* [RFC net-next 1/6] net: xdp: avoid output parameters when querying XDP prog
2017-11-24 2:36 [RFC net-next 0/6] xdp: make stack perform remove and tests Jakub Kicinski
@ 2017-11-24 2:36 ` Jakub Kicinski
2017-11-24 2:36 ` [RFC net-next 2/6] net: xdp: report flags program was installed with on query Jakub Kicinski
` (6 subsequent siblings)
7 siblings, 0 replies; 22+ messages in thread
From: Jakub Kicinski @ 2017-11-24 2:36 UTC (permalink / raw)
To: netdev; +Cc: oss-drivers, daniel, alexei.starovoitov, jiri, Jakub Kicinski
The output parameters will get unwieldy if we want to add more
information about the program. Simply pass the entire
struct netdev_bpf in.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
---
include/linux/netdevice.h | 3 ++-
net/core/dev.c | 24 ++++++++++++++----------
net/core/rtnetlink.c | 6 +++++-
3 files changed, 21 insertions(+), 12 deletions(-)
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index ef789e1d679e..667bdd3ad33e 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -3330,7 +3330,8 @@ struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
int fd, u32 flags);
-u8 __dev_xdp_attached(struct net_device *dev, bpf_op_t xdp_op, u32 *prog_id);
+void __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
+ struct netdev_bpf *xdp);
int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
diff --git a/net/core/dev.c b/net/core/dev.c
index 07ed21d64f92..3f271c9cb5e0 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -7073,17 +7073,21 @@ int dev_change_proto_down(struct net_device *dev, bool proto_down)
}
EXPORT_SYMBOL(dev_change_proto_down);
-u8 __dev_xdp_attached(struct net_device *dev, bpf_op_t bpf_op, u32 *prog_id)
+void __dev_xdp_query(struct net_device *dev, bpf_op_t bpf_op,
+ struct netdev_bpf *xdp)
{
- struct netdev_bpf xdp;
-
- memset(&xdp, 0, sizeof(xdp));
- xdp.command = XDP_QUERY_PROG;
+ memset(xdp, 0, sizeof(*xdp));
+ xdp->command = XDP_QUERY_PROG;
/* Query must always succeed. */
- WARN_ON(bpf_op(dev, &xdp) < 0);
- if (prog_id)
- *prog_id = xdp.prog_id;
+ WARN_ON(bpf_op(dev, xdp) < 0);
+}
+
+static u8 __dev_xdp_attached(struct net_device *dev, bpf_op_t bpf_op)
+{
+ struct netdev_bpf xdp;
+
+ __dev_xdp_query(dev, bpf_op, &xdp);
return xdp.prog_attached;
}
@@ -7134,10 +7138,10 @@ int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
bpf_chk = generic_xdp_install;
if (fd >= 0) {
- if (bpf_chk && __dev_xdp_attached(dev, bpf_chk, NULL))
+ if (bpf_chk && __dev_xdp_attached(dev, bpf_chk))
return -EEXIST;
if ((flags & XDP_FLAGS_UPDATE_IF_NOEXIST) &&
- __dev_xdp_attached(dev, bpf_op, NULL))
+ __dev_xdp_attached(dev, bpf_op))
return -EBUSY;
prog = bpf_prog_get_type_dev(fd, BPF_PROG_TYPE_XDP,
diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
index dabba2a91fc8..9c4cb584bfb0 100644
--- a/net/core/rtnetlink.c
+++ b/net/core/rtnetlink.c
@@ -1261,6 +1261,7 @@ static u8 rtnl_xdp_attached_mode(struct net_device *dev, u32 *prog_id)
{
const struct net_device_ops *ops = dev->netdev_ops;
const struct bpf_prog *generic_xdp_prog;
+ struct netdev_bpf xdp;
ASSERT_RTNL();
@@ -1273,7 +1274,10 @@ static u8 rtnl_xdp_attached_mode(struct net_device *dev, u32 *prog_id)
if (!ops->ndo_bpf)
return XDP_ATTACHED_NONE;
- return __dev_xdp_attached(dev, ops->ndo_bpf, prog_id);
+ __dev_xdp_query(dev, ops->ndo_bpf, &xdp);
+ *prog_id = xdp.prog_id;
+
+ return xdp.prog_attached;
}
static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
--
2.14.1
^ permalink raw reply related [flat|nested] 22+ messages in thread* [RFC net-next 2/6] net: xdp: report flags program was installed with on query
2017-11-24 2:36 [RFC net-next 0/6] xdp: make stack perform remove and tests Jakub Kicinski
2017-11-24 2:36 ` [RFC net-next 1/6] net: xdp: avoid output parameters when querying XDP prog Jakub Kicinski
@ 2017-11-24 2:36 ` Jakub Kicinski
2017-11-24 2:36 ` [RFC net-next 3/6] net: xdp: make the stack take care of the tear down Jakub Kicinski
` (5 subsequent siblings)
7 siblings, 0 replies; 22+ messages in thread
From: Jakub Kicinski @ 2017-11-24 2:36 UTC (permalink / raw)
To: netdev; +Cc: oss-drivers, daniel, alexei.starovoitov, jiri, Jakub Kicinski
Some drivers enforce that flags on program replacement and
removal must match the flags passed on install. This leaves
the possibility open to enable simultaneous loading
of XDP programs both to HW and DRV.
Allow such drivers to report the flags back to the stack.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
---
drivers/net/ethernet/netronome/nfp/nfp_net_common.c | 1 +
include/linux/netdevice.h | 2 ++
2 files changed, 3 insertions(+)
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c
index 1a603fdd9e80..ea6bbf1efefc 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c
+++ b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c
@@ -3392,6 +3392,7 @@ static int nfp_net_xdp(struct net_device *netdev, struct netdev_bpf *xdp)
if (nn->dp.bpf_offload_xdp)
xdp->prog_attached = XDP_ATTACHED_HW;
xdp->prog_id = nn->xdp_prog ? nn->xdp_prog->aux->id : 0;
+ xdp->flags = nn->xdp_prog ? nn->xdp_flags : 0;
return 0;
case BPF_OFFLOAD_VERIFIER_PREP:
return nfp_app_bpf_verifier_prep(nn->app, nn, xdp);
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index 667bdd3ad33e..cc4ce7456e38 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -820,6 +820,8 @@ struct netdev_bpf {
struct {
u8 prog_attached;
u32 prog_id;
+ /* flags with which program was installed */
+ u32 prog_flags;
};
/* BPF_OFFLOAD_VERIFIER_PREP */
struct {
--
2.14.1
^ permalink raw reply related [flat|nested] 22+ messages in thread* [RFC net-next 3/6] net: xdp: make the stack take care of the tear down
2017-11-24 2:36 [RFC net-next 0/6] xdp: make stack perform remove and tests Jakub Kicinski
2017-11-24 2:36 ` [RFC net-next 1/6] net: xdp: avoid output parameters when querying XDP prog Jakub Kicinski
2017-11-24 2:36 ` [RFC net-next 2/6] net: xdp: report flags program was installed with on query Jakub Kicinski
@ 2017-11-24 2:36 ` Jakub Kicinski
2017-11-24 23:24 ` Daniel Borkmann
2017-11-24 2:36 ` [RFC net-next 4/6] netdevsim: add software driver for testing offloads Jakub Kicinski
` (4 subsequent siblings)
7 siblings, 1 reply; 22+ messages in thread
From: Jakub Kicinski @ 2017-11-24 2:36 UTC (permalink / raw)
To: netdev
Cc: oss-drivers, daniel, alexei.starovoitov, jiri, Jakub Kicinski,
Saeed Mahameed, Michael Chan, Ariel Elior, John Fastabend
Since day one of XDP drivers had to remember to free the program
on the remove path. This leads to code duplication and is error
prone. Make the stack query the installed programs on unregister
and if something is installed, remove the program.
Because the remove will now be called before notifiers are
invoked, BPF offload state of the program will not get destroyed
before uninstall.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
---
CC: Saeed Mahameed <saeedm@mellanox.com>
CC: Michael Chan <michael.chan@broadcom.com>
CC: Ariel Elior <Ariel.Elior@cavium.com>
CC: John Fastabend <john.fastabend@gmail.com>
drivers/net/ethernet/broadcom/bnxt/bnxt.c | 2 --
drivers/net/ethernet/mellanox/mlx5/core/en_main.c | 3 ---
drivers/net/ethernet/netronome/nfp/bpf/main.c | 7 -------
drivers/net/ethernet/netronome/nfp/nfp_net_common.c | 3 ---
drivers/net/ethernet/qlogic/qede/qede_main.c | 4 ----
drivers/net/tun.c | 4 ----
net/core/dev.c | 18 ++++++++++++++++++
7 files changed, 18 insertions(+), 23 deletions(-)
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 33c49ad697e4..413ad2444ba2 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -7800,8 +7800,6 @@ static void bnxt_remove_one(struct pci_dev *pdev)
bnxt_dcb_free(bp);
kfree(bp->edev);
bp->edev = NULL;
- if (bp->xdp_prog)
- bpf_prog_put(bp->xdp_prog);
bnxt_cleanup_pci(bp);
free_netdev(dev);
}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
index d2b057a3e512..0f5c012de52e 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
@@ -4308,9 +4308,6 @@ static void mlx5e_nic_cleanup(struct mlx5e_priv *priv)
{
mlx5e_ipsec_cleanup(priv);
mlx5e_vxlan_cleanup(priv);
-
- if (priv->channels.params.xdp_prog)
- bpf_prog_put(priv->channels.params.xdp_prog);
}
static int mlx5e_init_nic_rx(struct mlx5e_priv *priv)
diff --git a/drivers/net/ethernet/netronome/nfp/bpf/main.c b/drivers/net/ethernet/netronome/nfp/bpf/main.c
index e379b78e86ef..54bfd7846f6d 100644
--- a/drivers/net/ethernet/netronome/nfp/bpf/main.c
+++ b/drivers/net/ethernet/netronome/nfp/bpf/main.c
@@ -82,12 +82,6 @@ static const char *nfp_bpf_extra_cap(struct nfp_app *app, struct nfp_net *nn)
return nfp_net_ebpf_capable(nn) ? "BPF" : "";
}
-static void nfp_bpf_vnic_free(struct nfp_app *app, struct nfp_net *nn)
-{
- if (nn->dp.bpf_offload_xdp)
- nfp_bpf_xdp_offload(app, nn, NULL);
-}
-
static int nfp_bpf_setup_tc_block_cb(enum tc_setup_type type,
void *type_data, void *cb_priv)
{
@@ -168,7 +162,6 @@ const struct nfp_app_type app_bpf = {
.extra_cap = nfp_bpf_extra_cap,
.vnic_alloc = nfp_app_nic_vnic_alloc,
- .vnic_free = nfp_bpf_vnic_free,
.setup_tc = nfp_bpf_setup_tc,
.tc_busy = nfp_bpf_tc_busy,
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c
index ea6bbf1efefc..ad3e9f6a61e5 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c
+++ b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c
@@ -3562,9 +3562,6 @@ struct nfp_net *nfp_net_alloc(struct pci_dev *pdev, bool needs_netdev,
*/
void nfp_net_free(struct nfp_net *nn)
{
- if (nn->xdp_prog)
- bpf_prog_put(nn->xdp_prog);
-
if (nn->dp.netdev)
free_netdev(nn->dp.netdev);
else
diff --git a/drivers/net/ethernet/qlogic/qede/qede_main.c b/drivers/net/ethernet/qlogic/qede/qede_main.c
index 8f9b3eb82137..57332b3e5e64 100644
--- a/drivers/net/ethernet/qlogic/qede/qede_main.c
+++ b/drivers/net/ethernet/qlogic/qede/qede_main.c
@@ -1068,10 +1068,6 @@ static void __qede_remove(struct pci_dev *pdev, enum qede_remove_mode mode)
pci_set_drvdata(pdev, NULL);
- /* Release edev's reference to XDP's bpf if such exist */
- if (edev->xdp_prog)
- bpf_prog_put(edev->xdp_prog);
-
/* Use global ops since we've freed edev */
qed_ops->common->slowpath_stop(cdev);
if (system_state == SYSTEM_POWER_OFF)
diff --git a/drivers/net/tun.c b/drivers/net/tun.c
index 6a7bde9bc4b2..6f7e8e45c961 100644
--- a/drivers/net/tun.c
+++ b/drivers/net/tun.c
@@ -673,7 +673,6 @@ static void tun_detach(struct tun_file *tfile, bool clean)
static void tun_detach_all(struct net_device *dev)
{
struct tun_struct *tun = netdev_priv(dev);
- struct bpf_prog *xdp_prog = rtnl_dereference(tun->xdp_prog);
struct tun_file *tfile, *tmp;
int i, n = tun->numqueues;
@@ -708,9 +707,6 @@ static void tun_detach_all(struct net_device *dev)
}
BUG_ON(tun->numdisabled != 0);
- if (xdp_prog)
- bpf_prog_put(xdp_prog);
-
if (tun->flags & IFF_PERSIST)
module_put(THIS_MODULE);
}
diff --git a/net/core/dev.c b/net/core/dev.c
index 3f271c9cb5e0..a3e932f98419 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -7110,6 +7110,23 @@ static int dev_xdp_install(struct net_device *dev, bpf_op_t bpf_op,
return bpf_op(dev, &xdp);
}
+static void dev_xdp_uninstall(struct net_device *dev)
+{
+ struct netdev_bpf xdp;
+ bpf_op_t ndo_bpf;
+
+ ndo_bpf = dev->netdev_ops->ndo_bpf;
+ if (!ndo_bpf)
+ return;
+
+ __dev_xdp_query(dev, ndo_bpf, &xdp);
+ if (xdp.prog_attached == XDP_ATTACHED_NONE)
+ return;
+
+ /* Program removal should always succeed */
+ WARN_ON(dev_xdp_install(dev, ndo_bpf, NULL, xdp.prog_flags, NULL));
+}
+
/**
* dev_change_xdp_fd - set or clear a bpf program for a device rx path
* @dev: device
@@ -7240,6 +7257,7 @@ static void rollback_registered_many(struct list_head *head)
/* Shutdown queueing discipline. */
dev_shutdown(dev);
+ dev_xdp_uninstall(dev);
/* Notify protocols, that we are about to destroy
* this device. They should clean all the things.
