus_dev->ofdev.node, "compatible", NULL);
+ loc_code = (char *)get_property(ebus_dev->ofdev.node,
+ "ibm,loc-code", NULL);
+ if (!(name && cp && loc_code))
+ return -ENODEV;
+
+ envp[0] = buffer;
+ length = scnprintf(buffer, buffer_size,
+
[PATCH 2.6.21-rc1 5/5] ehca: query_port() returns LINK_UP instead UNKNOWN
set port phys state as a result of ehca_query_port() to LINK_UP. On pSeries ehca actually represents a logical HCA, whose phys/link state always is LINK_UP. Signed-off-by: Hoang-Nam Nguyen <[EMAIL PROTECTED]> --- ehca_hca.c |3 +++ 1 files changed, 3 insertions(+) diff --git a/drivers/infiniband/hw/ehca/ehca_hca.c b/drivers/infiniband/hw/ehca/ehca_hca.c index b7be950..30eb45d 100644 --- a/drivers/infiniband/hw/ehca/ehca_hca.c +++ b/drivers/infiniband/hw/ehca/ehca_hca.c @@ -162,6 +162,9 @@ int ehca_query_port(struct ib_device *ib props->active_width= IB_WIDTH_12X; props->active_speed= 0x1; + /* at the moment (logical) link state is always LINK_UP */ + props->phys_state = 0x5; + query_port1: ehca_free_fw_ctrlblock(rblock); - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
[PATCH 2.6.21-rc1 3/5] ehca: allow en/disabling scaling code via module parameter
allow users to en/disable scaling code when loading ib_ehca module
Signed-off-by: Hoang-Nam Nguyen <[EMAIL PROTECTED]>
---
Kconfig|8
ehca_classes.h |1 +
ehca_irq.c | 47 +--
ehca_main.c|4
4 files changed, 26 insertions(+), 34 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/Kconfig
b/drivers/infiniband/hw/ehca/Kconfig
index 727b10d..1a85459 100644
--- a/drivers/infiniband/hw/ehca/Kconfig
+++ b/drivers/infiniband/hw/ehca/Kconfig
@@ -7,11 +7,3 @@ config INFINIBAND_EHCA
To compile the driver as a module, choose M here. The module
will be called ib_ehca.
-config INFINIBAND_EHCA_SCALING
- bool "Scaling support (EXPERIMENTAL)"
- depends on IBMEBUS && INFINIBAND_EHCA && HOTPLUG_CPU && EXPERIMENTAL
- default y
- ---help---
- eHCA scaling support schedules the CQ callbacks to different CPUs.
-
- To enable this feature choose Y here.
diff --git a/drivers/infiniband/hw/ehca/ehca_classes.h
b/drivers/infiniband/hw/ehca/ehca_classes.h
index f08ad6f..40404c9 100644
--- a/drivers/infiniband/hw/ehca/ehca_classes.h
+++ b/drivers/infiniband/hw/ehca/ehca_classes.h
@@ -277,6 +277,7 @@ extern struct idr ehca_cq_idr;
extern int ehca_static_rate;
extern int ehca_port_act_time;
extern int ehca_use_hp_mr;
+extern int ehca_scaling_code;
struct ipzu_queue_resp {
u32 qe_size; /* queue entry size */
diff --git a/drivers/infiniband/hw/ehca/ehca_irq.c
b/drivers/infiniband/hw/ehca/ehca_irq.c
index 9679b07..3ec53c6 100644
--- a/drivers/infiniband/hw/ehca/ehca_irq.c
+++ b/drivers/infiniband/hw/ehca/ehca_irq.c
@@ -63,15 +63,11 @@
#define ERROR_DATA_LENGTH EHCA_BMASK_IBM(52,63)
#define ERROR_DATA_TYPEEHCA_BMASK_IBM(0,7)
-#ifdef CONFIG_INFINIBAND_EHCA_SCALING
-
static void queue_comp_task(struct ehca_cq *__cq);
static struct ehca_comp_pool* pool;
static struct notifier_block comp_pool_callback_nb;
-#endif
-
static inline void comp_event_callback(struct ehca_cq *cq)
{
if (!cq->ib_cq.comp_handler)
@@ -423,13 +419,13 @@ static inline void process_eqe(struct eh
return;
}
reset_eq_pending(cq);
-#ifdef CONFIG_INFINIBAND_EHCA_SCALING
- queue_comp_task(cq);
- spin_unlock_irqrestore(_cq_idr_lock, flags);
-#else
- spin_unlock_irqrestore(_cq_idr_lock, flags);
- comp_event_callback(cq);
-#endif
+ if (ehca_scaling_code) {
+ queue_comp_task(cq);
+ spin_unlock_irqrestore(_cq_idr_lock, flags);
+ } else {
+ spin_unlock_irqrestore(_cq_idr_lock, flags);
+ comp_event_callback(cq);
+ }
} else {
ehca_dbg(>ib_device,
"Got non completion event");
@@ -508,13 +504,12 @@ void ehca_process_eq(struct ehca_shca *s
/* call completion handler for cached eqes */
for (i = 0; i < eqe_cnt; i++)
if (eq->eqe_cache[i].cq) {
-#ifdef CONFIG_INFINIBAND_EHCA_SCALING
- spin_lock(_cq_idr_lock);
- queue_comp_task(eq->eqe_cache[i].cq);
- spin_unlock(_cq_idr_lock);
-#else
- comp_event_callback(eq->eqe_cache[i].cq);
-#endif
+ if (ehca_scaling_code) {
+ spin_lock(_cq_idr_lock);
+ queue_comp_task(eq->eqe_cache[i].cq);
