* [PATCH rdma-next] RDMA/core: Fix find_gid_index to use the proper API to retrieve the vlan ID
@ 2019-09-25 11:23 Yuval Basson
2019-09-25 17:17 ` Parav Pandit
0 siblings, 1 reply; 3+ messages in thread
From: Yuval Basson @ 2019-09-25 11:23 UTC (permalink / raw)
To: dledford, jgg; +Cc: leon, mkalderon, linux-rdma, Yuval Basson, Ariel Elior
When working over a macvlan device which was itself
created over a vlan device, the roce CM traffic should use
the vlan from the lower vlan device, but instead it simply queries
the macvlan device as to whether it is itself a vlan device.
Since it is not, the roce CM traffic is sent without any vlan, which
causes it not to be accepted by the peer which is running directly
over a vlan device, and will thus only accept roce CM traffic carrying
the vlan.
Fixes: dbf727de7440 ("Use GID table in AH creation and dmac resolution")
Signed-off-by: Ariel Elior <aelior@marvell.com>
Signed-off-by: Yuval Basson <ybason@marvell.com>
---
drivers/infiniband/core/verbs.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c
index f974b68..1d2d9be0 100644
--- a/drivers/infiniband/core/verbs.c
+++ b/drivers/infiniband/core/verbs.c
@@ -662,13 +662,13 @@ static bool find_gid_index(const union ib_gid *gid,
void *context)
{
struct find_gid_index_context *ctx = context;
+ u16 vlan_id;
if (ctx->gid_type != gid_attr->gid_type)
return false;
- if ((!!(ctx->vlan_id != 0xffff) == !is_vlan_dev(gid_attr->ndev)) ||
- (is_vlan_dev(gid_attr->ndev) &&
- vlan_dev_vlan_id(gid_attr->ndev) != ctx->vlan_id))
+ rdma_read_gid_l2_fields(gid_attr, &vlan_id, NULL);
+ if (ctx->vlan_id != vlan_id)
return false;
return true;
--
1.8.3.1
^ permalink raw reply related [flat|nested] 3+ messages in thread* RE: [PATCH rdma-next] RDMA/core: Fix find_gid_index to use the proper API to retrieve the vlan ID
2019-09-25 11:23 [PATCH rdma-next] RDMA/core: Fix find_gid_index to use the proper API to retrieve the vlan ID Yuval Basson
@ 2019-09-25 17:17 ` Parav Pandit
2019-09-26 8:00 ` Leon Romanovsky
0 siblings, 1 reply; 3+ messages in thread
From: Parav Pandit @ 2019-09-25 17:17 UTC (permalink / raw)
To: Yuval Basson, dledford@redhat.com, jgg@ziepe.ca
Cc: leon@kernel.org, mkalderon@marvell.com,
linux-rdma@vger.kernel.org, Ariel Elior
Hi Yuval,
> -----Original Message-----
> From: linux-rdma-owner@vger.kernel.org <linux-rdma-
> owner@vger.kernel.org> On Behalf Of Yuval Basson
> Sent: Wednesday, September 25, 2019 6:23 AM
> To: dledford@redhat.com; jgg@ziepe.ca
> Cc: leon@kernel.org; mkalderon@marvell.com; linux-rdma@vger.kernel.org;
> Yuval Basson <ybason@marvell.com>; Ariel Elior <aelior@marvell.com>
> Subject: [PATCH rdma-next] RDMA/core: Fix find_gid_index to use the proper
> API to retrieve the vlan ID
>
> When working over a macvlan device which was itself created over a vlan
> device, the roce CM traffic should use the vlan from the lower vlan device, but
> instead it simply queries the macvlan device as to whether it is itself a vlan
> device.
> Since it is not, the roce CM traffic is sent without any vlan, which causes it not
> to be accepted by the peer which is running directly over a vlan device, and will
> thus only accept roce CM traffic carrying the vlan.
>
> Fixes: dbf727de7440 ("Use GID table in AH creation and dmac resolution")
> Signed-off-by: Ariel Elior <aelior@marvell.com>
> Signed-off-by: Yuval Basson <ybason@marvell.com>
> ---
> drivers/infiniband/core/verbs.c | 6 +++---
> 1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c
> index f974b68..1d2d9be0 100644
> --- a/drivers/infiniband/core/verbs.c
> +++ b/drivers/infiniband/core/verbs.c
> @@ -662,13 +662,13 @@ static bool find_gid_index(const union ib_gid *gid,
> void *context)
> {
> struct find_gid_index_context *ctx = context;
> + u16 vlan_id;
>
> if (ctx->gid_type != gid_attr->gid_type)
> return false;
>
> - if ((!!(ctx->vlan_id != 0xffff) == !is_vlan_dev(gid_attr->ndev)) ||
> - (is_vlan_dev(gid_attr->ndev) &&
> - vlan_dev_vlan_id(gid_attr->ndev) != ctx->vlan_id))
> + rdma_read_gid_l2_fields(gid_attr, &vlan_id, NULL);
> + if (ctx->vlan_id != vlan_id)
> return false;
A little shorter version [1] along with rdmacm support [2] are in Leon's queue which is needed for macvlan to work properly.