--
2.14.1
^ permalink raw reply related [flat|nested] 22+ messages in thread* Re: [RFC net-next 3/6] net: xdp: make the stack take care of the tear down
2017-11-24 2:36 ` [RFC net-next 3/6] net: xdp: make the stack take care of the tear down Jakub Kicinski
@ 2017-11-24 23:24 ` Daniel Borkmann
2017-11-25 3:28 ` Jakub Kicinski
0 siblings, 1 reply; 22+ messages in thread
From: Daniel Borkmann @ 2017-11-24 23:24 UTC (permalink / raw)
To: Jakub Kicinski, netdev
Cc: oss-drivers, alexei.starovoitov, jiri, Saeed Mahameed,
Michael Chan, Ariel Elior, John Fastabend
On 11/24/2017 03:36 AM, Jakub Kicinski wrote:
> Since day one of XDP drivers had to remember to free the program
> on the remove path. This leads to code duplication and is error
> prone. Make the stack query the installed programs on unregister
> and if something is installed, remove the program.
>
> Because the remove will now be called before notifiers are
> invoked, BPF offload state of the program will not get destroyed
> before uninstall.
>
> Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
> Reviewed-by: Simon Horman <simon.horman@netronome.com>
[...]
Nice work, series looks good to me! One really just minor
comment below:
> diff --git a/net/core/dev.c b/net/core/dev.c
> index 3f271c9cb5e0..a3e932f98419 100644
> --- a/net/core/dev.c
> +++ b/net/core/dev.c
> @@ -7110,6 +7110,23 @@ static int dev_xdp_install(struct net_device *dev, bpf_op_t bpf_op,
> return bpf_op(dev, &xdp);
> }
>
> +static void dev_xdp_uninstall(struct net_device *dev)
> +{
> + struct netdev_bpf xdp;
> + bpf_op_t ndo_bpf;
Can you add a comment here stating that generic XDP does not
need to be handled since we drop the prog from free_netdev()?
Potentially we could also drop the generic one from here, that
way we'd make no difference and have a dev_xdp_install() and
one dev_xdp_uninstall() for all kind of attach types. Given
generic XDP should simulate native XDP anyway, probably better
to just do that.
> + ndo_bpf = dev->netdev_ops->ndo_bpf;
> + if (!ndo_bpf)
> + return;
> +
> + __dev_xdp_query(dev, ndo_bpf, &xdp);
> + if (xdp.prog_attached == XDP_ATTACHED_NONE)
> + return;
> +
> + /* Program removal should always succeed */
> + WARN_ON(dev_xdp_install(dev, ndo_bpf, NULL, xdp.prog_flags, NULL));
> +}
> +
> /**
> * dev_change_xdp_fd - set or clear a bpf program for a device rx path
> * @dev: device
> @@ -7240,6 +7257,7 @@ static void rollback_registered_many(struct list_head *head)
> /* Shutdown queueing discipline. */
> dev_shutdown(dev);
>
> + dev_xdp_uninstall(dev);
>
> /* Notify protocols, that we are about to destroy
> * this device. They should clean all the things.
>
Thanks,
Daniel
^ permalink raw reply [flat|nested] 22+ messages in thread* Re: [RFC net-next 3/6] net: xdp: make the stack take care of the tear down
2017-11-24 23:24 ` Daniel Borkmann
@ 2017-11-25 3:28 ` Jakub Kicinski
0 siblings, 0 replies; 22+ messages in thread
From: Jakub Kicinski @ 2017-11-25 3:28 UTC (permalink / raw)
To: Daniel Borkmann
Cc: netdev, oss-drivers, alexei.starovoitov, jiri, Saeed Mahameed,
Michael Chan, Ariel Elior, John Fastabend
On Sat, 25 Nov 2017 00:24:50 +0100, Daniel Borkmann wrote:
> > +static void dev_xdp_uninstall(struct net_device *dev)
> > +{
> > + struct netdev_bpf xdp;
> > + bpf_op_t ndo_bpf;
>
> Can you add a comment here stating that generic XDP does not
> need to be handled since we drop the prog from free_netdev()?
> Potentially we could also drop the generic one from here, that
> way we'd make no difference and have a dev_xdp_install() and
> one dev_xdp_uninstall() for all kind of attach types. Given
> generic XDP should simulate native XDP anyway, probably better
> to just do that.
I will move the freeing of generic XDP here and add a simple test to
the last patch. Thanks!
> > + ndo_bpf = dev->netdev_ops->ndo_bpf;
> > + if (!ndo_bpf)
> > + return;
> > +
> > + __dev_xdp_query(dev, ndo_bpf, &xdp);
> > + if (xdp.prog_attached == XDP_ATTACHED_NONE)
> > + return;
> > +
> > + /* Program removal should always succeed */
> > + WARN_ON(dev_xdp_install(dev, ndo_bpf, NULL, xdp.prog_flags, NULL));
> > +}
^ permalink raw reply [flat|nested] 22+ messages in thread
* [RFC net-next 4/6] netdevsim: add software driver for testing offloads
2017-11-24 2:36 [RFC net-next 0/6] xdp: make stack perform remove and tests Jakub Kicinski
` (2 preceding siblings ...)
2017-11-24 2:36 ` [RFC net-next 3/6] net: xdp: make the stack take care of the tear down Jakub Kicinski
@ 2017-11-24 2:36 ` Jakub Kicinski
2017-11-24 7:24 ` Jiri Pirko
2017-11-27 19:30 ` David Miller
2017-11-24 2:36 ` [RFC net-next 5/6] netdevsim: add bpf offload support Jakub Kicinski
` (3 subsequent siblings)
7 siblings, 2 replies; 22+ messages in thread
From: Jakub Kicinski @ 2017-11-24 2:36 UTC (permalink / raw)
To: netdev; +Cc: oss-drivers, daniel, alexei.starovoitov, jiri, Jakub Kicinski
To be able to run selftests without any hardware required we
need a software model. The model can also serve as an example
implementation for those implementing actual HW offloads.
The dummy driver have previously been extended to test SR-IOV,
but the general consensus seems to be against adding further
features to it.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
---
MAINTAINERS | 5 ++
drivers/net/Kconfig | 10 +++
drivers/net/Makefile | 1 +
drivers/net/netdevsim/Makefile | 6 ++
drivers/net/netdevsim/netdev.c | 136 ++++++++++++++++++++++++++++++++++++++
drivers/net/netdevsim/netdevsim.h | 44 ++++++++++++
6 files changed, 202 insertions(+)
create mode 100644 drivers/net/netdevsim/Makefile
create mode 100644 drivers/net/netdevsim/netdev.c
create mode 100644 drivers/net/netdevsim/netdevsim.h
diff --git a/MAINTAINERS b/MAINTAINERS
index 44512c346206..e70e9c6e2595 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -9596,6 +9596,11 @@ NETWORKING [WIRELESS]
L: linux-wireless@vger.kernel.org
Q: http://patchwork.kernel.org/project/linux-wireless/list/
+NETDEVSIM
+M: Jakub Kicinski <jakub.kicinski@netronome.com>
+S: Maintained
+F: drivers/net/netdevsim/*
+
NETXEN (1/10) GbE SUPPORT
M: Manish Chopra <manish.chopra@cavium.com>
M: Rahul Verma <rahul.verma@cavium.com>
diff --git a/drivers/net/Kconfig b/drivers/net/Kconfig
index 0936da592e12..0aeab01d2429 100644
--- a/drivers/net/Kconfig
+++ b/drivers/net/Kconfig
@@ -497,4 +497,14 @@ config THUNDERBOLT_NET
source "drivers/net/hyperv/Kconfig"
+config NETDEVSIM
+ tristate "Simulated networking device"
+ help
+ This driver is a developer testing tool and software model that can
+ be used to test various control path networking APIs, especially
+ HW-offload related.
+
+ To compile this driver as a module, choose M here: the module
+ will be called netdevsim.
+
endif # NETDEVICES
diff --git a/drivers/net/Makefile b/drivers/net/Makefile
index 766f62d02a0b..04c3b747812c 100644
--- a/drivers/net/Makefile
+++ b/drivers/net/Makefile
@@ -78,3 +78,4 @@ obj-$(CONFIG_FUJITSU_ES) += fjes/
thunderbolt-net-y += thunderbolt.o
obj-$(CONFIG_THUNDERBOLT_NET) += thunderbolt-net.o
+obj-$(CONFIG_NETDEVSIM) += netdevsim/
diff --git a/drivers/net/netdevsim/Makefile b/drivers/net/netdevsim/Makefile
new file mode 100644
index 000000000000..07867bfe873b
--- /dev/null
+++ b/drivers/net/netdevsim/Makefile
@@ -0,0 +1,6 @@
+# SPDX-License-Identifier: GPL-2.0
+
+obj-$(CONFIG_NETDEVSIM) += netdevsim.o
+
+netdevsim-objs := \
+ netdev.o \
diff --git a/drivers/net/netdevsim/netdev.c b/drivers/net/netdevsim/netdev.c
new file mode 100644
index 000000000000..6f6c78bb076b
--- /dev/null
+++ b/drivers/net/netdevsim/netdev.c
@@ -0,0 +1,136 @@
+/*
+ * Copyright (C) 2017 Netronome Systems, Inc.
+ *
+ * This software is dual licensed under the GNU General License Version 2,
+ * June 1991 as shown in the file COPYING in the top-level directory of this
+ * source tree or the BSD 2-Clause License provided below. You have the
+ * option to license this software under the complete terms of either license.
+ *
+ * The BSD 2-Clause License:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * 2. 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.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/etherdevice.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/netdevice.h>
+#include <linux/slab.h>
+#include <net/netlink.h>
+#include <net/rtnetlink.h>
+
+#include "netdevsim.h"
+
+static netdev_tx_t nsim_start_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+ struct netdevsim *ns = netdev_priv(dev);
+
+ u64_stats_update_begin(&ns->syncp);
+ ns->tx_packets++;
+ ns->tx_bytes += skb->len;
+ u64_stats_update_end(&ns->syncp);
+
+ dev_kfree_skb(skb);
+
+ return NETDEV_TX_OK;
+}
+
+static void nsim_set_rx_mode(struct net_device *dev)
+{
+}
+
+static void
+nsim_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats)
+{
+ struct netdevsim *ns = netdev_priv(dev);
+ unsigned int start;
+
+ do {
+ start = u64_stats_fetch_begin(&ns->syncp);
+ stats->tx_bytes = ns->tx_bytes;
+ stats->tx_packets = ns->tx_packets;
+ } while (u64_stats_fetch_retry(&ns->syncp, start));
+}
+
+static const struct net_device_ops nsim_netdev_ops = {
+ .ndo_start_xmit = nsim_start_xmit,
+ .ndo_set_rx_mode = nsim_set_rx_mode,
+ .ndo_set_mac_address = eth_mac_addr,
+ .ndo_validate_addr = eth_validate_addr,
+ .ndo_get_stats64 = nsim_get_stats64,
+};
+
+static void nsim_setup(struct net_device *dev)
+{
+ ether_setup(dev);
+ eth_hw_addr_random(dev);
+
+ dev->netdev_ops = &nsim_netdev_ops;
+ dev->needs_free_netdev = true;
+
+ dev->tx_queue_len = 0;
+ dev->flags |= IFF_NOARP;
+ dev->flags &= ~IFF_MULTICAST;
+ dev->priv_flags |= IFF_LIVE_ADDR_CHANGE |
+ IFF_NO_QUEUE;
+ dev->features |= NETIF_F_HIGHDMA |
+ NETIF_F_SG |
+ NETIF_F_FRAGLIST |
+ NETIF_F_HW_CSUM |
+ NETIF_F_TSO;
+ dev->max_mtu = ETH_MAX_MTU;
+}
+
+static int nsim_validate(struct nlattr *tb[], struct nlattr *data[],
+ struct netlink_ext_ack *extack)
+{
+ if (tb[IFLA_ADDRESS]) {
+ if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
+ return -EINVAL;
+ if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
+ return -EADDRNOTAVAIL;
+ }
+ return 0;
+}
+
+static struct rtnl_link_ops nsim_link_ops __read_mostly = {
+ .kind = DRV_NAME,
+ .priv_size = sizeof(struct netdevsim),
+ .setup = nsim_setup,
+ .validate = nsim_validate,
+};
+
+static int __init nsim_module_init(void)
+{
+ return rtnl_link_register(&nsim_link_ops);
+}
+
+static void __exit nsim_module_exit(void)
+{
+ rtnl_link_unregister(&nsim_link_ops);
+}
+
+module_init(nsim_module_init);
+module_exit(nsim_module_exit);
+MODULE_LICENSE("GPL");
+MODULE_ALIAS_RTNL_LINK(DRV_NAME);
diff --git a/drivers/net/netdevsim/netdevsim.h b/drivers/net/netdevsim/netdevsim.h
new file mode 100644
index 000000000000..8cf174ed55c5
--- /dev/null
+++ b/drivers/net/netdevsim/netdevsim.h
@@ -0,0 +1,44 @@
+/*
+ * Copyright (C) 2017 Netronome Systems, Inc.
+ *
+ * This software is dual licensed under the GNU General License Version 2,
+ * June 1991 as shown in the file COPYING in the top-level directory of this
+ * source tree or the BSD 2-Clause License provided below. You have the
+ * option to license this software under the complete terms of either license.
+ *
+ * The BSD 2-Clause License:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * 2. 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.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/kernel.h>
+#include <linux/netdevice.h>
+#include <linux/u64_stats_sync.h>
+
+#define DRV_NAME "netdevsim"
+
+struct netdevsim {
+ u64 tx_packets;
+ u64 tx_bytes;
+ struct u64_stats_sync syncp;
+};
--
2.14.1
^ permalink raw reply related [flat|nested] 22+ messages in thread* Re: [RFC net-next 4/6] netdevsim: add software driver for testing offloads
2017-11-24 2:36 ` [RFC net-next 4/6] netdevsim: add software driver for testing offloads Jakub Kicinski
@ 2017-11-24 7:24 ` Jiri Pirko
2017-11-24 7:49 ` Jakub Kicinski
2017-11-27 19:30 ` David Miller
1 sibling, 1 reply; 22+ messages in thread
From: Jiri Pirko @ 2017-11-24 7:24 UTC (permalink / raw)
To: Jakub Kicinski; +Cc: netdev, oss-drivers, daniel, alexei.starovoitov
Fri, Nov 24, 2017 at 03:36:11AM CET, jakub.kicinski@netronome.com wrote:
>To be able to run selftests without any hardware required we
>need a software model. The model can also serve as an example
>implementation for those implementing actual HW offloads.
>The dummy driver have previously been extended to test SR-IOV,
>but the general consensus seems to be against adding further
>features to it.
>
>Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
>Reviewed-by: Simon Horman <simon.horman@netronome.com>
>---
[...]
>+++ b/drivers/net/netdevsim/netdev.c
>@@ -0,0 +1,136 @@
>+/*
>+ * Copyright (C) 2017 Netronome Systems, Inc.
>+ *
>+ * This software is dual licensed under the GNU General License Version 2,
>+ * June 1991 as shown in the file COPYING in the top-level directory of this
>+ * source tree or the BSD 2-Clause License provided below. You have the
>+ * option to license this software under the complete terms of either license.
>+ *
>+ * The BSD 2-Clause License:
Why gpl2 is not enough for this?
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [RFC net-next 4/6] netdevsim: add software driver for testing offloads
2017-11-24 7:24 ` Jiri Pirko
@ 2017-11-24 7:49 ` Jakub Kicinski
2017-11-24 8:07 ` Jiri Pirko
0 siblings, 1 reply; 22+ messages in thread
From: Jakub Kicinski @ 2017-11-24 7:49 UTC (permalink / raw)
To: Jiri Pirko
Cc: Linux Netdev List, oss-drivers, Daniel Borkmann, Alexei Starovoitov
On Thu, Nov 23, 2017 at 11:24 PM, Jiri Pirko <jiri@resnulli.us> wrote:
> Fri, Nov 24, 2017 at 03:36:11AM CET, jakub.kicinski@netronome.com wrote:
>>To be able to run selftests without any hardware required we
>>need a software model. The model can also serve as an example
>>implementation for those implementing actual HW offloads.
>>The dummy driver have previously been extended to test SR-IOV,
>>but the general consensus seems to be against adding further
>>features to it.
>>
>>Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
>>Reviewed-by: Simon Horman <simon.horman@netronome.com>
>>---
>
> [...]
>
>
>>+++ b/drivers/net/netdevsim/netdev.c
>>@@ -0,0 +1,136 @@
>>+/*
>>+ * Copyright (C) 2017 Netronome Systems, Inc.
>>+ *
>>+ * This software is dual licensed under the GNU General License Version 2,
>>+ * June 1991 as shown in the file COPYING in the top-level directory of this
>>+ * source tree or the BSD 2-Clause License provided below. You have the
>>+ * option to license this software under the complete terms of either license.
>>+ *
>>+ * The BSD 2-Clause License:
>
> Why gpl2 is not enough for this?
It's the license I got from legal, I will request permission to use
pure gpl2. Thanks!
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [RFC net-next 4/6] netdevsim: add software driver for testing offloads
2017-11-24 7:49 ` Jakub Kicinski
@ 2017-11-24 8:07 ` Jiri Pirko
0 siblings, 0 replies; 22+ messages in thread
From: Jiri Pirko @ 2017-11-24 8:07 UTC (permalink / raw)
To: Jakub Kicinski
Cc: Linux Netdev List, oss-drivers, Daniel Borkmann, Alexei Starovoitov
Fri, Nov 24, 2017 at 08:49:17AM CET, jakub.kicinski@netronome.com wrote:
>On Thu, Nov 23, 2017 at 11:24 PM, Jiri Pirko <jiri@resnulli.us> wrote:
>> Fri, Nov 24, 2017 at 03:36:11AM CET, jakub.kicinski@netronome.com wrote:
>>>To be able to run selftests without any hardware required we
>>>need a software model. The model can also serve as an example
>>>implementation for those implementing actual HW offloads.
>>>The dummy driver have previously been extended to test SR-IOV,
>>>but the general consensus seems to be against adding further
>>>features to it.
>>>
>>>Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
>>>Reviewed-by: Simon Horman <simon.horman@netronome.com>
>>>---
>>
>> [...]
>>
>>
>>>+++ b/drivers/net/netdevsim/netdev.c
>>>@@ -0,0 +1,136 @@
>>>+/*
>>>+ * Copyright (C) 2017 Netronome Systems, Inc.
>>>+ *
>>>+ * This software is dual licensed under the GNU General License Version 2,
>>>+ * June 1991 as shown in the file COPYING in the top-level directory of this
>>>+ * source tree or the BSD 2-Clause License provided below. You have the
>>>+ * option to license this software under the complete terms of either license.
>>>+ *
>>>+ * The BSD 2-Clause License:
>>
>> Why gpl2 is not enough for this?
>
>It's the license I got from legal, I will request permission to use
>pure gpl2. Thanks!
Yeah, I semi-understand need for bsd for actual hw driver (we have it
for mlxsw as well). But for this testing driver, it really does not make
sense.
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [RFC net-next 4/6] netdevsim: add software driver for testing offloads
2017-11-24 2:36 ` [RFC net-next 4/6] netdevsim: add software driver for testing offloads Jakub Kicinski
2017-11-24 7:24 ` Jiri Pirko
@ 2017-11-27 19:30 ` David Miller
2017-11-27 19:42 ` Jakub Kicinski
1 sibling, 1 reply; 22+ messages in thread
From: David Miller @ 2017-11-27 19:30 UTC (permalink / raw)
To: jakub.kicinski; +Cc: netdev, oss-drivers, daniel, alexei.starovoitov, jiri
From: Jakub Kicinski <jakub.kicinski@netronome.com>
Date: Thu, 23 Nov 2017 18:36:11 -0800
> The dummy driver have previously been extended to test SR-IOV,
> but the general consensus seems to be against adding further
> features to it.
I guess this is fine, but then is it going to be the case that
every time we want to add a software offload implementation for
testing driver paths we add yet another dummy driver?
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [RFC net-next 4/6] netdevsim: add software driver for testing offloads
2017-11-27 19:30 ` David Miller
@ 2017-11-27 19:42 ` Jakub Kicinski
2017-11-28 14:55 ` Phil Sutter
2017-11-28 16:20 ` Jiri Pirko
0 siblings, 2 replies; 22+ messages in thread
From: Jakub Kicinski @ 2017-11-27 19:42 UTC (permalink / raw)
To: David Miller
Cc: netdev, oss-drivers, daniel, alexei.starovoitov, jiri,
Sabrina Dubroca, Phil Sutter
On Tue, 28 Nov 2017 04:30:26 +0900 (KST), David Miller wrote:
> From: Jakub Kicinski <jakub.kicinski@netronome.com>
> Date: Thu, 23 Nov 2017 18:36:11 -0800
>
> > The dummy driver have previously been extended to test SR-IOV,
> > but the general consensus seems to be against adding further
> > features to it.
>
> I guess this is fine, but then is it going to be the case that
> every time we want to add a software offload implementation for
> testing driver paths we add yet another dummy driver?
I would expect others to extend the netdevsim driver, so all test/SW
implementation would live there.
I was considering moving the SR-IOV code over, too, but I'm worried
I will break existing tests. Phil, Sabrina would you be OK with that?
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [RFC net-next 4/6] netdevsim: add software driver for testing offloads
2017-11-27 19:42 ` Jakub Kicinski
@ 2017-11-28 14:55 ` Phil Sutter
2017-11-28 16:20 ` Jiri Pirko
1 sibling, 0 replies; 22+ messages in thread
From: Phil Sutter @ 2017-11-28 14:55 UTC (permalink / raw)
To: Jakub Kicinski
Cc: David Miller, netdev, oss-drivers, daniel, alexei.starovoitov,
jiri, Sabrina Dubroca
Hi,
On Mon, Nov 27, 2017 at 11:42:49AM -0800, Jakub Kicinski wrote:
> On Tue, 28 Nov 2017 04:30:26 +0900 (KST), David Miller wrote:
> > From: Jakub Kicinski <jakub.kicinski@netronome.com>
> > Date: Thu, 23 Nov 2017 18:36:11 -0800
> >
> > > The dummy driver have previously been extended to test SR-IOV,
> > > but the general consensus seems to be against adding further
> > > features to it.
> >
> > I guess this is fine, but then is it going to be the case that
> > every time we want to add a software offload implementation for
> > testing driver paths we add yet another dummy driver?
>
> I would expect others to extend the netdevsim driver, so all test/SW
> implementation would live there.
>
> I was considering moving the SR-IOV code over, too, but I'm worried
> I will break existing tests. Phil, Sabrina would you be OK with that?
Yes, that's fine with us!
Thanks, Phil
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [RFC net-next 4/6] netdevsim: add software driver for testing offloads
2017-11-27 19:42 ` Jakub Kicinski
2017-11-28 14:55 ` Phil Sutter
@ 2017-11-28 16:20 ` Jiri Pirko
1 sibling, 0 replies; 22+ messages in thread
From: Jiri Pirko @ 2017-11-28 16:20 UTC (permalink / raw)
To: Jakub Kicinski
Cc: David Miller, netdev, oss-drivers, daniel, alexei.starovoitov,
Sabrina Dubroca, Phil Sutter
Mon, Nov 27, 2017 at 08:42:49PM CET, jakub.kicinski@netronome.com wrote:
>On Tue, 28 Nov 2017 04:30:26 +0900 (KST), David Miller wrote:
>> From: Jakub Kicinski <jakub.kicinski@netronome.com>
>> Date: Thu, 23 Nov 2017 18:36:11 -0800
>>
>> > The dummy driver have previously been extended to test SR-IOV,
>> > but the general consensus seems to be against adding further
>> > features to it.
>>
>> I guess this is fine, but then is it going to be the case that
>> every time we want to add a software offload implementation for
>> testing driver paths we add yet another dummy driver?
>
>I would expect others to extend the netdevsim driver, so all test/SW
>implementation would live there.
>
>I was considering moving the SR-IOV code over, too, but I'm worried
Agreed. Leave dummy be dummy. netdevsim should serve this purpose.
Thanks!
>I will break existing tests. Phil, Sabrina would you be OK with that?
^ permalink raw reply [flat|nested] 22+ messages in thread
* [RFC net-next 5/6] netdevsim: add bpf offload support
2017-11-24 2:36 [RFC net-next 0/6] xdp: make stack perform remove and tests Jakub Kicinski
` (3 preceding siblings ...)
2017-11-24 2:36 ` [RFC net-next 4/6] netdevsim: add software driver for testing offloads Jakub Kicinski
@ 2017-11-24 2:36 ` Jakub Kicinski
2017-11-24 2:36 ` [RFC net-next 6/6] selftests/bpf: add offload test based on netdevsim Jakub Kicinski
` (2 subsequent siblings)
7 siblings, 0 replies; 22+ messages in thread
From: Jakub Kicinski @ 2017-11-24 2:36 UTC (permalink / raw)
To: netdev; +Cc: oss-drivers, daniel, alexei.starovoitov, jiri, Jakub Kicinski
Add support for loading programs for netdevsim devices and
expose the related information via DebugFS. Both offload
of XDP and cls_bpf programs is supported.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
---
drivers/net/netdevsim/Makefile | 1 +
drivers/net/netdevsim/bpf.c | 391 ++++++++++++++++++++++++++++++++++++++
drivers/net/netdevsim/netdev.c | 107 ++++++++++-
drivers/net/netdevsim/netdevsim.h | 40 ++++
4 files changed, 538 insertions(+), 1 deletion(-)
create mode 100644 drivers/net/netdevsim/bpf.c
diff --git a/drivers/net/netdevsim/Makefile b/drivers/net/netdevsim/Makefile
index 07867bfe873b..074ddebbc41d 100644
--- a/drivers/net/netdevsim/Makefile
+++ b/drivers/net/netdevsim/Makefile
@@ -4,3 +4,4 @@ obj-$(CONFIG_NETDEVSIM) += netdevsim.o
netdevsim-objs := \
netdev.o \
+ bpf.o \
diff --git a/drivers/net/netdevsim/bpf.c b/drivers/net/netdevsim/bpf.c
new file mode 100644
index 000000000000..7738412c8e5a
--- /dev/null
+++ b/drivers/net/netdevsim/bpf.c
@@ -0,0 +1,391 @@
+/*
+ * Copyright (C) 2017 Netronome Systems, Inc.
+ *
+ * This software is dual licensed under the GNU General License Version 2,
+ * June 1991 as shown in the file COPYING in the top-level directory of this
+ * source tree or the BSD 2-Clause License provided below. You have the
+ * option to license this software under the complete terms of either license.
+ *
+ * The BSD 2-Clause License:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * 2. 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.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/bpf.h>
+#include <linux/bpf_verifier.h>
+#include <linux/debugfs.h>
+#include <linux/kernel.h>
+#include <linux/rtnetlink.h>
+#include <net/pkt_cls.h>
+
+#include "netdevsim.h"
+
+struct nsim_bpf_bound_prog {
+ struct netdevsim *ns;
+ struct bpf_prog *prog;
+ struct dentry *ddir;
+ const char *state;
+ bool is_loaded;
+ struct list_head l;
+};
+
+static int nsim_debugfs_bpf_string_read(struct seq_file *file, void *data)
+{
+ const char **str = file->private;
+
+ if (*str)
+ seq_printf(file, "%s\n", *str);
+
+ return 0;
+}
+
+static int nsim_debugfs_bpf_string_open(struct inode *inode, struct file *f)
+{
+ return single_open(f, nsim_debugfs_bpf_string_read, inode->i_private);
+}
+
+static const struct file_operations nsim_bpf_string_fops = {
+ .owner = THIS_MODULE,
+ .open = nsim_debugfs_bpf_string_open,
+ .release = single_release,
+ .read = seq_read,
+ .llseek = seq_lseek
+};
+
+static int
+nsim_bpf_verify_insn(struct bpf_verifier_env *env, int insn_idx, int prev_insn)
+{
+ struct nsim_bpf_bound_prog *state;
+
+ state = env->prog->aux->offload->dev_priv;
+ if (state->ns->bpf_bind_verifier_delay && !insn_idx)
+ msleep(state->ns->bpf_bind_verifier_delay);
+
+ return 0;
+}
+
+static const struct bpf_ext_analyzer_ops nsim_bpf_analyzer_ops = {
+ .insn_hook = nsim_bpf_verify_insn,
+};
+
+static bool nsim_xdp_offload_active(struct netdevsim *ns)
+{
+ return ns->xdp_prog_mode == XDP_ATTACHED_HW;
+}
+
+static void nsim_prog_set_loaded(struct bpf_prog *prog, bool loaded)
+{
+ struct nsim_bpf_bound_prog *state;
+
+ if (!prog || !prog->aux->offload)
+ return;
+
+ state = prog->aux->offload->dev_priv;
+ state->is_loaded = loaded;
+}
+
+static int
+nsim_bpf_offload(struct netdevsim *ns, struct bpf_prog *prog, bool oldprog)
+{
+ nsim_prog_set_loaded(ns->bpf_offloaded, false);
+
+ WARN(!!ns->bpf_offloaded != oldprog,
+ "bad offload state, expected offload %sto be active",
+ oldprog ? "" : "not ");
+ ns->bpf_offloaded = prog;
+ ns->bpf_offloaded_id = prog ? prog->aux->id : 0;
+ nsim_prog_set_loaded(prog, true);
+
+ return 0;
+}
+
+int nsim_bpf_setup_tc_block_cb(enum tc_setup_type type,
+ void *type_data, void *cb_priv)
+{
+ struct tc_cls_bpf_offload *cls_bpf = type_data;
+ struct netdevsim *ns = cb_priv;
+ bool skip_sw;
+
+ if (type != TC_SETUP_CLSBPF ||
+ !tc_can_offload(ns->netdev) ||
+ cls_bpf->common.protocol != htons(ETH_P_ALL) ||
+ cls_bpf->common.chain_index)
+ return -EOPNOTSUPP;
+
+ skip_sw = cls_bpf->gen_flags & TCA_CLS_FLAGS_SKIP_SW;
+
+ if (nsim_xdp_offload_active(ns))
+ return -EBUSY;
+
+ if (!ns->bpf_tc_accept)
+ return -EOPNOTSUPP;
+ /* Note: progs without skip_sw will probably not be dev bound */
+ if (!ns->bpf_tc_no_skip_accept && !skip_sw)
+ return -EOPNOTSUPP;
+
+ switch (cls_bpf->command) {
+ case TC_CLSBPF_REPLACE:
+ return nsim_bpf_offload(ns, cls_bpf->prog, true);
+ case TC_CLSBPF_ADD:
+ return nsim_bpf_offload(ns, cls_bpf->prog, false);
+ case TC_CLSBPF_DESTROY:
+ return nsim_bpf_offload(ns, NULL, true);
+ default:
+ return -EOPNOTSUPP;
+ }
+}
+
+int nsim_bpf_disable_tc(struct netdevsim *ns)
+{
+ if (ns->bpf_offloaded && !nsim_xdp_offload_active(ns))
+ return -EBUSY;
+ return 0;
+}
+
+static int nsim_xdp_offload_prog(struct netdevsim *ns, struct netdev_bpf *bpf)
+{
+ if (!nsim_xdp_offload_active(ns) && !bpf->prog)
+ return 0;
+ if (!nsim_xdp_offload_active(ns) && bpf->prog && ns->bpf_offloaded) {
+ NSIM_EA(bpf->extack, "TC program is already loaded");
+ return -EBUSY;
+ }
+
+ return nsim_bpf_offload(ns, bpf->prog, nsim_xdp_offload_active(ns));
+}
+
+static int nsim_xdp_set_prog(struct netdevsim *ns, struct netdev_bpf *bpf)
+{
+ int err;
+
+ if (ns->xdp_prog && (bpf->flags ^ ns->xdp_flags) & XDP_FLAGS_MODES) {
+ NSIM_EA(bpf->extack, "program loaded with different flags");
+ return -EBUSY;
+ }
+
+ if (bpf->command == XDP_SETUP_PROG && !ns->bpf_xdpdrv_accept) {
+ NSIM_EA(bpf->extack, "driver XDP disabled in DebugFS");
+ return -EOPNOTSUPP;
+ }
+ if (bpf->command == XDP_SETUP_PROG_HW && !ns->bpf_xdpoffload_accept) {
+ NSIM_EA(bpf->extack, "XDP offload disabled in DebugFS");
+ return -EOPNOTSUPP;
+ }
+
+ if (bpf->command == XDP_SETUP_PROG_HW) {
+ err = nsim_xdp_offload_prog(ns, bpf);
+ if (err)
+ return err;
+ }
+
+ if (ns->xdp_prog)
+ bpf_prog_put(ns->xdp_prog);
+
+ ns->xdp_prog = bpf->prog;
+ ns->xdp_flags = bpf->flags;
+
+ if (!bpf->prog)
+ ns->xdp_prog_mode = XDP_ATTACHED_NONE;
+ else if (bpf->command == XDP_SETUP_PROG)
+ ns->xdp_prog_mode = XDP_ATTACHED_DRV;
+ else
+ ns->xdp_prog_mode = XDP_ATTACHED_HW;
+
+ return 0;
+}
+
+int nsim_bpf_create_prog(struct netdevsim *ns, struct bpf_prog *prog)
+{
+ struct nsim_bpf_bound_prog *state;
+ char name[16];
+ int err;
+
+ state = kzalloc(sizeof(*state), GFP_KERNEL);
+ if (!state)
+ return -ENOMEM;
+
+ state->ns = ns;
+ state->prog = prog;
+ state->state = "verify";
+
+ /* Program id is not populated yet when we create the state. */
+ sprintf(name, "%u", ns->prog_id_gen++);
+ state->ddir = debugfs_create_dir(name, ns->ddir_bpf_bound_progs);
+ if (IS_ERR(state->ddir)) {
+ err = PTR_ERR(state->ddir);
+ kfree(state);
+ return err;
+ }
+
+ debugfs_create_u32("id", 0400, state->ddir, &prog->aux->id);
+ debugfs_create_file("state", 0400, state->ddir,
+ &state->state, &nsim_bpf_string_fops);
+ debugfs_create_bool("loaded", 0400, state->ddir, &state->is_loaded);
+
+ list_add_tail(&state->l, &ns->bpf_bound_progs);
+
+ prog->aux->offload->dev_priv = state;
+
+ return 0;
+}
+
+void nsim_bpf_destroy_prog(struct bpf_prog *prog)
+{
+ struct nsim_bpf_bound_prog *state;
+
+ state = prog->aux->offload->dev_priv;
+ WARN(state->is_loaded,
+ "offload state destroyed while program still bound");
+ debugfs_remove_recursive(state->ddir);
+ list_del(&state->l);
+ kfree(state);
+}
+
+static int nsim_setup_prog_checks(struct netdevsim *ns, struct netdev_bpf *bpf)
+{
+ if (bpf->prog && bpf->prog->aux->offload) {
+ NSIM_EA(bpf->extack, "attempt to load offloaded prog to drv");
+ return -EINVAL;
+ }
+ if (ns->netdev->mtu > NSIM_XDP_MAX_MTU) {
+ NSIM_EA(bpf->extack, "MTU too large w/ XDP enabled");
+ return -EINVAL;
+ }
+ if (nsim_xdp_offload_active(ns)) {
+ NSIM_EA(bpf->extack, "xdp offload active, can't load drv prog");
+ return -EBUSY;
+ }
+ return 0;
+}
+
+static int
+nsim_setup_prog_hw_checks(struct netdevsim *ns, struct netdev_bpf *bpf)
+{
+ struct nsim_bpf_bound_prog *state;
+
+ if (!bpf->prog)
+ return 0;
+
+ if (!bpf->prog->aux->offload) {
+ NSIM_EA(bpf->extack, "xdpoffload of non-bound program");
+ return -EINVAL;
+ }
+ if (bpf->prog->aux->offload->netdev != ns->netdev) {
+ NSIM_EA(bpf->extack, "program bound to different dev");
+ return -EINVAL;
+ }
+
+ state = bpf->prog->aux->offload->dev_priv;
+ if (WARN_ON(strcmp(state->state, "xlated"))) {
+ NSIM_EA(bpf->extack, "offloading program in bad state");
+ return -EINVAL;
+ }
+ return 0;
+}
+
+int nsim_bpf(struct net_device *dev, struct netdev_bpf *bpf)
+{
+ struct netdevsim *ns = netdev_priv(dev);
+ struct nsim_bpf_bound_prog *state;
+ int err;
+
+ ASSERT_RTNL();
+
+ switch (bpf->command) {
+ case BPF_OFFLOAD_VERIFIER_PREP:
+ if (!ns->bpf_bind_accept)
+ return -EOPNOTSUPP;
+
+ err = nsim_bpf_create_prog(ns, bpf->verifier.prog);
+ if (err)
+ return err;
+
+ bpf->verifier.ops = &nsim_bpf_analyzer_ops;
+ return 0;
+ case BPF_OFFLOAD_TRANSLATE:
+ state = bpf->offload.prog->aux->offload->dev_priv;
+
+ state->state = "xlated";
+ return 0;
+ case BPF_OFFLOAD_DESTROY:
+ nsim_bpf_destroy_prog(bpf->offload.prog);
+ return 0;
+ case XDP_QUERY_PROG:
+ bpf->prog_attached = ns->xdp_prog_mode;
+ bpf->prog_id = ns->xdp_prog ? ns->xdp_prog->aux->id : 0;
+ bpf->prog_flags = ns->xdp_prog ? ns->xdp_flags : 0;
+ return 0;
+ case XDP_SETUP_PROG:
+ err = nsim_setup_prog_checks(ns, bpf);
+ if (err)
+ return err;
+
+ return nsim_xdp_set_prog(ns, bpf);
+ case XDP_SETUP_PROG_HW:
+ err = nsim_setup_prog_hw_checks(ns, bpf);
+ if (err)
+ return err;
+
+ return nsim_xdp_set_prog(ns, bpf);
+ default:
+ return -EINVAL;
+ }
+}
+
+int nsim_bpf_init(struct netdevsim *ns)
+{
+ INIT_LIST_HEAD(&ns->bpf_bound_progs);
+
+ debugfs_create_u32("bpf_offloaded_id", 0400, ns->ddir,
+ &ns->bpf_offloaded_id);
+
+ ns->bpf_bind_accept = true;
+ debugfs_create_bool("bpf_bind_accept", 0600, ns->ddir,
+ &ns->bpf_bind_accept);
+ debugfs_create_u32("bpf_bind_verifier_delay", 0600, ns->ddir,
+ &ns->bpf_bind_verifier_delay);
+ ns->ddir_bpf_bound_progs =
+ debugfs_create_dir("bpf_bound_progs", ns->ddir);
+
+ ns->bpf_tc_accept = true;
+ debugfs_create_bool("bpf_tc_accept", 0600, ns->ddir,
+ &ns->bpf_tc_accept);
+ ns->bpf_tc_no_skip_accept = true;
+ debugfs_create_bool("bpf_tc_no_skip_accept", 0600, ns->ddir,
+ &ns->bpf_tc_no_skip_accept);
+ ns->bpf_xdpdrv_accept = true;
+ debugfs_create_bool("bpf_xdpdrv_accept", 0600, ns->ddir,
+ &ns->bpf_xdpdrv_accept);
+ ns->bpf_xdpoffload_accept = true;
+ debugfs_create_bool("bpf_xdpoffload_accept", 0600, ns->ddir,
+ &ns->bpf_xdpoffload_accept);
+
+ return 0;
+}
+
+void nsim_bpf_uninit(struct netdevsim *ns)
+{
+ WARN_ON(!list_empty(&ns->bpf_bound_progs));
+ WARN_ON(ns->xdp_prog);
+ WARN_ON(ns->bpf_offloaded);
+}
diff --git a/drivers/net/netdevsim/netdev.c b/drivers/net/netdevsim/netdev.c
index 6f6c78bb076b..93987879b73c 100644
--- a/drivers/net/netdevsim/netdev.c
+++ b/drivers/net/netdevsim/netdev.c
@@ -31,16 +31,36 @@
* SOFTWARE.
*/
+#include <linux/debugfs.h>
#include <linux/etherdevice.h>
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/netdevice.h>
#include <linux/slab.h>
#include <net/netlink.h>
+#include <net/pkt_cls.h>
#include <net/rtnetlink.h>
#include "netdevsim.h"
+static int nsim_init(struct net_device *dev)
+{
+ struct netdevsim *ns = netdev_priv(dev);
+
+ ns->netdev = dev;
+ ns->ddir = debugfs_create_dir(netdev_name(dev), nsim_ddir);
+
+ return nsim_bpf_init(ns);
+}
+
+static void nsim_uninit(struct net_device *dev)
+{
+ struct netdevsim *ns = netdev_priv(dev);
+
+ debugfs_remove_recursive(ns->ddir);
+ nsim_bpf_uninit(ns);
+}
+
static netdev_tx_t nsim_start_xmit(struct sk_buff *skb, struct net_device *dev)
{
struct netdevsim *ns = netdev_priv(dev);
@@ -59,6 +79,19 @@ static void nsim_set_rx_mode(struct net_device *dev)
{
}
+static int nsim_change_mtu(struct net_device *dev, int new_mtu)
+{
+ struct netdevsim *ns = netdev_priv(dev);
+
+ if (ns->xdp_prog_mode == XDP_ATTACHED_DRV &&
+ new_mtu > NSIM_XDP_MAX_MTU)
+ return -EBUSY;
+
+ dev->mtu = new_mtu;
+
+ return 0;
+}
+
static void
nsim_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats)
{
@@ -72,12 +105,66 @@ nsim_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats)
} while (u64_stats_fetch_retry(&ns->syncp, start));
}
+static int
+nsim_setup_tc_block_cb(enum tc_setup_type type, void *type_data, void *cb_priv)
+{
+ return nsim_bpf_setup_tc_block_cb(type, type_data, cb_priv);
+}
+
+static int
+nsim_setup_tc_block(struct net_device *dev, struct tc_block_offload *f)
+{
+ struct netdevsim *ns = netdev_priv(dev);
+
+ if (f->binder_type != TCF_BLOCK_BINDER_TYPE_CLSACT_INGRESS)
+ return -EOPNOTSUPP;
+
+ switch (f->command) {
+ case TC_BLOCK_BIND:
+ return tcf_block_cb_register(f->block, nsim_setup_tc_block_cb,
+ ns, ns);
+ case TC_BLOCK_UNBIND:
+ tcf_block_cb_unregister(f->block, nsim_setup_tc_block_cb, ns);
+ return 0;
+ default:
+ return -EOPNOTSUPP;
+ }
+}
+
+static int
+nsim_setup_tc(struct net_device *dev, enum tc_setup_type type, void *type_data)
+{
+ switch (type) {
+ case TC_SETUP_BLOCK:
+ return nsim_setup_tc_block(dev, type_data);
+ default:
+ return -EOPNOTSUPP;
+ }
+}
+
+static int
+nsim_set_features(struct net_device *dev, netdev_features_t features)
+{
+ struct netdevsim *ns = netdev_priv(dev);
+
+ if ((dev->features & NETIF_F_HW_TC) > (features & NETIF_F_HW_TC))
+ return nsim_bpf_disable_tc(ns);
+
+ return 0;
+}
+
static const struct net_device_ops nsim_netdev_ops = {
+ .ndo_init = nsim_init,
+ .ndo_uninit = nsim_uninit,
.ndo_start_xmit = nsim_start_xmit,
.ndo_set_rx_mode = nsim_set_rx_mode,
.ndo_set_mac_address = eth_mac_addr,
.ndo_validate_addr = eth_validate_addr,
+ .ndo_change_mtu = nsim_change_mtu,
.ndo_get_stats64 = nsim_get_stats64,
+ .ndo_setup_tc = nsim_setup_tc,
+ .ndo_set_features = nsim_set_features,
+ .ndo_bpf = nsim_bpf,
};
static void nsim_setup(struct net_device *dev)
@@ -98,6 +185,7 @@ static void nsim_setup(struct net_device *dev)
NETIF_F_FRAGLIST |
NETIF_F_HW_CSUM |
NETIF_F_TSO;
+ dev->hw_features |= NETIF_F_HW_TC;
dev->max_mtu = ETH_MAX_MTU;
}
@@ -120,14 +208,31 @@ static struct rtnl_link_ops nsim_link_ops __read_mostly = {
.validate = nsim_validate,
};
+struct dentry *nsim_ddir;
+
static int __init nsim_module_init(void)
{
- return rtnl_link_register(&nsim_link_ops);
+ int err;
+
+ nsim_ddir = debugfs_create_dir(DRV_NAME, NULL);
+ if (IS_ERR(nsim_ddir))
+ return PTR_ERR(nsim_ddir);
+
+ err = rtnl_link_register(&nsim_link_ops);
+ if (err)
+ goto err_debugfs_destroy;
+
+ return 0;
+
+err_debugfs_destroy:
+ debugfs_remove_recursive(nsim_ddir);
+ return err;
}
static void __exit nsim_module_exit(void)
{
rtnl_link_unregister(&nsim_link_ops);
+ debugfs_remove_recursive(nsim_ddir);
}
module_init(nsim_module_init);
diff --git a/drivers/net/netdevsim/netdevsim.h b/drivers/net/netdevsim/netdevsim.h
index 8cf174ed55c5..502e6ff7bf01 100644
--- a/drivers/net/netdevsim/netdevsim.h
+++ b/drivers/net/netdevsim/netdevsim.h
@@ -32,13 +32,53 @@
*/
#include <linux/kernel.h>
+#include <linux/list.h>
#include <linux/netdevice.h>
#include <linux/u64_stats_sync.h>
#define DRV_NAME "netdevsim"
+#define NSIM_XDP_MAX_MTU 4000
+
+#define NSIM_EA(extack, msg) NL_SET_ERR_MSG_MOD((extack), msg)
+
+struct bpf_prog;
+struct dentry;
+
struct netdevsim {
+ struct net_device *netdev;
+
u64 tx_packets;
u64 tx_bytes;
struct u64_stats_sync syncp;
+
+ struct dentry *ddir;
+
+ struct bpf_prog *bpf_offloaded;
+ u32 bpf_offloaded_id;
+
+ u32 xdp_flags;
+ int xdp_prog_mode;
+ struct bpf_prog *xdp_prog;
+
+ u32 prog_id_gen;
+
+ bool bpf_bind_accept;
+ u32 bpf_bind_verifier_delay;
+ struct dentry *ddir_bpf_bound_progs;
+ struct list_head bpf_bound_progs;
+
+ bool bpf_tc_accept;
+ bool bpf_tc_no_skip_accept;
+ bool bpf_xdpdrv_accept;
+ bool bpf_xdpoffload_accept;
};
+
+extern struct dentry *nsim_ddir;
+
+int nsim_bpf_init(struct netdevsim *ns);
+void nsim_bpf_uninit(struct netdevsim *ns);
+int nsim_bpf(struct net_device *dev, struct netdev_bpf *bpf);
+int nsim_bpf_disable_tc(struct netdevsim *ns);
+int nsim_bpf_setup_tc_block_cb(enum tc_setup_type type,
+ void *type_data, void *cb_priv);
--
2.14.1
^ permalink raw reply related [flat|nested] 22+ messages in thread* [RFC net-next 6/6] selftests/bpf: add offload test based on netdevsim
2017-11-24 2:36 [RFC net-next 0/6] xdp: make stack perform remove and tests Jakub Kicinski
` (4 preceding siblings ...)
2017-11-24 2:36 ` [RFC net-next 5/6] netdevsim: add bpf offload support Jakub Kicinski
@ 2017-11-24 2:36 ` Jakub Kicinski
2017-11-24 7:45 ` [RFC net-next 0/6] xdp: make stack perform remove and tests Jiri Pirko
2017-11-26 1:02 ` Alexei Starovoitov
7 siblings, 0 replies; 22+ messages in thread
From: Jakub Kicinski @ 2017-11-24 2:36 UTC (permalink / raw)
To: netdev; +Cc: oss-drivers, daniel, alexei.starovoitov, jiri, Jakub Kicinski
Add a test of BPF offload control path interfaces based on
just-added netdevsim driver. Perform various checks of both
the stack and the expected driver behaviour.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
---
tools/testing/selftests/bpf/Makefile | 5 +-
tools/testing/selftests/bpf/sample_ret0.c | 8 +
tools/testing/selftests/bpf/test_offload.py | 634 ++++++++++++++++++++++++++++
3 files changed, 645 insertions(+), 2 deletions(-)
create mode 100644 tools/testing/selftests/bpf/sample_ret0.c
create mode 100755 tools/testing/selftests/bpf/test_offload.py
diff --git a/tools/testing/selftests/bpf/Makefile b/tools/testing/selftests/bpf/Makefile
index 333a48655ee0..2c9d8c63c6fa 100644
--- a/tools/testing/selftests/bpf/Makefile
+++ b/tools/testing/selftests/bpf/Makefile
@@ -17,9 +17,10 @@ TEST_GEN_PROGS = test_verifier test_tag test_maps test_lru_map test_lpm_map test
TEST_GEN_FILES = test_pkt_access.o test_xdp.o test_l4lb.o test_tcp_estats.o test_obj_id.o \
test_pkt_md_access.o test_xdp_redirect.o test_xdp_meta.o sockmap_parse_prog.o \
- sockmap_verdict_prog.o dev_cgroup.o
+ sockmap_verdict_prog.o dev_cgroup.o sample_ret0.o
-TEST_PROGS := test_kmod.sh test_xdp_redirect.sh test_xdp_meta.sh
+TEST_PROGS := test_kmod.sh test_xdp_redirect.sh test_xdp_meta.sh \
+ test_offload.py
include ../lib.mk
diff --git a/tools/testing/selftests/bpf/sample_ret0.c b/tools/testing/selftests/bpf/sample_ret0.c
new file mode 100644
index 000000000000..62b071426a1e
--- /dev/null
+++ b/tools/testing/selftests/bpf/sample_ret0.c
@@ -0,0 +1,8 @@
+/* SPDX-License-Identifier: (GPL-2.0 OR BSD-2-Clause) */
+/* Copyright (C) 2017 Netronome Systems, Inc. */
+
+/* Sample program which should always load for testing control paths. */
+int func()
+{
+ return 0;
+}
diff --git a/tools/testing/selftests/bpf/test_offload.py b/tools/testing/selftests/bpf/test_offload.py
new file mode 100755
index 000000000000..ceb18958fc33
--- /dev/null
+++ b/tools/testing/selftests/bpf/test_offload.py
@@ -0,0 +1,634 @@
+#!/usr/bin/python3
+
+# Copyright (C) 2015-2017 Netronome Systems, Inc.
+#
+# This software is dual licensed under the GNU General License Version 2,
+# June 1991 as shown in the file COPYING in the top-level directory of this
+# source tree or the BSD 2-Clause License provided below. You have the
+# option to license this software under the complete terms of either license.
+#
+# The BSD 2-Clause License:
+#
+# Redistribution and use in source and binary forms, with or
+# without modification, are permitted provided that the following
+# conditions are met:
+#
+# 1. Redistributions of source code must retain the above
+# copyright notice, this list of conditions and the following
+# disclaimer.
+#
+# 2. 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.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+# EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+# NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+# BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+# ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+# CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+# SOFTWARE.
+
+import argparse
+import json
+import os
+import pprint
+import subprocess
+import time
+
+logfile = None
+log_level = 1
+bpf_test_dir = os.path.dirname(os.path.realpath(__file__))
+pp = pprint.PrettyPrinter()
+devs = [] # devices we created for clean up
+files = [] # files to be removed
+
+def skip(cond, msg):
+ if not cond:
+ return
+ print("SKIP: " + msg)
+ os.sys.exit(0)
+
+def fail(cond, msg):
+ if not cond:
+ return
+ print("FAIL: " + msg)
+ os.sys.exit(1)
+
+def start_test(msg):
+ global log_level
+
+ # Reset the log level in case other test was running
+ log_level = 1
+ log(msg, "")
+ log_level = 2
+
+ print(msg)
+
+def log_get_sec(level=0):
+ return "*" * (log_level + level)
+
+def log_level_inc(add=1):
+ global log_level
+ log_level += add
+
+def log_level_dec(sub=1):
+ global log_level
+ log_level -= sub
+
+def log(header, data):
+ """
+ Output to an optional log.
+ """
+ if logfile is None:
+ return
+
+ if not isinstance(data, str):
+ data = pp.pformat(data)
+
+ if len(header):
+ logfile.write("\n" + log_get_sec() + " ")
+ logfile.write(header)
+ if len(header) and len(data.strip()):
+ logfile.write("\n")
+ logfile.write(data)
+
+def cmd(cmd, shell=True, include_stderr=False, background=False, fail=False):
+ """
+ Run a command in subprocess and return tuple of (retval, stdout);
+ optionally return stderr as well as third value.
+ """
+ proc = subprocess.Popen(cmd, shell=shell, stdout=subprocess.PIPE,
+ stderr=subprocess.PIPE)
+ if background:
+ log("BKG " + proc.args, "")
+ return proc
+
+ return cmd_result(proc, include_stderr=include_stderr, fail=fail)
+
+def cmd_result(proc, include_stderr=False, fail=False):
+ stdout, stderr = proc.communicate()
+ stdout = stdout.decode("utf-8")
+ stderr = stderr.decode("utf-8")
+ proc.stdout.close()
+ proc.stderr.close()
+
+ sec = log_get_sec(1)
+ log("CMD " + proc.args, "retcode: %d\n%s stdout:\n%s%s stderr:\n%s" %
+ (proc.returncode, sec, stdout, sec, stderr))
+
+ if proc.returncode != 0 and fail:
+ if len(stderr) > 0 and stderr[-1] == "\n":
+ stderr = stderr[:-1]
+ raise Exception("Command failed: %s\n%s" % (proc.args, stderr))
+
+ if include_stderr:
+ return proc.returncode, stdout, stderr
+ else:
+ return proc.returncode, stdout
+
+def rm(f):
+ cmd("rm -f %s" % (f))
+ if f in files:
+ files.remove(f)
+
+def tool(name, args, flags, JSON=True, fail=True):
+ params = ""
+ if JSON:
+ params += "%s " % (flags["json"])
+
+ ret, out = cmd(name + " " + params + args, fail=fail)
+ if JSON and len(out.strip()) != 0:
+ return ret, json.loads(out)
+ else:
+ return ret, out
+
+def bpftool(args, JSON=True, fail=True):
+ return tool("bpftool", args, {"json":"-p"}, JSON=JSON, fail=fail)
+
+def bpftool_prog_list(expected=None):
+ _, progs = bpftool("prog show", JSON=True, fail=True)
+ if expected is not None:
+ if len(progs) != expected:
+ fail(True, "%d BPF programs loaded, expected %d" %
+ (len(progs), expected))
+ return progs
+
+def ip(args, force=False, JSON=True, fail=True):
+ if force:
+ args = "-force " + args
+ return tool("ip", args, {"json":"-j"}, JSON=JSON, fail=fail)
+
+def tc(args, JSON=True, fail=True):
+ return tool("tc", args, {"json":"-p"}, JSON=JSON, fail=fail)
+
+def ethtool(dev, opt, args, fail=True):
+ return cmd("ethtool %s %s %s" % (opt, dev["ifname"], args), fail=fail)
+
+def bpf_obj(name, sec=".text", path=bpf_test_dir,):
+ return "obj %s sec %s" % (os.path.join(path, name), sec)
+
+def bpf_pinned(name):
+ return "pinned %s" % (name)
+
+def bpf_bytecode(bytecode):
+ return "bytecode %s" % (bytecode)
+
+class DebugfsDir:
+ """
+ Class for accessing DebugFS directories as a dictionary.
+ """
+
+ def __init__(self, path):
+ self.path = path
+ self._dict = self._debugfs_dir_read(path)
+
+ def __len__(self):
+ return len(self._dict.keys())
+
+ def __getitem__(self, key):
+ if type(key) is int:
+ key = list(self._dict.keys())[key]
+ return self._dict[key]
+
+ def __setitem__(self, key, value):
+ log("DebugFS set %s = %s" % (key, value), "")
+ log_level_inc()
+
+ cmd("echo '%s' > %s/%s" % (value, self.path, key))
+ log_level_dec()
+
+ self._dict[key] = value
+
+ def _debugfs_dir_read(self, path):
+ dfs = {}
+
+ log("DebugFS state for %s" % (path), "")
+ log_level_inc(add=2)
+
+ _, out = cmd('ls ' + path)
+ for f in out.split():
+ p = os.path.join(path, f)
+ if os.path.isfile(p):
+ _, out = cmd('cat %s/%s' % (path, f))
+ dfs[f] = out.strip()
+ elif os.path.isdir(p):
+ dfs[f] = DebugfsDir(p)
+ else:
+ raise Exception("%s is neither file nor directory" % (p))
+
+ log_level_dec()
+ log("DebugFS state", dfs)
+ log_level_dec()
+
+ return dfs
+
+class NetdevSim:
+ """
+ Class for netdevsim netdevice and its attributes.
+ """
+
+ def __init__(self):
+ self.dev = self._netdevsim_create()
+ devs.append(self)
+
+ self.dfs_dir = '/sys/kernel/debug/netdevsim/%s' % (self.dev['ifname'])
+ self.dfs_refresh()
+
+ def __getitem__(self, key):
+ return self.dev[key]
+
+ def _netdevsim_create(self):
+ _, old = ip("link show")
+ ip("link add sim%d type netdevsim")
+ _, new = ip("link show")
+
+ for dev in new:
+ f = filter(lambda x: x["ifname"] == dev["ifname"], old)
+ if len(list(f)) == 0:
+ return dev
+
+ raise Exception("failed to create netdevsim device")
+
+ def remove(self):
+ devs.remove(self)
+ ip("link del dev %s" % (self.dev["ifname"]))
+ time.sleep(0.1)
+
+ def dfs_refresh(self):
+ self.dfs = DebugfsDir(self.dfs_dir)
+ return self.dfs
+
+ def dfs_get_bound_progs(self, expected):
+ progs = DebugfsDir(os.path.join(self.dfs_dir, "bpf_bound_progs"))
+ if expected is not None:
+ if len(progs) != expected:
+ fail(True, "%d BPF programs bound, expected %d" %
+ (len(progs), expected))
+ return progs
+
+ def set_mtu(self, mtu, fail=True):
+ return ip("link set dev %s mtu %d" % (self.dev["ifname"], mtu),
+ fail=fail)
+
+ def set_xdp(self, bpf, mode, force=False, fail=True):
+ return ip("link set dev %s xdp%s %s" % (self.dev["ifname"], mode, bpf),
+ force=force, fail=fail)
+
+ def unset_xdp(self, mode, force=False, fail=True):
+ return ip("link set dev %s xdp%s off" % (self.dev["ifname"], mode),
+ force=force, fail=fail)
+
+ def ip_link_show(self, xdp):
+ _, link = ip("link show dev %s" % (self['ifname']))
+ if len(link) > 1:
+ raise Exception("Multiple objects on ip link show")
+ if len(link) < 1:
+ return {}
+ fail(xdp != "xdp" in link,
+ "XDP program not reporting in iplink (reported %s, expected %s)" %
+ ("xdp" in link, xdp))
+ return link[0]
+
+ def tc_add_ingress(self):
+ tc("qdisc add dev %s ingress" % (self['ifname']))
+
+ def tc_del_ingress(self):
+ tc("qdisc del dev %s ingress" % (self['ifname']))
+ time.sleep(0.1)
+
+ def tc_flush_filters(self):
+ self.tc_del_ingress()
+ self.tc_add_ingress()
+
+ def tc_show_ingress(self, expected=None):
+ # No JSON support, oh well...
+ flags = ["skip_sw", "skip_hw", "in_hw"]
+ named = ["protocol", "pref", "chain", "handle", "id", "tag"]
+
+ args = "-s filter show dev %s ingress" % (self['ifname'])
+ _, out = tc(args, JSON=False)
+
+ filters = []
+ lines = out.split('\n')
+ for line in lines:
+ words = line.split()
+ if "handle" not in words:
+ continue
+ fltr = {}
+ for flag in flags:
+ fltr[flag] = flag in words
+ for name in named:
+ try:
+ idx = words.index(name)
+ fltr[name] = words[idx + 1]
+ except ValueError:
+ pass
+ filters.append(fltr)
+
+ if expected is not None:
+ fail(len(filters) != expected,
+ "%d ingress filters loaded, expected %d" %
+ (len(filters), expected))
+ return filters
+
+ def cls_bpf_add_filter(self, bpf, da=False, skip_sw=False, skip_hw=False,
+ fail=True):
+ params = ""
+ if da:
+ params += " da"
+ if skip_sw:
+ params += " skip_sw"
+ if skip_hw:
+ params += " skip_hw"
+ return tc("filter add dev %s ingress bpf %s %s" %
+ (self['ifname'], bpf, params), fail=fail)
+
+ def set_ethtool_tc_offloads(self, enable, fail=True):
+ args = "hw-tc-offload %s" % ("on" if enable else "off")
+ return ethtool(self, "-K", args, fail=fail)
+
+################################################################################
+def clean_up():
+ for dev in devs:
+ dev.remove()
+ for f in files:
+ cmd("rm -f %s" % (f))
+
+def check_no_progs(sim):
+ dfs = sim.dfs_get_bound_progs(expected=0)
+ progs = bpftool_prog_list(expected=0)
+ ingress = sim.tc_show_ingress(expected=0)
+
+def pin_prog(file_name, idx=0):
+ progs = bpftool_prog_list(expected=(idx + 1))
+ prog = progs[idx]
+ bpftool("prog pin id %d %s" % (prog["id"], file_name))
+ files.append(file_name)
+
+ return file_name, bpf_pinned(file_name)
+
+# Parse command line
+parser = argparse.ArgumentParser()
+parser.add_argument("--log", help="output verbose log to given file")
+args = parser.parse_args()
+if args.log:
+ logfile = open(args.log, 'w+')
+ logfile.write("# -*-Org-*-")
+
+# Check permissions
+skip(os.getuid() != 0, "test must be run as root")
+
+# Check tools
+ret, progs = bpftool("prog", fail=False)
+skip(ret != 0, "bpftool not installed")
+# Check no BPF programs are loaded
+skip(len(progs) != 0, "BPF programs already loaded on the system")
+
+# Check netdevsim
+ret, out = cmd("modprobe netdevsim", fail=False)
+skip(ret != 0, "netdevsim module could not be loaded")
+
+# Check debugfs
+_, out = cmd("mount")
+if out.find("/sys/kernel/debug type debugfs") == -1:
+ cmd("mount -t debugfs none /sys/kernel/debug")
+
+# Check samples are compiled
+samples = ["sample_ret0.o"]
+for s in samples:
+ ret, out = cmd("ls %s/%s" % (bpf_test_dir, s), fail=False)
+ skip(ret != 0, "sample %s/%s not found, please compile it" %
+ (bpf_test_dir, s))
+
+try:
+ obj = bpf_obj("sample_ret0.o")
+ bytecode = bpf_bytecode("1,6 0 0 4294967295,")
+
+ sim = NetdevSim()
+ sim.tc_add_ingress()
+
+ start_test("Check TC non-offloaded...")