+ spin_unlock(_cq_idr_lock);
+ } else
+ comp_event_callback(eq->eqe_cache[i].cq);
} else {
ehca_dbg(>ib_device, "Got non completion event");
parse_identifier(shca, eq->eqe_cache[i].eqe->entry);
@@ -540,8 +535,6 @@ void ehca_tasklet_eq(unsigned long data)
ehca_process_eq((struct ehca_shca*)data, 1);
}
-#ifdef CONFIG_INFINIBAND_EHCA_SCALING
-
static inline int find_next_online_cpu(struct ehca_comp_pool* pool)
{
int cpu;
@@ -764,14 +757,14 @@ static int comp_pool_callback(struct not
return NOTIFY_OK;
}
-#endif
-
int ehca_create_comp_pool(void)
{
-#ifdef CONFIG_INFINIBAND_EHCA_SCALING
int cpu;
struct task_struct *task;
+ if (!ehca_scaling_code)
+ return 0;
+
pool = kzalloc(sizeof(struct ehca_comp_pool), GFP_KERNEL);
if (pool == NULL)
return -ENOMEM;
@@ -796,16 +789,19 @@ int ehca_create_comp_pool(void)
comp_pool_callback_nb.notifier_call = comp_pool_callback;
comp_pool_callback_nb.priority =0;
register_cpu_notifier(_pool_callback_nb);
-#endif
+
+ printk(KERN_INFO "eHCA scaling code enabled\n");
re
[PATCH 2.6.21-rc1 4/5] ehca: replace yield() by wait_for_completion()
remove yield() and use wait_for_completion() in order to wait for running
completion handlers finished before destroying associated completion queue
Signed-off-by: Hoang-Nam Nguyen <[EMAIL PROTECTED]>
---
ehca_classes.h |3 +++
ehca_cq.c |5 +++--
ehca_irq.c |6 +-
3 files changed, 11 insertions(+), 3 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_classes.h
b/drivers/infiniband/hw/ehca/ehca_classes.h
index 40404c9..d8ce0c8 100644
--- a/drivers/infiniband/hw/ehca/ehca_classes.h
+++ b/drivers/infiniband/hw/ehca/ehca_classes.h
@@ -52,6 +52,8 @@ struct ehca_mw;
struct ehca_pd;
struct ehca_av;
+#include
+
#include
#include
@@ -154,6 +156,7 @@ struct ehca_cq {
struct hlist_head qp_hashtab[QP_HASHTAB_LEN];
struct list_head entry;
u32 nr_callbacks;
+ struct completion zero_callbacks;
spinlock_t task_lock;
u32 ownpid;
/* mmap counter for resources mapped into user space */
diff --git a/drivers/infiniband/hw/ehca/ehca_cq.c
b/drivers/infiniband/hw/ehca/ehca_cq.c
index 9291a86..906bd5b 100644
--- a/drivers/infiniband/hw/ehca/ehca_cq.c
+++ b/drivers/infiniband/hw/ehca/ehca_cq.c
@@ -147,6 +147,7 @@ struct ib_cq *ehca_create_cq(struct ib_d
spin_lock_init(_cq->spinlock);
spin_lock_init(_cq->cb_lock);
spin_lock_init(_cq->task_lock);
+ init_completion(_cq->zero_callbacks);
my_cq->ownpid = current->tgid;
cq = _cq->ib_cq;
@@ -330,9 +331,9 @@ int ehca_destroy_cq(struct ib_cq *cq)
}
spin_lock_irqsave(_cq_idr_lock, flags);
- while (my_cq->nr_callbacks) {
+ if (my_cq->nr_callbacks) {
spin_unlock_irqrestore(_cq_idr_lock, flags);
- yield();
+ wait_for_completion(_cq->zero_callbacks);
spin_lock_irqsave(_cq_idr_lock, flags);
}
diff --git a/drivers/infiniband/hw/ehca/ehca_irq.c
b/drivers/infiniband/hw/ehca/ehca_irq.c
index 3ec53c6..7db39b7 100644
--- a/drivers/infiniband/hw/ehca/ehca_irq.c
+++ b/drivers/infiniband/hw/ehca/ehca_irq.c
@@ -605,6 +605,7 @@ static void run_comp_task(struct ehca_cp
spin_lock_irqsave(>task_lock, flags);
while (!list_empty(>cq_list)) {
+ int is_complete = 0;
cq = list_entry(cct->cq_list.next, struct ehca_cq, entry);
spin_unlock_irqrestore(>task_lock, flags);
comp_event_callback(cq);
@@ -612,11 +613,14 @@ static void run_comp_task(struct ehca_cp
spin_lock(>task_lock);
cq->nr_callbacks--;
- if (cq->nr_callbacks == 0) {
+ is_complete = (cq->nr_callbacks == 0);
+ if (is_complete) {
list_del_init(cct->cq_list.next);
cct->cq_jobs--;
}
spin_unlock(>task_lock);
+ if (is_complete) /* wake up waiting destroy_cq() */
+ complete(>zero_callbacks);
}
spin_unlock_irqrestore(>task_lock, flags);
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[PATCH 2.6.21-rc1 2/5] ehca: fix race condition/locking issues in scaling code
fix a race condition in find_next_cpu_online() and some
other locking issues in scaling code
Signed-off-by: Hoang-Nam Nguyen <[EMAIL PROTECTED]>
---
ehca_irq.c | 68 +
1 files changed, 33 insertions(+), 35 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_irq.c
b/drivers/infiniband/hw/ehca/ehca_irq.c
index b923b5d..9679b07 100644
--- a/drivers/infiniband/hw/ehca/ehca_irq.c
+++ b/drivers/infiniband/hw/ehca/ehca_irq.c
@@ -544,28 +544,30 @@ void ehca_tasklet_eq(unsigned long data)
static inline int find_next_online_cpu(struct ehca_comp_pool* pool)
{
- unsigned long flags_last_cpu;
+ int cpu;
+ unsigned long flags;
+ WARN_ON_ONCE(!in_interrupt());
if (ehca_debug_level)
ehca_dmp(_online_map, sizeof(cpumask_t), "");
- spin_lock_irqsave(>last_cpu_lock, flags_last_cpu);
- pool->last_cpu = next_cpu(pool->last_cpu, cpu_online_map);
- if (pool->last_cpu == NR_CPUS)
- pool->last_cpu = first_cpu(cpu_online_map);
- spin_unlock_irqrestore(>last_cpu_lock, flags_last_cpu);
+ spin_lock_irqsave(>last_cpu_lock, flags);
+ cpu = next_cpu(pool->last_cpu, cpu_online_map);
+ if (cpu == NR_CPUS)
+ cpu = first_cpu(cpu_online_map);
+ pool->last_cpu = cpu;
+ spin_unlock_irqrestore(>last_cpu_lock, flags);
- return pool->last_cpu;
+ return cpu;
}
static void __queue_comp_task(struct ehca_cq *__cq,
struct ehca_cpu_comp_task *cct)
{
- unsigned long flags_cct;
- unsigned long flags_cq;
+ unsigned long flags;
- spin_lock_irqsave(>task_lock, flags_cct);
- spin_lock_irqsave(&__cq->task_lock, flags_cq);
+ spin_lock_irqsave(>task_lock, flags);
+ spin_lock(&__cq->task_lock);
if (__cq->nr_callbacks == 0) {
__cq->nr_callbacks++;
@@ -576,8 +578,8 @@ static void __queue_comp_task(struct ehc
else
__cq->nr_callbacks++;
- spin_unlock_irqrestore(&__cq->task_lock, flags_cq);
- spin_unlock_irqrestore(>task_lock, flags_cct);
+ spin_unlock(&__cq->task_lock);
+ spin_unlock_irqrestore(>task_lock, flags);
}
static void queue_comp_task(struct ehca_cq *__cq)
@@ -588,69 +590,69 @@ static void queue_comp_task(struct ehca_
cpu = get_cpu();
cpu_id = find_next_online_cpu(pool);
-
BUG_ON(!cpu_online(cpu_id));
cct = per_cpu_ptr(pool->cpu_comp_tasks, cpu_id);
+ BUG_ON(!cct);
if (cct->cq_jobs > 0) {
cpu_id = find_next_online_cpu(pool);
cct = per_cpu_ptr(pool->cpu_comp_tasks, cpu_id);
+ BUG_ON(!cct);
}
__queue_comp_task(__cq, cct);
-
- put_cpu();
-
- return;
}
static void run_comp_task(struct ehca_cpu_comp_task* cct)
{
struct ehca_cq *cq;
- unsigned long flags_cct;
- unsigned long flags_cq;
+ unsigned long flags;
- spin_lock_irqsave(>task_lock, flags_cct);
+ spin_lock_irqsave(>task_lock, flags);
while (!list_empty(>cq_list)) {
cq = list_entry(cct->cq_list.next, struct ehca_cq, entry);
- spin_unlock_irqrestore(>task_lock, flags_cct);
+ spin_unlock_irqrestore(>task_lock, flags);
comp_event_callback(cq);
- spin_lock_irqsave(>task_lock, flags_cct);
+ spin_lock_irqsave(>task_lock, flags);
- spin_lock_irqsave(>task_lock, flags_cq);
+ spin_lock(>task_lock);
cq->nr_callbacks--;
if (cq->nr_callbacks == 0) {
list_del_init(cct->cq_list.next);
cct->cq_jobs--;
}
- spin_unlock_irqrestore(>task_lock, flags_cq);
-
+ spin_unlock(>task_lock);
}
- spin_unlock_irqrestore(>task_lock, flags_cct);
-
- return;
+ spin_unlock_irqrestore(>task_lock, flags);
}
static int comp_task(void *__cct)
{
struct ehca_cpu_comp_task* cct = __cct;
+ int cql_empty;
DECLARE_WAITQUEUE(wait, current);
set_current_state(TASK_INTERRUPTIBLE);
while(!kthread_should_stop()) {
add_wait_queue(>wait_queue, );
- if (list_empty(>cq_list))
+ spin_lock_irq(>task_lock);
+ cql_empty = list_empty(>cq_list);
+ spin_unlock_irq(>task_lock);
+ if (cql_empty)
schedule();
else
__set_current_state(TASK_RUNNING);
remove_wait_queue(>wait_queue, );
- if (!list_empty(>cq_list))
+ spin_lock_ir
[PATCH 2.6.21-rc1 1/5] ehca: reworked irq handler to avoid/reduce missed irq events
reworked irq handler to avoid/reduce missed irq events
Signed-off-by: Hoang-Nam Nguyen <[EMAIL PROTECTED]>
---
ehca_classes.h | 18 +++-
ehca_eq.c |1
ehca_irq.c | 214 +++--
ehca_irq.h |1
ehca_main.c| 28 +--
ipz_pt_fn.h| 11 ++
6 files changed, 182 insertions(+), 91 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_classes.h
b/drivers/infiniband/hw/ehca/ehca_classes.h
index cf95ee4..f08ad6f 100644
--- a/drivers/infiniband/hw/ehca/ehca_classes.h
+++ b/drivers/infiniband/hw/ehca/ehca_classes.h
@@ -42,8 +42,6 @@
#ifndef __EHCA_CLASSES_H__
#define __EHCA_CLASSES_H__
-#include "ehca_classes.h"
-#include "ipz_pt_fn.h"
struct ehca_module;
struct ehca_qp;
@@ -54,14 +52,22 @@ struct ehca_mw;
struct ehca_pd;
struct ehca_av;
+#include
+#include
+
#ifdef CONFIG_PPC64
#include "ehca_classes_pSeries.h"
#endif
+#include "ipz_pt_fn.h"
+#include "ehca_qes.h"
+#include "ehca_irq.h"
-#include
-#include
+#define EHCA_EQE_CACHE_SIZE 20
-#include "ehca_irq.h"
+struct ehca_eqe_cache_entry {
+ struct ehca_eqe *eqe;
+ struct ehca_cq *cq;
+};
struct ehca_eq {
u32 length;
@@ -74,6 +80,8 @@ struct ehca_eq {
spinlock_t spinlock;
struct tasklet_struct interrupt_task;
u32 ist;
+ spinlock_t irq_spinlock;
+ struct ehca_eqe_cache_entry eqe_cache[EHCA_EQE_CACHE_SIZE];
};
struct ehca_sport {
diff --git a/drivers/infiniband/hw/ehca/ehca_eq.c
b/drivers/infiniband/hw/ehca/ehca_eq.c
index 5281dec..33c822e 100644
--- a/drivers/infiniband/hw/ehca/ehca_eq.c
+++ b/drivers/infiniband/hw/ehca/ehca_eq.c
@@ -61,6 +61,7 @@ int ehca_create_eq(struct ehca_shca *shc
struct ib_device *ib_dev = >ib_device;
spin_lock_init(>spinlock);
+ spin_lock_init(>irq_spinlock);
eq->is_initialized = 0;
if (type != EHCA_EQ && type != EHCA_NEQ) {
diff --git a/drivers/infiniband/hw/ehca/ehca_irq.c
b/drivers/infiniband/hw/ehca/ehca_irq.c
index 6c4f9f9..b923b5d 100644
--- a/drivers/infiniband/hw/ehca/ehca_irq.c
+++ b/drivers/infiniband/hw/ehca/ehca_irq.c
@@ -206,7 +206,7 @@ static void qp_event_callback(struct ehc
}
static void cq_event_callback(struct ehca_shca *shca,
- u64 eqe)
+ u64 eqe)
{
struct ehca_cq *cq;
unsigned long flags;
@@ -318,7 +318,7 @@ static void parse_ec(struct ehca_shca *s
"disruptive port %x configuration change", port);
ehca_info(>ib_device,
-"port %x is inactive.", port);
+ "port %x is inactive.", port);
event.device = >ib_device;
event.event = IB_EVENT_PORT_ERR;
event.element.port_num = port;
@@ -326,7 +326,7 @@ static void parse_ec(struct ehca_shca *s
ib_dispatch_event();
ehca_info(>ib_device,
-"port %x is active.", port);
+ "port %x is active.", port);
event.device = >ib_device;
event.event = IB_EVENT_PORT_ACTIVE;
event.element.port_num = port;
@@ -401,87 +401,143 @@ irqreturn_t ehca_interrupt_eq(int irq, v
return IRQ_HANDLED;
}
-void ehca_tasklet_eq(unsigned long data)
-{
- struct ehca_shca *shca = (struct ehca_shca*)data;
- struct ehca_eqe *eqe;
- int int_state;
- int query_cnt = 0;
- do {
- eqe = (struct ehca_eqe *)ehca_poll_eq(shca, >eq);
+static inline void process_eqe(struct ehca_shca *shca, struct ehca_eqe *eqe)
+{
+ u64 eqe_value;
+ u32 token;
+ unsigned long flags;
+ struct ehca_cq *cq;
+ eqe_value = eqe->entry;
+ ehca_dbg(>ib_device, "eqe_value=%lx", eqe_value);
+ if (EHCA_BMASK_GET(EQE_COMPLETION_EVENT, eqe_value)) {
+ ehca_dbg(>ib_device, "... completion event");
+ token = EHCA_BMASK_GET(EQE_CQ_TOKEN, eqe_value);
+ spin_lock_irqsave(_cq_idr_lock, flags);
+ cq = idr_find(_cq_idr, token);
+ if (cq == NULL) {
+ spin_unlock_irqrestore(_cq_idr_lock, flags);
+ ehca_err(>ib_device,
+"Invalid eqe for non-existing cq token=%x",
+token);
+ return;
+ }
+ reset_eq_pending(cq);
+#ifdef CONFIG_INFINIBAND_EHCA_SCALING
+ queue_comp_task(cq);
+ spin_unlock_irqrestore(_cq_idr_lock, flags);
+#else
+ spin_unlock_irqrestore(_cq_idr_lock, flags);
+ comp_event_callback(cq);
+#endif
+ } e
Re: [PATCH 2.6.21-rc1 1/5] ehca: reworked irq handler to avoid/reduce missed irq events
> Looks fine but this patch at least has serious whitespace > damage... please resend a fixed version. Sorry for this. Resending the patches 1-5. Nam - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [openib-general] 32-bit build for ppc64 is required