[1] https://git.kernel.org/pub/scm/linux/kernel/git/leon/linux-rdma.git/commit/?h=rdma-next&id=9811bdac481cf6278512a8d052c5ba91cfb69c71
[2] https://git.kernel.org/pub/scm/linux/kernel/git/leon/linux-rdma.git/commit/?h=rdma-next&id=d190d04808be104a2a3ef71841ff03c5ffaa16f8
>
> return true;
> --
> 1.8.3.1
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: [PATCH rdma-next] RDMA/core: Fix find_gid_index to use the proper API to retrieve the vlan ID
2019-09-25 17:17 ` Parav Pandit
@ 2019-09-26 8:00 ` Leon Romanovsky
0 siblings, 0 replies; 3+ messages in thread
From: Leon Romanovsky @ 2019-09-26 8:00 UTC (permalink / raw)
To: Parav Pandit
Cc: Yuval Basson, dledford@redhat.com, jgg@ziepe.ca,
mkalderon@marvell.com, linux-rdma@vger.kernel.org, Ariel Elior
On Wed, Sep 25, 2019 at 05:17:49PM +0000, Parav Pandit wrote:
> Hi Yuval,
>
>
> > -----Original Message-----
> > From: linux-rdma-owner@vger.kernel.org <linux-rdma-
> > owner@vger.kernel.org> On Behalf Of Yuval Basson
> > Sent: Wednesday, September 25, 2019 6:23 AM
> > To: dledford@redhat.com; jgg@ziepe.ca
> > Cc: leon@kernel.org; mkalderon@marvell.com; linux-rdma@vger.kernel.org;
> > Yuval Basson <ybason@marvell.com>; Ariel Elior <aelior@marvell.com>
> > Subject: [PATCH rdma-next] RDMA/core: Fix find_gid_index to use the proper
> > API to retrieve the vlan ID
> >
> > When working over a macvlan device which was itself created over a vlan
> > device, the roce CM traffic should use the vlan from the lower vlan device, but
> > instead it simply queries the macvlan device as to whether it is itself a vlan
> > device.
> > Since it is not, the roce CM traffic is sent without any vlan, which causes it not
> > to be accepted by the peer which is running directly over a vlan device, and will
> > thus only accept roce CM traffic carrying the vlan.
> >
> > Fixes: dbf727de7440 ("Use GID table in AH creation and dmac resolution")
> > Signed-off-by: Ariel Elior <aelior@marvell.com>
> > Signed-off-by: Yuval Basson <ybason@marvell.com>
> > ---
> > drivers/infiniband/core/verbs.c | 6 +++---
> > 1 file changed, 3 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c
> > index f974b68..1d2d9be0 100644
> > --- a/drivers/infiniband/core/verbs.c
> > +++ b/drivers/infiniband/core/verbs.c
> > @@ -662,13 +662,13 @@ static bool find_gid_index(const union ib_gid *gid,
> > void *context)
> > {
> > struct find_gid_index_context *ctx = context;
> > + u16 vlan_id;
> >
> > if (ctx->gid_type != gid_attr->gid_type)
> > return false;
> >
> > - if ((!!(ctx->vlan_id != 0xffff) == !is_vlan_dev(gid_attr->ndev)) ||
> > - (is_vlan_dev(gid_attr->ndev) &&
> > - vlan_dev_vlan_id(gid_attr->ndev) != ctx->vlan_id))
> > + rdma_read_gid_l2_fields(gid_attr, &vlan_id, NULL);
> > + if (ctx->vlan_id != vlan_id)
> > return false;
>
> A little shorter version [1] along with rdmacm support [2] are in Leon's queue which is needed for macvlan to work properly.
I'll start sending after merge window will end.
Thanks
>
> [1] https://git.kernel.org/pub/scm/linux/kernel/git/leon/linux-rdma.git/commit/?h=rdma-next&id=9811bdac481cf6278512a8d052c5ba91cfb69c71
> [2] https://git.kernel.org/pub/scm/linux/kernel/git/leon/linux-rdma.git/commit/?h=rdma-next&id=d190d04808be104a2a3ef71841ff03c5ffaa16f8
>
> >
> > return true;
> > --
> > 1.8.3.1
>
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2019-09-26 8:00 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2019-09-25 11:23 [PATCH rdma-next] RDMA/core: Fix find_gid_index to use the proper API to retrieve the vlan ID Yuval Basson
2019-09-25 17:17 ` Parav Pandit
2019-09-26 8:00 ` Leon Romanovsky
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox