2018-04-25 22:37:41

by Greg Thelen

[permalink] [raw]
Subject: [PATCH 5/5] IB: make INFINIBAND_ADDR_TRANS configurable

Allow INFINIBAND without INFINIBAND_ADDR_TRANS because fuzzing has been
finding fair number of CM bugs. So provide option to disable it.

Signed-off-by: Greg Thelen <[email protected]>
Cc: Tarick Bedeir <[email protected]>
---
drivers/infiniband/Kconfig | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/drivers/infiniband/Kconfig b/drivers/infiniband/Kconfig
index ee270e065ba9..2a972ed6851b 100644
--- a/drivers/infiniband/Kconfig
+++ b/drivers/infiniband/Kconfig
@@ -61,9 +61,12 @@ config INFINIBAND_ON_DEMAND_PAGING
pages on demand instead.

config INFINIBAND_ADDR_TRANS
- bool
+ bool "RDMA/CM"
depends on INFINIBAND
default y
+ ---help---
+ Support for RDMA communication manager (CM).
+ This allows for a generic connection abstraction over RDMA.

config INFINIBAND_ADDR_TRANS_CONFIGFS
bool
--
2.17.0.441.gb46fe60e1d-goog



2018-04-26 02:14:48

by Bart Van Assche

[permalink] [raw]
Subject: Re: [PATCH 5/5] IB: make INFINIBAND_ADDR_TRANS configurable

On Wed, 2018-04-25 at 15:34 -0700, Greg Thelen wrote:
> Allow INFINIBAND without INFINIBAND_ADDR_TRANS because fuzzing has been
> finding fair number of CM bugs. So provide option to disable it.
>
> Signed-off-by: Greg Thelen <[email protected]>
> Cc: Tarick Bedeir <[email protected]>
> ---
> drivers/infiniband/Kconfig | 5 ++++-
> 1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/infiniband/Kconfig b/drivers/infiniband/Kconfig
> index ee270e065ba9..2a972ed6851b 100644
> --- a/drivers/infiniband/Kconfig
> +++ b/drivers/infiniband/Kconfig
> @@ -61,9 +61,12 @@ config INFINIBAND_ON_DEMAND_PAGING
> pages on demand instead.
>
> config INFINIBAND_ADDR_TRANS
> - bool
> + bool "RDMA/CM"
> depends on INFINIBAND
> default y
> + ---help---
> + Support for RDMA communication manager (CM).
> + This allows for a generic connection abstraction over RDMA.

Hello Greg,

Please provide a cover letter when posting a patch series. Such a cover
letter is not only informative but also makes it easy for people who want
to comment on a patch series as a whole. I have a question that applies
to the entire patch series. The RDMA/CM code defines functions like
rdma_create_id() and rdma_create_qp(). If I search through the kernel tree
for callers of these functions then I find several more kernel modules than
the ones that are modified by this patch series:

$ git grep -lE '[[:blank:]](rdma_create_id|rdma_create_qp)\('
drivers/infiniband/core/cma.c
drivers/infiniband/ulp/iser/iser_verbs.c
drivers/infiniband/ulp/isert/ib_isert.c
drivers/infiniband/ulp/srp/ib_srp.c
drivers/infiniband/ulp/srpt/ib_srpt.c
drivers/nvme/host/rdma.c
drivers/nvme/target/rdma.c
drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd.c
drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd.h
fs/cifs/smbdirect.c
include/rdma/rdma_cm.h
net/9p/trans_rdma.c
net/rds/ib.c
net/rds/ib_cm.c
net/rds/rdma_transport.c
net/sunrpc/xprtrdma/svc_rdma_transport.c
net/sunrpc/xprtrdma/verbs.c

Are you sure that this patch series is complete?

Thanks,

Bart.


2018-04-26 02:26:37

by Greg Thelen

[permalink] [raw]
Subject: Re: [PATCH 5/5] IB: make INFINIBAND_ADDR_TRANS configurable

On Wed, Apr 25, 2018 at 7:13 PM Bart Van Assche <[email protected]>
wrote:

> On Wed, 2018-04-25 at 15:34 -0700, Greg Thelen wrote:
> > Allow INFINIBAND without INFINIBAND_ADDR_TRANS because fuzzing has been
> > finding fair number of CM bugs. So provide option to disable it.
> >
> > Signed-off-by: Greg Thelen <[email protected]>
> > Cc: Tarick Bedeir <[email protected]>
> > ---
> > drivers/infiniband/Kconfig | 5 ++++-
> > 1 file changed, 4 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/infiniband/Kconfig b/drivers/infiniband/Kconfig
> > index ee270e065ba9..2a972ed6851b 100644
> > --- a/drivers/infiniband/Kconfig
> > +++ b/drivers/infiniband/Kconfig
> > @@ -61,9 +61,12 @@ config INFINIBAND_ON_DEMAND_PAGING
> > pages on demand instead.
> >
> > config INFINIBAND_ADDR_TRANS
> > - bool
> > + bool "RDMA/CM"
> > depends on INFINIBAND
> > default y
> > + ---help---
> > + Support for RDMA communication manager (CM).
> > + This allows for a generic connection abstraction over RDMA.

> Hello Greg,

> Please provide a cover letter when posting a patch series. Such a cover
> letter is not only informative but also makes it easy for people who want
> to comment on a patch series as a whole. I have a question that applies
> to the entire patch series. The RDMA/CM code defines functions like
> rdma_create_id() and rdma_create_qp(). If I search through the kernel tree
> for callers of these functions then I find several more kernel modules
than
> the ones that are modified by this patch series:

> $ git grep -lE '[[:blank:]](rdma_create_id|rdma_create_qp)\('
> drivers/infiniband/core/cma.c
> drivers/infiniband/ulp/iser/iser_verbs.c
> drivers/infiniband/ulp/isert/ib_isert.c
> drivers/infiniband/ulp/srp/ib_srp.c
> drivers/infiniband/ulp/srpt/ib_srpt.c
> drivers/nvme/host/rdma.c
> drivers/nvme/target/rdma.c
> drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd.c
> drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd.h
> fs/cifs/smbdirect.c
> include/rdma/rdma_cm.h
> net/9p/trans_rdma.c
> net/rds/ib.c
> net/rds/ib_cm.c
> net/rds/rdma_transport.c
> net/sunrpc/xprtrdma/svc_rdma_transport.c
> net/sunrpc/xprtrdma/verbs.c

> Are you sure that this patch series is complete?

I'll check your cited files. I'll resend with cover letter.

FYI: I already rand this series through the 0-day builder, which presumably
would've caught most config dep issues. But I'll research your list before
reposting.