And the customers can pick the one(s) they like. I see your point regarding QA effort. Is it really twice? Probably more - I'm reasonably sure most scripts written so far assume stuff is installed in prefix/bin, so testing harness etc would need to be changed. And how to make sure the *correct* set of binaries was actually QA'd? As far as I understood each component owner is responsible for QA of her/his component and supported platforms. For PPC we cover that. If we have 32bit execs only for development resp. testing, why don't we also give them to customers in order to do basic test or diagnosis of their setup? Because of the confusion this would create. For shared libraries, the 32/64 bit issues seem to be automatically figured out by ld.so, but there's no such solution for binaries. This is true as we have discussed at ofed-1.1. See also http://openib.org/pipermail/openfabrics-ewg/2006-October/001831.html I agree with you in that if there is a standard for binaries dir struct let's go for it. If there is no such one, let's agree on one approach: either bin32 resp bin or appl resp appl64 or... To me it's worse if customers have to fix or write build scripts by themselves in order to build 32bit binaries. Nam ___ openib-general mailing list openib-general@openib.org http://openib.org/mailman/listinfo/openib-general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
Re: [openib-general] [PATCH 2.6.21-rc1 1/5] ehca: reworked irq handler to avoid/reduce missed irq events
Looks fine but this patch at least has serious whitespace damage... please resend a fixed version. Sorry for this. Resending the patches 1-5. Nam ___ openib-general mailing list openib-general@openib.org http://openib.org/mailman/listinfo/openib-general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
[openib-general] [PATCH 2.6.21-rc1 1/5] ehca: reworked irq handler to avoid/reduce missed irq events
reworked irq handler to avoid/reduce missed irq events
Signed-off-by: Hoang-Nam Nguyen [EMAIL PROTECTED]
---
ehca_classes.h | 18 +++-
ehca_eq.c |1
ehca_irq.c | 214 +++--
ehca_irq.h |1
ehca_main.c| 28 +--
ipz_pt_fn.h| 11 ++
6 files changed, 182 insertions(+), 91 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_classes.h
b/drivers/infiniband/hw/ehca/ehca_classes.h
index cf95ee4..f08ad6f 100644
--- a/drivers/infiniband/hw/ehca/ehca_classes.h
+++ b/drivers/infiniband/hw/ehca/ehca_classes.h
@@ -42,8 +42,6 @@
#ifndef __EHCA_CLASSES_H__
#define __EHCA_CLASSES_H__
-#include ehca_classes.h
-#include ipz_pt_fn.h
struct ehca_module;
struct ehca_qp;
@@ -54,14 +52,22 @@ struct ehca_mw;
struct ehca_pd;
struct ehca_av;
+#include rdma/ib_verbs.h
+#include rdma/ib_user_verbs.h
+
#ifdef CONFIG_PPC64
#include ehca_classes_pSeries.h
#endif
+#include ipz_pt_fn.h
+#include ehca_qes.h
+#include ehca_irq.h
-#include rdma/ib_verbs.h
-#include rdma/ib_user_verbs.h
+#define EHCA_EQE_CACHE_SIZE 20
-#include ehca_irq.h
+struct ehca_eqe_cache_entry {
+ struct ehca_eqe *eqe;
+ struct ehca_cq *cq;
+};
struct ehca_eq {
u32 length;
@@ -74,6 +80,8 @@ struct ehca_eq {
spinlock_t spinlock;
struct tasklet_struct interrupt_task;
u32 ist;
+ spinlock_t irq_spinlock;
+ struct ehca_eqe_cache_entry eqe_cache[EHCA_EQE_CACHE_SIZE];
};
struct ehca_sport {
diff --git a/drivers/infiniband/hw/ehca/ehca_eq.c
b/drivers/infiniband/hw/ehca/ehca_eq.c
index 5281dec..33c822e 100644
--- a/drivers/infiniband/hw/ehca/ehca_eq.c
+++ b/drivers/infiniband/hw/ehca/ehca_eq.c
@@ -61,6 +61,7 @@ int ehca_create_eq(struct ehca_shca *shc
struct ib_device *ib_dev = shca-ib_device;
spin_lock_init(eq-spinlock);
+ spin_lock_init(eq-irq_spinlock);
eq-is_initialized = 0;
if (type != EHCA_EQ type != EHCA_NEQ) {
diff --git a/drivers/infiniband/hw/ehca/ehca_irq.c
b/drivers/infiniband/hw/ehca/ehca_irq.c
index 6c4f9f9..b923b5d 100644
--- a/drivers/infiniband/hw/ehca/ehca_irq.c
+++ b/drivers/infiniband/hw/ehca/ehca_irq.c
@@ -206,7 +206,7 @@ static void qp_event_callback(struct ehc
}
static void cq_event_callback(struct ehca_shca *shca,
- u64 eqe)
+ u64 eqe)
{
struct ehca_cq *cq;
unsigned long flags;
@@ -318,7 +318,7 @@ static void parse_ec(struct ehca_shca *s
disruptive port %x configuration change, port);
ehca_info(shca-ib_device,
-port %x is inactive., port);
+ port %x is inactive., port);
event.device = shca-ib_device;
event.event = IB_EVENT_PORT_ERR;
event.element.port_num = port;
@@ -326,7 +326,7 @@ static void parse_ec(struct ehca_shca *s
ib_dispatch_event(event);
ehca_info(shca-ib_device,
-port %x is active., port);
+ port %x is active., port);
event.device = shca-ib_device;
