This v7 of patch series is used to read the port_attribute subnet_prefix
from a valid cache entry instead of having to call
device->ops.query_gid() for Infiniband link-layer devices in
__ib_query_port().
In the event of a cache update, the value for subnet_prefix gets read
using device->ops.query_gid() in config_non_roce_gid_cache().
Anand Khoje (3):
IB/core: Updating cache for subnet_prefix in
config_non_roce_gid_cache()
IB/core: Shifting initialization of device->cache_lock.
IB/core: Read subnet_prefix in ib_query_port via cache.
---
v1 -> v2:
- Split the v1 patch in 3 patches as per Leon's suggestion.
v2 -> v3:
- Added changes as per Mark Zhang's suggestion of clearing
flags in git_table_cleanup_one().
v3 -> v4:
- Removed the enum ib_port_data_flags and 8 byte flags from
struct ib_port_data, and the set_bit()/clear_bit() API
used to update this flag as that was not necessary.
Done to keep the code simple.
- Added code to read subnet_prefix from updated GID cache in the
event of cache update. Prior to this change, ib_cache_update
was reading the value for subnet_prefix via ib_query_port(),
due to this patch, we ended up reading a stale cached value of
subnet_prefix.
v4 -> v5:
- Removed the code to reset cache_is_initialised bit from cleanup
as per Leon's suggestion.
- Removed ib_cache_is_initialised() function.
v5 -> v6:
- Added changes as per Jason's suggestion of updating subnet_prefix
in config_non_roce_gid_cache() and removing the flag
cache_is_initialized in __ib_query_port().
v6 -> v7:
- Reordering the initialization of cache_lock, as the previous
version caused an access to uninitialized cache_lock.
---
drivers/infiniband/core/cache.c | 10 +++++-----
drivers/infiniband/core/device.c | 10 ++++------
2 files changed, 9 insertions(+), 11 deletions(-)
--
1.8.3.1
Currently cache for subnet_prefix was getting updated by reading port
attributes via ib_query_port. ib_query_port() calls ops.query_gid()
to get subnet_prefix and returns it via port_attr.
In ib_cache_update(), config_non_roce_gid_cache() obtains GIDs
by calling ops.query_gid(). We utilize this to store subnet_prefix
in cache.
Suggested-by: Jason Gunthorpe <[email protected]>
Suggested-by: Aru Kolappan <[email protected]>
Signed-off-by: Anand Khoje <[email protected]>
Signed-off-by: Haakon Bugge <[email protected]>
---
drivers/infiniband/core/cache.c | 8 +++++---
1 file changed, 5 insertions(+), 3 deletions(-)
diff --git a/drivers/infiniband/core/cache.c b/drivers/infiniband/core/cache.c
index c9e9fc8..929399e 100644
--- a/drivers/infiniband/core/cache.c
+++ b/drivers/infiniband/core/cache.c
@@ -1429,7 +1429,7 @@ int rdma_read_gid_l2_fields(const struct ib_gid_attr *attr,
EXPORT_SYMBOL(rdma_read_gid_l2_fields);
static int config_non_roce_gid_cache(struct ib_device *device,
- u32 port, int gid_tbl_len)
+ u32 port, struct ib_port_attr *tprops)
{
struct ib_gid_attr gid_attr = {};
struct ib_gid_table *table;
@@ -1441,7 +1441,7 @@ static int config_non_roce_gid_cache(struct ib_device *device,
table = rdma_gid_table(device, port);
mutex_lock(&table->lock);
- for (i = 0; i < gid_tbl_len; ++i) {
+ for (i = 0; i < tprops->gid_tbl_len; ++i) {
if (!device->ops.query_gid)
continue;
ret = device->ops.query_gid(device, port, i, &gid_attr.gid);
@@ -1452,6 +1452,8 @@ static int config_non_roce_gid_cache(struct ib_device *device,
goto err;
}
gid_attr.index = i;
+ tprops->subnet_prefix =
+ be64_to_cpu(gid_attr.gid.global.subnet_prefix);
add_modify_gid(table, &gid_attr);
}
err:
@@ -1484,7 +1486,7 @@ static int config_non_roce_gid_cache(struct ib_device *device,
if (!rdma_protocol_roce(device, port) && update_gids) {
ret = config_non_roce_gid_cache(device, port,
- tprops->gid_tbl_len);
+ tprops);
if (ret)
goto err;
}
--
1.8.3.1
The lock cache_lock of struct ib_device is initialized in function
ib_cache_setup_one(). This is much later than the device initialization
in _ib_alloc_device().
This change shifts initialization of cache_lock in _ib_alloc_device().
Suggested-by: Haakon Bugge <[email protected]>
Signed-off-by: Anand Khoje <[email protected]>
---
drivers/infiniband/core/cache.c | 2 --
drivers/infiniband/core/device.c | 2 ++
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/infiniband/core/cache.c b/drivers/infiniband/core/cache.c
index 929399e..0c98dd3 100644
--- a/drivers/infiniband/core/cache.c
+++ b/drivers/infiniband/core/cache.c
@@ -1621,8 +1621,6 @@ int ib_cache_setup_one(struct ib_device *device)
u32 p;
int err;
- rwlock_init(&device->cache_lock);
-
err = gid_table_setup_one(device);
if (err)
return err;
diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c
index fa20b18..ba0ad72 100644
--- a/drivers/infiniband/core/device.c
+++ b/drivers/infiniband/core/device.c
@@ -607,6 +607,8 @@ struct ib_device *_ib_alloc_device(size_t size)
for (i = 0; i < ARRAY_SIZE(device->cq_pools); i++)
INIT_LIST_HEAD(&device->cq_pools[i]);
+ rwlock_init(&device->cache_lock);
+
device->uverbs_cmd_mask =
BIT_ULL(IB_USER_VERBS_CMD_ALLOC_MW) |
BIT_ULL(IB_USER_VERBS_CMD_ALLOC_PD) |
--
1.8.3.1
ib_query_port() calls device->ops.query_port() to get the port
attributes. The method of querying is device driver specific.
The same function calls device->ops.query_gid() to get the GID and
extract the subnet_prefix (gid_prefix).
The GID and subnet_prefix are stored in a cache. But they do not get
read from the cache if the device is an Infiniband device. The
following change takes advantage of the cached subnet_prefix.
Testing with RDBMS has shown a significant improvement in performance
with this change.
Fixes: fad61ad ("IB/core: Add subnet prefix to port info")
Signed-off-by: Anand Khoje <[email protected]>
Signed-off-by: Haakon Bugge <[email protected]>
---
drivers/infiniband/core/device.c | 8 ++------
1 file changed, 2 insertions(+), 6 deletions(-)
diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c
index ba0ad72..9056f48 100644
--- a/drivers/infiniband/core/device.c
+++ b/drivers/infiniband/core/device.c
@@ -2052,7 +2052,6 @@ static int __ib_query_port(struct ib_device *device,
u32 port_num,
struct ib_port_attr *port_attr)
{
- union ib_gid gid = {};
int err;
memset(port_attr, 0, sizeof(*port_attr));
@@ -2065,11 +2064,8 @@ static int __ib_query_port(struct ib_device *device,
IB_LINK_LAYER_INFINIBAND)
return 0;
- err = device->ops.query_gid(device, port_num, 0, &gid);
- if (err)
- return err;
-
- port_attr->subnet_prefix = be64_to_cpu(gid.global.subnet_prefix);
+ ib_get_cached_subnet_prefix(device, port_num,
+ &port_attr->subnet_prefix);
return 0;
}
--
1.8.3.1
On 6/30/2021 3:16 PM, Anand Khoje wrote:
> This v7 of patch series is used to read the port_attribute subnet_prefix
> from a valid cache entry instead of having to call
> device->ops.query_gid() for Infiniband link-layer devices in
> __ib_query_port().
>
> In the event of a cache update, the value for subnet_prefix gets read
> using device->ops.query_gid() in config_non_roce_gid_cache().
>
> Anand Khoje (3):
> IB/core: Updating cache for subnet_prefix in
> config_non_roce_gid_cache()
> IB/core: Shifting initialization of device->cache_lock.
> IB/core: Read subnet_prefix in ib_query_port via cache.
> ---
> v1 -> v2:
> - Split the v1 patch in 3 patches as per Leon's suggestion.
>
> v2 -> v3:
> - Added changes as per Mark Zhang's suggestion of clearing
> flags in git_table_cleanup_one().
> v3 -> v4:
> - Removed the enum ib_port_data_flags and 8 byte flags from
> struct ib_port_data, and the set_bit()/clear_bit() API
> used to update this flag as that was not necessary.
> Done to keep the code simple.
> - Added code to read subnet_prefix from updated GID cache in the
> event of cache update. Prior to this change, ib_cache_update
> was reading the value for subnet_prefix via ib_query_port(),
> due to this patch, we ended up reading a stale cached value of
> subnet_prefix.
> v4 -> v5:
> - Removed the code to reset cache_is_initialised bit from cleanup
> as per Leon's suggestion.
> - Removed ib_cache_is_initialised() function.
>
> v5 -> v6:
> - Added changes as per Jason's suggestion of updating subnet_prefix
> in config_non_roce_gid_cache() and removing the flag
> cache_is_initialized in __ib_query_port().
>
> v6 -> v7:
> - Reordering the initialization of cache_lock, as the previous
> version caused an access to uninitialized cache_lock.
> ---
>
> drivers/infiniband/core/cache.c | 10 +++++-----
> drivers/infiniband/core/device.c | 10 ++++------
> 2 files changed, 9 insertions(+), 11 deletions(-)
>
Hi,
This is just a reminder note requesting review for this patch-set.
Thanks,
Anand
On Tue, Jul 06, 2021 at 12:55:48PM +0530, Anand Khoje wrote:
> On 6/30/2021 3:16 PM, Anand Khoje wrote:
> > This v7 of patch series is used to read the port_attribute subnet_prefix
> > from a valid cache entry instead of having to call
> > device->ops.query_gid() for Infiniband link-layer devices in
> > __ib_query_port().