+ ret, _ = sim.cls_bpf_add_filter(obj, skip_hw=True, fail=False)
+ fail(ret != 0, "Software TC filter did not load")
+
+ start_test("Check TC non-offloaded isn't getting bound...")
+ ret, _ = sim.cls_bpf_add_filter(obj, fail=False)
+ fail(ret != 0, "Software TC filter did not load")
+ sim.dfs_get_bound_progs(expected=0)
+
+ sim.tc_flush_filters()
+
+ start_test("Check TC offloads are off by default...")
+ ret, _ = sim.cls_bpf_add_filter(obj, skip_sw=True, fail=False)
+ fail(ret == 0, "TC filter loaded without enabling TC offloads")
+
+ sim.set_ethtool_tc_offloads(True)
+
+ start_test("Check TC offload by default...")
+ sim.dfs["bpf_tc_no_skip_accept"] = "Y"
+ ret, _ = sim.cls_bpf_add_filter(obj, fail=False)
+ fail(ret != 0, "Software TC filter did not load")
+ sim.dfs_get_bound_progs(expected=0)
+ ingress = sim.tc_show_ingress(expected=1)
+ fltr = ingress[0]
+ fail(not fltr["in_hw"], "Filter not offloaded by default")
+
+ sim.tc_flush_filters()
+
+ start_test("Check cBPF bytecode doesn't offload...")
+ ret, _ = sim.cls_bpf_add_filter(bytecode, skip_sw=True, fail=False)
+ fail(ret == 0, "TC bytecode loaded for offload")
+
+ start_test("Check TC offloads work...")
+ ret, _ = sim.cls_bpf_add_filter(obj, skip_sw=True, fail=False)
+ fail(ret != 0, "TC filter did not load with TC offloads enabled")
+
+ start_test("Check TC offload basics...")
+ dfs = sim.dfs_get_bound_progs(expected=1)
+ progs = bpftool_prog_list(expected=1)
+ ingress = sim.tc_show_ingress(expected=1)
+
+ dprog = dfs[0]
+ prog = progs[0]
+ fltr = ingress[0]
+ fail(fltr["skip_hw"], "TC does reports 'skip_hw' on offloaded filter")
+ fail(not fltr["in_hw"], "TC does not report 'in_hw' for offloaded filter")
+ fail(not fltr["skip_sw"], "TC does not report 'skip_sw' back")
+
+ start_test("Check TC offload is device-bound...")
+ fail(str(prog["id"]) != fltr["id"], "Program IDs don't match")
+ fail(prog["tag"] != fltr["tag"], "Program tags don't match")
+ fail(fltr["id"] != dprog["id"], "Program IDs don't match")
+ fail(dprog["state"] != "xlated", "Offloaded program state not translated")
+ fail(dprog["loaded"] != "Y", "Offloaded program is not loaded")
+
+ start_test("Check disabling TC offloads is rejected while filters installed...")
+ ret, _ = sim.set_ethtool_tc_offloads(False, fail=False)
+ fail(ret == 0, "Driver should refuse to disable TC offloads with filters installed...")
+
+ start_test("Check qdisc removal frees things...")
+ sim.tc_flush_filters()
+ check_no_progs(sim)
+
+ start_test("Check disabling TC offloads is OK without filters...")
+ ret, _ = sim.set_ethtool_tc_offloads(False, fail=False)
+ fail(ret != 0,
+ "Driver refused to disable TC offloads without filters installed...")
+
+ sim.set_ethtool_tc_offloads(True)
+
+ start_test("Check destroying device gets rid of TC filters...")
+ sim.cls_bpf_add_filter(obj, skip_sw=True)
+ sim.remove()
+ progs = bpftool_prog_list(expected=0)
+
+ sim = NetdevSim()
+ sim.set_ethtool_tc_offloads(True)
+
+ start_test("Check destroying device gets rid of XDP...")
+ sim.set_xdp(obj, "offload")
+ sim.remove()
+ progs = bpftool_prog_list(expected=0)
+
+ sim = NetdevSim()
+ sim.set_ethtool_tc_offloads(True)
+
+ start_test("Check XDP prog reporting...")
+ sim.set_xdp(obj, "drv")
+ ipl = sim.ip_link_show(xdp=True)
+ progs = bpftool_prog_list(expected=1)
+ fail(ipl["xdp"]["prog"]["id"] != progs[0]["id"],
+ "Loaded program has wrong ID")
+
+ start_test("Check XDP prog replace without force...")
+ ret, _ = sim.set_xdp(obj, "drv", fail=False)
+ fail(ret == 0, "Replaced XDP program without -force")
+
+ start_test("Check XDP prog replace with force...")
+ ret, _ = sim.set_xdp(obj, "drv", force=True, fail=False)
+ fail(ret != 0, "Could not replace XDP program with -force")
+ ipl = sim.ip_link_show(xdp=True)
+ progs = bpftool_prog_list(expected=1)
+ fail(ipl["xdp"]["prog"]["id"] != progs[0]["id"],
+ "Loaded program has wrong ID")
+
+ start_test("Check XDP prog replace with bad flags...")
+ ret, _ = sim.set_xdp(obj, "offload", force=True, fail=False)
+ fail(ret == 0, "Replaced XDP program with a program in different mode")
+ ret, _ = sim.set_xdp(obj, "", force=True, fail=False)
+ fail(ret == 0, "Replaced XDP program with a program in different mode")
+
+ start_test("Check XDP prog remove with bad flags...")
+ ret, _ = sim.unset_xdp("offload", force=True, fail=False)
+ fail(ret == 0, "Removed program with a bad mode mode")
+ ret, _ = sim.unset_xdp("", force=True, fail=False)
+ fail(ret == 0, "Removed program with a bad mode mode")
+
+ start_test("Check MTU restrictions...")
+ ret, _ = sim.set_mtu(9000, fail=False)
+ fail(ret == 0,
+ "Driver should refuse to increase MTU to 9000 with XDP loaded...")
+ sim.unset_xdp("drv")
+ sim.set_mtu(9000)
+ ret, _ = sim.set_xdp(obj, "drv", fail=False)
+ fail(ret == 0, "Driver should refuse to load program with MTU of 9000...")
+ sim.set_mtu(1500)
+
+ start_test("Check XDP offload...")
+ sim.set_xdp(obj, "offload")
+ ipl = sim.ip_link_show(xdp=True)
+ link_xdp = ipl["xdp"]["prog"]
+ progs = bpftool_prog_list(expected=1)
+ prog = progs[0]
+ fail(link_xdp["id"] != prog["id"], "Loaded program has wrong ID")
+
+ start_test("Check XDP offload is device bound...")
+ dfs = sim.dfs_get_bound_progs(expected=1)
+ dprog = dfs[0]
+
+ fail(prog["id"] != link_xdp["id"], "Program IDs don't match")
+ fail(prog["tag"] != link_xdp["tag"], "Program tags don't match")
+ fail(str(link_xdp["id"]) != dprog["id"], "Program IDs don't match")
+ fail(dprog["state"] != "xlated", "Offloaded program state not translated")
+ fail(dprog["loaded"] != "Y", "Offloaded program is not loaded")
+
+ start_test("Check removing XDP program many times...")
+ sim.unset_xdp("offload")
+ sim.unset_xdp("offload")
+ sim.unset_xdp("drv")
+ sim.unset_xdp("drv")
+ sim.unset_xdp("")
+ sim.unset_xdp("")
+
+ start_test("Check attempt to use a program for a wrong device...")
+ sim2 = NetdevSim()
+ sim2.set_xdp(obj, "offload")
+ pin_file, pinned = pin_prog("/sys/fs/bpf/tmp")
+
+ ret, _ = sim.set_xdp(pinned, "offload", fail=False)
+ fail(ret == 0, "Pinned program loaded for a different device accepted")
+ sim2.remove()
+ ret, _ = sim.set_xdp(pinned, "offload", fail=False)
+ fail(ret == 0, "Pinned program loaded for a removed device accepted")
+ rm(pin_file)
+
+ start_test("Check mixing of TC and XDP...")
+ sim.tc_add_ingress()
+ sim.set_xdp(obj, "offload")
+ ret, _ = sim.cls_bpf_add_filter(obj, skip_sw=True, fail=False)
+ fail(ret == 0, "Loading TC when XDP active should fail")
+ sim.unset_xdp("offload")
+
+ sim.cls_bpf_add_filter(obj, skip_sw=True)
+ ret, _ = sim.set_xdp(obj, "offload", fail=False)
+ fail(ret == 0, "Loading XDP when TC active should fail")
+
+ start_test("Check binding TC from pinned...")
+ pin_file, pinned = pin_prog("/sys/fs/bpf/tmp")
+ sim.tc_flush_filters()
+ sim.cls_bpf_add_filter(pinned, da=True, skip_sw=True)
+ sim.tc_flush_filters()
+
+ start_test("Check binding XDP from pinned...")
+ sim.set_xdp(obj, "offload")
+ pin_file, pinned = pin_prog("/sys/fs/bpf/tmp2", idx=1)
+
+ sim.set_xdp(pinned, "offload", force=True)
+ sim.unset_xdp("offload")
+ sim.set_xdp(pinned, "offload", force=True)
+ sim.unset_xdp("offload")
+
+ start_test("Check offload of wrong type fails...")
+ ret, _ = sim.cls_bpf_add_filter(pinned, da=True, skip_sw=True, fail=False)
+ fail(ret == 0, "Managed to attach XDP program to TC")
+
+ start_test("Check asking for TC offload of two filters...")
+ sim.cls_bpf_add_filter(obj, da=True, skip_sw=True)
+ sim.cls_bpf_add_filter(obj, da=True, skip_sw=True)
+ # The above will trigger a splat until TC cls_bpf drivers are fixed
+
+ sim.tc_flush_filters()
+
+ start_test("Check that netdev removal waits for translation...")
+ sim.dfs["bpf_bind_verifier_delay"] = 1000 # msec
+ start = time.time()
+ cmd_line = "tc filter add dev %s ingress bpf %s da skip_sw" % \
+ (sim['ifname'], obj)
+ tc_proc = cmd(cmd_line, background=True, fail=False)
+ time.sleep(0.2)
+ sim.remove()
+ ret, _ = cmd_result(tc_proc, fail=False)
+ end = time.time()
+ time_diff = end - start
+ log("Time", "start: %s\nend: %s\ndiff: %s" % (start, end, time_diff))
+
+ fail(ret == 0, "Managed to load TC filter on a unregistering device")
+ fail(time_diff < 1.0, "Removal process took %s, expected %s" %
+ (time_diff, 1.0))
+
+finally:
+ log_level = 1
+ clean_up()
--
2.14.1
^ permalink raw reply related [flat|nested] 22+ messages in thread* Re: [RFC net-next 0/6] xdp: make stack perform remove and tests
2017-11-24 2:36 [RFC net-next 0/6] xdp: make stack perform remove and tests Jakub Kicinski
` (5 preceding siblings ...)