event.event = IB_EVENT_PORT_ACTIVE;
event.element.port_num = port;
@@ -401,87 +401,143 @@ irqreturn_t ehca_interrupt_eq(int irq, v
return IRQ_HANDLED;
}
-void ehca_tasklet_eq(unsigned long data)
-{
- struct ehca_shca *shca = (struct ehca_shca*)data;
- struct ehca_eqe *eqe;
- int int_state;
- int query_cnt = 0;
- do {
- eqe = (struct ehca_eqe *)ehca_poll_eq(shca, shca-eq);
+static inline void process_eqe(struct ehca_shca *shca, struct ehca_eqe *eqe)
+{
+ u64 eqe_value;
+ u32 token;
+ unsigned long flags;
+ struct ehca_cq *cq;
+ eqe_value = eqe-entry;
+ ehca_dbg(shca-ib_device, eqe_value=%lx, eqe_value);
+ if (EHCA_BMASK_GET(EQE_COMPLETION_EVENT, eqe_value)) {
+ ehca_dbg(shca-ib_device, ... completion event);
+ token = EHCA_BMASK_GET(EQE_CQ_TOKEN, eqe_value);
+ spin_lock_irqsave(ehca_cq_idr_lock, flags);
+ cq = idr_find(ehca_cq_idr, token);
+ if (cq == NULL) {
+ spin_unlock_irqrestore(ehca_cq_idr_lock, flags);
+ ehca_err(shca-ib_device,
+Invalid eqe for non-existing cq token=%x,
+token);
+ return;
+ }
+ reset_eq_pending(cq);
+#ifdef CONFIG_INFINIBAND_EHCA_SCALING
+ queue_comp_task(cq);
+ spin_unlock_irqrestore(ehca_cq_idr_lock, flags);
+#else
+ spin_unlock_irqrestore(ehca_cq_idr_lock, flags);
+ comp_event_callback(cq);
+#endif
+ } else {
+ ehca_dbg(shca-ib_device,
+Got non completion event
[openib-general] [PATCH 2.6.21-rc1 2/5] ehca: fix race condition/locking issues in scaling code
fix a race condition in find_next_cpu_online() and some
other locking issues in scaling code
Signed-off-by: Hoang-Nam Nguyen [EMAIL PROTECTED]
---
ehca_irq.c | 68 +
1 files changed, 33 insertions(+), 35 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_irq.c
b/drivers/infiniband/hw/ehca/ehca_irq.c
index b923b5d..9679b07 100644
--- a/drivers/infiniband/hw/ehca/ehca_irq.c
+++ b/drivers/infiniband/hw/ehca/ehca_irq.c
@@ -544,28 +544,30 @@ void ehca_tasklet_eq(unsigned long data)
static inline int find_next_online_cpu(struct ehca_comp_pool* pool)
{
- unsigned long flags_last_cpu;
+ int cpu;
+ unsigned long flags;
+ WARN_ON_ONCE(!in_interrupt());
if (ehca_debug_level)
ehca_dmp(cpu_online_map, sizeof(cpumask_t), );
- spin_lock_irqsave(pool-last_cpu_lock, flags_last_cpu);
- pool-last_cpu = next_cpu(pool-last_cpu, cpu_online_map);
- if (pool-last_cpu == NR_CPUS)
- pool-last_cpu = first_cpu(cpu_online_map);
- spin_unlock_irqrestore(pool-last_cpu_lock, flags_last_cpu);
+ spin_lock_irqsave(pool-last_cpu_lock, flags);
+ cpu = next_cpu(pool-last_cpu, cpu_online_map);
+ if (cpu == NR_CPUS)
+ cpu = first_cpu(cpu_online_map);
+ pool-last_cpu = cpu;
+ spin_unlock_irqrestore(pool-last_cpu_lock, flags);
- return pool-last_cpu;
+ return cpu;
}
static void __queue_comp_task(struct ehca_cq *__cq,
struct ehca_cpu_comp_task *cct)
{
- unsigned long flags_cct;
- unsigned long flags_cq;
+ unsigned long flags;
- spin_lock_irqsave(cct-task_lock, flags_cct);
- spin_lock_irqsave(__cq-task_lock, flags_cq);
+ spin_lock_irqsave(cct-task_lock, flags);
+ spin_lock(__cq-task_lock);
if (__cq-nr_callbacks == 0) {
__cq-nr_callbacks++;
@@ -576,8 +578,8 @@ static void __queue_comp_task(struct ehc
else
__cq-nr_callbacks++;
- spin_unlock_irqrestore(__cq-task_lock, flags_cq);
- spin_unlock_irqrestore(cct-task_lock, flags_cct);
+ spin_unlock(__cq-task_lock);
+ spin_unlock_irqrestore(cct-task_lock, flags);
}
static void queue_comp_task(struct ehca_cq *__cq)
@@ -588,69 +590,69 @@ static void queue_comp_task(struct ehca_
cpu = get_cpu();
cpu_id = find_next_online_cpu(pool);
-
BUG_ON(!cpu_online(cpu_id));
cct = per_cpu_ptr(pool-cpu_comp_tasks, cpu_id);
+ BUG_ON(!cct);
if (cct-cq_jobs 0) {
cpu_id = find_next_online_cpu(pool);
cct = per_cpu_ptr(pool-cpu_comp_tasks, cpu_id);
+ BUG_ON(!cct);
}
__queue_comp_task(__cq, cct);
-
- put_cpu();
-
- return;
}
static void run_comp_task(struct ehca_cpu_comp_task* cct)
{
struct ehca_cq *cq;
- unsigned long flags_cct;
- unsigned long flags_cq;
+ unsigned long flags;
- spin_lock_irqsave(cct-task_lock, flags_cct);
+ spin_lock_irqsave(cct-task_lock, flags);
while (!list_empty(cct-cq_list)) {
cq = list_entry(cct-cq_list.next, struct ehca_cq, entry);