> >
> > In the event of a cache update, the value for subnet_prefix gets read
> > using device->ops.query_gid() in config_non_roce_gid_cache().
> >
> > Anand Khoje (3):
> > IB/core: Updating cache for subnet_prefix in
> > config_non_roce_gid_cache()
> > IB/core: Shifting initialization of device->cache_lock.
> > IB/core: Read subnet_prefix in ib_query_port via cache.
> > v1 -> v2:
> > - Split the v1 patch in 3 patches as per Leon's suggestion.
> >
> > v2 -> v3:
> > - Added changes as per Mark Zhang's suggestion of clearing
> > flags in git_table_cleanup_one().
> > v3 -> v4:
> > - Removed the enum ib_port_data_flags and 8 byte flags from
> > struct ib_port_data, and the set_bit()/clear_bit() API
> > used to update this flag as that was not necessary.
> > Done to keep the code simple.
> > - Added code to read subnet_prefix from updated GID cache in the
> > event of cache update. Prior to this change, ib_cache_update
> > was reading the value for subnet_prefix via ib_query_port(),
> > due to this patch, we ended up reading a stale cached value of
> > subnet_prefix.
> > v4 -> v5:
> > - Removed the code to reset cache_is_initialised bit from cleanup
> > as per Leon's suggestion.
> > - Removed ib_cache_is_initialised() function.
> >
> > v5 -> v6:
> > - Added changes as per Jason's suggestion of updating subnet_prefix
> > in config_non_roce_gid_cache() and removing the flag
> > cache_is_initialized in __ib_query_port().
> >
> > v6 -> v7:
> > - Reordering the initialization of cache_lock, as the previous
> > version caused an access to uninitialized cache_lock.
> >
> > drivers/infiniband/core/cache.c | 10 +++++-----
> > drivers/infiniband/core/device.c | 10 ++++------
> > 2 files changed, 9 insertions(+), 11 deletions(-)
> >
>
> Hi,
>
> This is just a reminder note requesting review for this patch-set.
You'll probably have to resend it after the merge window closed on
Monday, rebased on the new rc1
Jason
On 7/9/2021 6:51 PM, Jason Gunthorpe wrote:
> On Tue, Jul 06, 2021 at 12:55:48PM +0530, Anand Khoje wrote:
>> On 6/30/2021 3:16 PM, Anand Khoje wrote:
>>> This v7 of patch series is used to read the port_attribute subnet_prefix
>>> from a valid cache entry instead of having to call
>>> device->ops.query_gid() for Infiniband link-layer devices in
>>> __ib_query_port().
>>>
>>> In the event of a cache update, the value for subnet_prefix gets read
>>> using device->ops.query_gid() in config_non_roce_gid_cache().
>>>
>>> Anand Khoje (3):
>>> IB/core: Updating cache for subnet_prefix in
>>> config_non_roce_gid_cache()
>>> IB/core: Shifting initialization of device->cache_lock.
>>> IB/core: Read subnet_prefix in ib_query_port via cache.
>>> v1 -> v2:
>>> - Split the v1 patch in 3 patches as per Leon's suggestion.
>>>
>>> v2 -> v3:
>>> - Added changes as per Mark Zhang's suggestion of clearing
>>> flags in git_table_cleanup_one().
>>> v3 -> v4:
>>> - Removed the enum ib_port_data_flags and 8 byte flags from
>>> struct ib_port_data, and the set_bit()/clear_bit() API
>>> used to update this flag as that was not necessary.
>>> Done to keep the code simple.
>>> - Added code to read subnet_prefix from updated GID cache in the
>>> event of cache update. Prior to this change, ib_cache_update
>>> was reading the value for subnet_prefix via ib_query_port(),
>>> due to this patch, we ended up reading a stale cached value of
>>> subnet_prefix.
>>> v4 -> v5:
>>> - Removed the code to reset cache_is_initialised bit from cleanup
>>> as per Leon's suggestion.
>>> - Removed ib_cache_is_initialised() function.
>>>
>>> v5 -> v6:
>>> - Added changes as per Jason's suggestion of updating subnet_prefix
>>> in config_non_roce_gid_cache() and removing the flag
>>> cache_is_initialized in __ib_query_port().
>>>
>>> v6 -> v7:
>>> - Reordering the initialization of cache_lock, as the previous
>>> version caused an access to uninitialized cache_lock.
>>>
>>> drivers/infiniband/core/cache.c | 10 +++++-----
>>> drivers/infiniband/core/device.c | 10 ++++------
>>> 2 files changed, 9 insertions(+), 11 deletions(-)
>>>
>>
>> Hi,
>>
>> This is just a reminder note requesting review for this patch-set.
>
> You'll probably have to resend it after the merge window closed on
> Monday, rebased on the new rc1
>
> Jason
>
Hi Jason,
Thanks for the response. Ok, I will rebase and resend this patchset.
Thx, Anand