- spin_unlock_irqrestore(cct-task_lock, flags_cct);
+ spin_unlock_irqrestore(cct-task_lock, flags);
comp_event_callback(cq);
- spin_lock_irqsave(cct-task_lock, flags_cct);
+ spin_lock_irqsave(cct-task_lock, flags);
- spin_lock_irqsave(cq-task_lock, flags_cq);
+ spin_lock(cq-task_lock);
cq-nr_callbacks--;
if (cq-nr_callbacks == 0) {
list_del_init(cct-cq_list.next);
cct-cq_jobs--;
}
- spin_unlock_irqrestore(cq-task_lock, flags_cq);
-
+ spin_unlock(cq-task_lock);
}
- spin_unlock_irqrestore(cct-task_lock, flags_cct);
-
- return;
+ spin_unlock_irqrestore(cct-task_lock, flags);
}
static int comp_task(void *__cct)
{
struct ehca_cpu_comp_task* cct = __cct;
+ int cql_empty;
DECLARE_WAITQUEUE(wait, current);
set_current_state(TASK_INTERRUPTIBLE);
while(!kthread_should_stop()) {
add_wait_queue(cct-wait_queue, wait);
- if (list_empty(cct-cq_list))
+ spin_lock_irq(cct-task_lock);
+ cql_empty = list_empty(cct-cq_list);
+ spin_unlock_irq(cct-task_lock);
+ if (cql_empty)
schedule();
else
__set_current_state(TASK_RUNNING);
remove_wait_queue(cct-wait_queue, wait);
- if (!list_empty(cct-cq_list))
+ spin_lock_irq(cct-task_lock);
+ cql_empty = list_empty(cct-cq_list);
+ spin_unlock_irq(cct-task_lock
[openib-general] [PATCH 2.6.21-rc1 3/5] ehca: allow en/disabling scaling code via module parameter
allow users to en/disable scaling code when loading ib_ehca module
Signed-off-by: Hoang-Nam Nguyen [EMAIL PROTECTED]
---
Kconfig|8
ehca_classes.h |1 +
ehca_irq.c | 47 +--
ehca_main.c|4
4 files changed, 26 insertions(+), 34 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/Kconfig
b/drivers/infiniband/hw/ehca/Kconfig
index 727b10d..1a85459 100644
--- a/drivers/infiniband/hw/ehca/Kconfig
+++ b/drivers/infiniband/hw/ehca/Kconfig
@@ -7,11 +7,3 @@ config INFINIBAND_EHCA
To compile the driver as a module, choose M here. The module
will be called ib_ehca.
-config INFINIBAND_EHCA_SCALING
- bool Scaling support (EXPERIMENTAL)
- depends on IBMEBUS INFINIBAND_EHCA HOTPLUG_CPU EXPERIMENTAL
- default y
- ---help---
- eHCA scaling support schedules the CQ callbacks to different CPUs.
-
- To enable this feature choose Y here.
diff --git a/drivers/infiniband/hw/ehca/ehca_classes.h
b/drivers/infiniband/hw/ehca/ehca_classes.h
index f08ad6f..40404c9 100644
--- a/drivers/infiniband/hw/ehca/ehca_classes.h
+++ b/drivers/infiniband/hw/ehca/ehca_classes.h
@@ -277,6 +277,7 @@ extern struct idr ehca_cq_idr;
extern int ehca_static_rate;
extern int ehca_port_act_time;
extern int ehca_use_hp_mr;
+extern int ehca_scaling_code;
struct ipzu_queue_resp {
u32 qe_size; /* queue entry size */
diff --git a/drivers/infiniband/hw/ehca/ehca_irq.c
b/drivers/infiniband/hw/ehca/ehca_irq.c
index 9679b07..3ec53c6 100644
--- a/drivers/infiniband/hw/ehca/ehca_irq.c
+++ b/drivers/infiniband/hw/ehca/ehca_irq.c
@@ -63,15 +63,11 @@
#define ERROR_DATA_LENGTH EHCA_BMASK_IBM(52,63)
#define ERROR_DATA_TYPEEHCA_BMASK_IBM(0,7)
-#ifdef CONFIG_INFINIBAND_EHCA_SCALING
-
static void queue_comp_task(struct ehca_cq *__cq);
static struct ehca_comp_pool* pool;
static struct notifier_block comp_pool_callback_nb;
-#endif
-
static inline void comp_event_callback(struct ehca_cq *cq)
{
if (!cq-ib_cq.comp_handler)
@@ -423,13 +419,13 @@ static inline void process_eqe(struct eh
return;
}
reset_eq_pending(cq);
-#ifdef CONFIG_INFINIBAND_EHCA_SCALING
- queue_comp_task(cq);
- spin_unlock_irqrestore(ehca_cq_idr_lock, flags);
-#else
- spin_unlock_irqrestore(ehca_cq_idr_lock, flags);
- comp_event_callback(cq);
-#endif
+ if (ehca_scaling_code) {
+ queue_comp_task(cq);
+ spin_unlock_irqrestore(ehca_cq_idr_lock, flags);
+ } else {
+ spin_unlock_irqrestore(ehca_cq_idr_lock, flags);
+ comp_event_callback(cq);
+ }
} else {
ehca_dbg(shca-ib_device,
Got non completion event);
@@ -508,13 +504,12 @@ void ehca_process_eq(struct ehca_shca *s
/* call completion handler for cached eqes */
for (i = 0; i eqe_cnt; i++)
if (eq-eqe_cache[i].cq) {
-#ifdef CONFIG_INFINIBAND_EHCA_SCALING
- spin_lock(ehca_cq_idr_lock);
- queue_comp_task(eq-eqe_cache[i].cq);
- spin_unlock(ehca_cq_idr_lock);
-#else
- comp_event_callback(eq-eqe_cache[i].cq);
-#endif
+ if (ehca_scaling_code) {
+ spin_lock(ehca_cq_idr_lock);
+ queue_comp_task(eq-eqe_cache[i].cq);
+ spin_unlock(ehca_cq_idr_lock);
+ } else
+ comp_event_callback(eq-eqe_cache[i].cq);
} else {
ehca_dbg(shca-ib_device, Got non completion event);
parse_identifier(shca, eq-eqe_cache[i].eqe-entry);
@@ -540,8 +535,6 @@ void ehca_tasklet_eq(unsigned long data)
ehca_process_eq((struct ehca_shca*)data, 1);
}
-#ifdef CONFIG_INFINIBAND_EHCA_SCALING
-
static inline int find_next_online_cpu(struct ehca_comp_pool* pool)
{
int cpu;
@@ -764,14 +757,14 @@ static int comp_pool_callback(struct not
return NOTIFY_OK;
}
-#endif
-
int ehca_create_comp_pool(void)
{
-#ifdef CONFIG_INFINIBAND_EHCA_SCALING
int cpu;
struct task_struct *task;
+ if (!ehca_scaling_code)
+ return 0;
+
pool = kzalloc(sizeof(struct ehca_comp_pool), GFP_KERNEL);
if (pool == NULL)
return -ENOMEM;
@@ -796,16 +789,19 @@ int ehca_create_comp_pool(void)
comp_pool_callback_nb.notifier_call = comp_pool_callback;
comp_pool_callback_nb.priority =0;
register_cpu_notifier(comp_pool_callback_nb);
-#endif
+
+ printk(KERN_INFO eHCA scaling code enabled\n);
return 0;
}
void ehca_destroy_comp_pool(void)
{
-#ifdef CONFIG_INFINIBAND_EHCA_SCALING
[openib-general] [PATCH 2.6.21-rc1 4/5] ehca: replace yield() by wait_for_completion()
remove yield() and use wait_for_completion() in order to wait for running
completion handlers finished before destroying associated completion queue
Signed-off-by: Hoang-Nam Nguyen [EMAIL PROTECTED]
---
ehca_classes.h |3 +++
ehca_cq.c |5 +++--
ehca_irq.c |6 +-
3 files changed, 11 insertions(+), 3 deletions(-)
diff --git a/drivers/infiniband/hw/ehca/ehca_classes.h
b/drivers/infiniband/hw/ehca/ehca_classes.h
index 40404c9..d8ce0c8 100644
--- a/drivers/infiniband/hw/ehca/ehca_classes.h
+++ b/drivers/infiniband/hw/ehca/ehca_classes.h
@@ -52,6 +52,8 @@ struct ehca_mw;
struct ehca_pd;
struct ehca_av;
+#include linux/completion.h
+
#include rdma/ib_verbs.h
#include rdma/ib_user_verbs.h
@@ -154,6 +156,7 @@ struct ehca_cq {
struct hlist_head qp_hashtab[QP_HASHTAB_LEN];
struct list_head entry;
u32 nr_callbacks;
+ struct completion zero_callbacks;
spinlock_t task_lock;
u32 ownpid;
/* mmap counter for resources mapped into user space */
diff --git a/drivers/infiniband/hw/ehca/ehca_cq.c
b/drivers/infiniband/hw/ehca/ehca_cq.c
index 9291a86..906bd5b 100644
--- a/drivers/infiniband/hw/ehca/ehca_cq.c
+++ b/drivers/infiniband/hw/ehca/ehca_cq.c
@@ -147,6 +147,7 @@ struct ib_cq *ehca_create_cq(struct ib_d
spin_lock_init(my_cq-spinlock);
spin_lock_init(my_cq-cb_lock);
spin_lock_init(my_cq-task_lock);
+ init_completion(my_cq-zero_callbacks);
my_cq-ownpid = current-tgid;
cq = my_cq-ib_cq;
@@ -330,9 +331,9 @@ int ehca_destroy_cq(struct ib_cq *cq)
}
spin_lock_irqsave(ehca_cq_idr_lock, flags);
- while (my_cq-nr_callbacks) {
+ if (my_cq-nr_callbacks) {
spin_unlock_irqrestore(ehca_cq_idr_lock, flags);
- yield();
+ wait_for_completion(my_cq-zero_callbacks);
spin_lock_irqsave(ehca_cq_idr_lock, flags);
}
diff --git a/drivers/infiniband/hw/ehca/ehca_irq.c
b/drivers/infiniband/hw/ehca/ehca_irq.c
index 3ec53c6..7db39b7 100644
--- a/drivers/infiniband/hw/ehca/ehca_irq.c
+++ b/drivers/infiniband/hw/ehca/ehca_irq.c
@@ -605,6 +605,7 @@ static void run_comp_task(struct ehca_cp
spin_lock_irqsave(cct-task_lock, flags);
while (!list_empty(cct-cq_list)) {
+ int is_complete = 0;
cq = list_entry(cct-cq_list.next, struct ehca_cq, entry);
spin_unlock_irqrestore(cct-task_lock, flags);
comp_event_callback(cq);
@@ -612,11 +613,14 @@ static void run_comp_task(struct ehca_cp
spin_lock(cq-task_lock);
cq-nr_callbacks--;
- if (cq-nr_callbacks == 0) {
+ is_complete = (cq-nr_callbacks == 0);
+ if (is_complete) {
list_del_init(cct-cq_list.next);
cct-cq_jobs--;
}
spin_unlock(cq-task_lock);
+ if (is_complete) /* wake up waiting destroy_cq() */
+ complete(cq-zero_callbacks);
}
spin_unlock_irqrestore(cct-task_lock, flags);
___
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/listinfo/openib-general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