2017-11-24 2:36 ` [RFC net-next 6/6] selftests/bpf: add offload test based on netdevsim Jakub Kicinski
@ 2017-11-24 7:45 ` Jiri Pirko
2017-11-24 8:02 ` Jakub Kicinski
2017-11-26 1:02 ` Alexei Starovoitov
7 siblings, 1 reply; 22+ messages in thread
From: Jiri Pirko @ 2017-11-24 7:45 UTC (permalink / raw)
To: Jakub Kicinski; +Cc: netdev, oss-drivers, daniel, alexei.starovoitov
Fri, Nov 24, 2017 at 03:36:07AM CET, jakub.kicinski@netronome.com wrote:
>Hi!
>
>The purpose of this series is to add a software model of BPF offloads
>to make it easier for everyone to test them and make some of the more
>arcane rules and assumptions more clear.
>
>The series starts with 3 patches aiming to make XDP handling in the
>drivers less error prone. Currently driver authors have to remember
>to free XDP programs if XDP is active during unregister. With this
>series the core will disable XDP on its own. It will take place
>after close, drivers are not expected to perform reconfiguration
>when disabling XDP on a downed device.
>
>Next two patches add the software netdev driver. Last but not least
I wonder if for this it is needed to split the driver into multiple
files. I think that a single file would be better as I don't expect the
driver would get big.
>there is a python test which exercises all the corner cases which
>came to my mind.
>
>Test needs to be run as root. It will print basic information to
>stdout, but can also create a more detailed log of all commands
>when --log option is passed. Log is in Emacs Org-mode format.
>
> ./tools/testing/selftests/bpf/test_offload.py --log /tmp/log
>
>Something I'm still battling with, and would appreciate help of
>wiser people is that occasionally during the test something makes
>the refcount of init_net drop to 0 :S I tried to create a simple
>reproducer, but seems like just running the script in the loop is
>the easiest way to go... Could it have something to do with the
>recent TC work? The driver is pretty simple and never touches
I don't see how...
>ref counts. The only slightly unusual thing is that the BPF code
>sleeps for a bit on remove in the netdev notifier.
>
>
>Jakub Kicinski (6):
> net: xdp: avoid output parameters when querying XDP prog
> net: xdp: report flags program was installed with on query
> net: xdp: make the stack take care of the tear down
> netdevsim: add software driver for testing offloads
> netdevsim: add bpf offload support
> selftests/bpf: add offload test based on netdevsim
Patchset looks fine to me.
Thanks for this!
>
> MAINTAINERS | 5 +
> drivers/net/Kconfig | 10 +
> drivers/net/Makefile | 1 +
> drivers/net/ethernet/broadcom/bnxt/bnxt.c | 2 -
> drivers/net/ethernet/mellanox/mlx5/core/en_main.c | 3 -
> drivers/net/ethernet/netronome/nfp/bpf/main.c | 7 -
> .../net/ethernet/netronome/nfp/nfp_net_common.c | 4 +-
> drivers/net/ethernet/qlogic/qede/qede_main.c | 4 -
> drivers/net/netdevsim/Makefile | 7 +
> drivers/net/netdevsim/bpf.c | 391 +++++++++++++
> drivers/net/netdevsim/netdev.c | 241 ++++++++
> drivers/net/netdevsim/netdevsim.h | 84 +++
> drivers/net/tun.c | 4 -
> include/linux/netdevice.h | 5 +-
> net/core/dev.c | 42 +-
> net/core/rtnetlink.c | 6 +-
> tools/testing/selftests/bpf/Makefile | 5 +-
> tools/testing/selftests/bpf/sample_ret0.c | 8 +
> tools/testing/selftests/bpf/test_offload.py | 634 +++++++++++++++++++++
> 19 files changed, 1426 insertions(+), 37 deletions(-)
> create mode 100644 drivers/net/netdevsim/Makefile
> create mode 100644 drivers/net/netdevsim/bpf.c
> create mode 100644 drivers/net/netdevsim/netdev.c
> create mode 100644 drivers/net/netdevsim/netdevsim.h
> create mode 100644 tools/testing/selftests/bpf/sample_ret0.c
> create mode 100755 tools/testing/selftests/bpf/test_offload.py
>
>--
>2.14.1
>
^ permalink raw reply [flat|nested] 22+ messages in thread* Re: [RFC net-next 0/6] xdp: make stack perform remove and tests
2017-11-24 7:45 ` [RFC net-next 0/6] xdp: make stack perform remove and tests Jiri Pirko
@ 2017-11-24 8:02 ` Jakub Kicinski
2017-11-25 4:55 ` Jakub Kicinski
0 siblings, 1 reply; 22+ messages in thread
From: Jakub Kicinski @ 2017-11-24 8:02 UTC (permalink / raw)
To: Jiri Pirko
Cc: Linux Netdev List, oss-drivers, Daniel Borkmann, Alexei Starovoitov
On Thu, Nov 23, 2017 at 11:45 PM, Jiri Pirko <jiri@resnulli.us> wrote:
> Fri, Nov 24, 2017 at 03:36:07AM CET, jakub.kicinski@netronome.com wrote:
>>Hi!
>>
>>The purpose of this series is to add a software model of BPF offloads
>>to make it easier for everyone to test them and make some of the more
>>arcane rules and assumptions more clear.
>>
>>The series starts with 3 patches aiming to make XDP handling in the
>>drivers less error prone. Currently driver authors have to remember
>>to free XDP programs if XDP is active during unregister. With this
>>series the core will disable XDP on its own. It will take place
>>after close, drivers are not expected to perform reconfiguration
>>when disabling XDP on a downed device.
>>
>>Next two patches add the software netdev driver. Last but not least
>
> I wonder if for this it is needed to split the driver into multiple
> files. I think that a single file would be better as I don't expect the
> driver would get big.
I was hoping other offloads will be added to their separate files, to
make it easier for people to find "all code relevant when implementing
X" easier.
Sort of related to your comment on the license, I'm hoping to be able
to use SPDX one-line header to lower the overhead of many files. Has
anyone managed to get an OK to do that?
>>there is a python test which exercises all the corner cases which
>>came to my mind.
>>
>>Test needs to be run as root. It will print basic information to
>>stdout, but can also create a more detailed log of all commands
>>when --log option is passed. Log is in Emacs Org-mode format.
>>
>> ./tools/testing/selftests/bpf/test_offload.py --log /tmp/log
>>
>>Something I'm still battling with, and would appreciate help of
>>wiser people is that occasionally during the test something makes
>>the refcount of init_net drop to 0 :S I tried to create a simple
>>reproducer, but seems like just running the script in the loop is
>>the easiest way to go... Could it have something to do with the
>>recent TC work? The driver is pretty simple and never touches
>
> I don't see how...
To be clear I meant the changes made to destruction of filters, not
your work. The BPF code doesn't touch ref counts and cls exts do seem
to hold a ref on the net... but perhaps that's just pointing the
finger unnecessarily :) I will try to investigate again tomorrow.
>>ref counts. The only slightly unusual thing is that the BPF code
>>sleeps for a bit on remove in the netdev notifier.
>>
>>
>>Jakub Kicinski (6):
>> net: xdp: avoid output parameters when querying XDP prog
>> net: xdp: report flags program was installed with on query
>> net: xdp: make the stack take care of the tear down
>> netdevsim: add software driver for testing offloads
>> netdevsim: add bpf offload support
>> selftests/bpf: add offload test based on netdevsim
>
> Patchset looks fine to me.
> Thanks for this!
Thanks!
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [RFC net-next 0/6] xdp: make stack perform remove and tests
2017-11-24 8:02 ` Jakub Kicinski
@ 2017-11-25 4:55 ` Jakub Kicinski
0 siblings, 0 replies; 22+ messages in thread
From: Jakub Kicinski @ 2017-11-25 4:55 UTC (permalink / raw)
To: Jiri Pirko
Cc: Linux Netdev List, oss-drivers, Daniel Borkmann, Alexei Starovoitov
On Fri, 24 Nov 2017 00:02:32 -0800, Jakub Kicinski wrote:
> >>Something I'm still battling with, and would appreciate help of
> >>wiser people is that occasionally during the test something makes
> >>the refcount of init_net drop to 0 :S I tried to create a simple
> >>reproducer, but seems like just running the script in the loop is
> >>the easiest way to go... Could it have something to do with the
> >>recent TC work? The driver is pretty simple and never touches
> >
> > I don't see how...
>
> To be clear I meant the changes made to destruction of filters, not
> your work. The BPF code doesn't touch ref counts and cls exts do seem
> to hold a ref on the net... but perhaps that's just pointing the
> finger unnecessarily :) I will try to investigate again tomorrow.
Looks like I was lazy when adding the offload and just called
__cls_bpf_delete_prog() instead of extending the error path.
Cong missed this extra call in aae2c35ec892 ("cls_bpf: use
tcf_exts_get_net() before call_rcu()"). We need something like
this:
diff --git a/net/sched/cls_bpf.c b/net/sched/cls_bpf.c
index a9f3e317055c..40d4289aea28 100644
--- a/net/sched/cls_bpf.c
+++ b/net/sched/cls_bpf.c
@@ -514,12 +514,8 @@ static int cls_bpf_change(struct net *net, struct sk_buff *in_skb,
goto errout_idr;
ret = cls_bpf_offload(tp, prog, oldprog);
- if (ret) {
- if (!oldprog)
- idr_remove_ext(&head->handle_idr, prog->handle);
- __cls_bpf_delete_prog(prog);
- return ret;
- }
+ if (ret)
+ goto errout_parms;
if (!tc_in_hw(prog->gen_flags))
prog->gen_flags |= TCA_CLS_FLAGS_NOT_IN_HW;
@@ -537,6 +533,13 @@ static int cls_bpf_change(struct net *net, struct sk_buff *in_skb,
*arg = prog;
return 0;
+errout_parms:
+ if (cls_bpf_is_ebpf(prog))
+ bpf_prog_put(prog->filter);
+ else
+ bpf_prog_destroy(prog->filter);
+ kfree(prog->bpf_name);
+ kfree(prog->bpf_ops);
errout_idr:
if (!oldprog)
idr_remove_ext(&head->handle_idr, prog->handle);
^ permalink raw reply related [flat|nested] 22+ messages in thread
* Re: [RFC net-next 0/6] xdp: make stack perform remove and tests
2017-11-24 2:36 [RFC net-next 0/6] xdp: make stack perform remove and tests Jakub Kicinski
` (6 preceding siblings ...)
2017-11-24 7:45 ` [RFC net-next 0/6] xdp: make stack perform remove and tests Jiri Pirko
@ 2017-11-26 1:02 ` Alexei Starovoitov
2017-11-26 1:43 ` Jakub Kicinski
7 siblings, 1 reply; 22+ messages in thread
From: Alexei Starovoitov @ 2017-11-26 1:02 UTC (permalink / raw)
To: Jakub Kicinski; +Cc: netdev, oss-drivers, daniel, jiri
On Thu, Nov 23, 2017 at 06:36:07PM -0800, Jakub Kicinski wrote:
> Hi!
>
> The purpose of this series is to add a software model of BPF offloads
> to make it easier for everyone to test them and make some of the more
> arcane rules and assumptions more clear.
>
> The series starts with 3 patches aiming to make XDP handling in the
> drivers less error prone. Currently driver authors have to remember
> to free XDP programs if XDP is active during unregister. With this
> series the core will disable XDP on its own. It will take place
> after close, drivers are not expected to perform reconfiguration
> when disabling XDP on a downed device.
>
> Next two patches add the software netdev driver. Last but not least
> there is a python test which exercises all the corner cases which
> came to my mind.
>
> Test needs to be run as root. It will print basic information to
> stdout, but can also create a more detailed log of all commands
> when --log option is passed. Log is in Emacs Org-mode format.
>
> ./tools/testing/selftests/bpf/test_offload.py --log /tmp/log
>
> Something I'm still battling with, and would appreciate help of
> wiser people is that occasionally during the test something makes
> the refcount of init_net drop to 0 :S I tried to create a simple
> reproducer, but seems like just running the script in the loop is
> the easiest way to go... Could it have something to do with the
> recent TC work? The driver is pretty simple and never touches
> ref counts. The only slightly unusual thing is that the BPF code
> sleeps for a bit on remove in the netdev notifier.
I like the direction.
Patch 3 in particular with auto prog_put is a great idea!
Patch 4 - if you want to do dual gpl+bsd. It's ok and really your call as an author.
Patch 5 - I'm only not excited about debugfs. Can you do the same with tracepoints?
Patch 6 - python 3 is a requirement? or can work with 2.7 ? I don't mind if it's 3+ only.
Just trying to understand the dependencies.
and thanks for adding it to selftests.
^ permalink raw reply [flat|nested] 22+ messages in thread* Re: [RFC net-next 0/6] xdp: make stack perform remove and tests
2017-11-26 1:02 ` Alexei Starovoitov
@ 2017-11-26 1:43 ` Jakub Kicinski
2017-11-26 7:43 ` Jiri Pirko
0 siblings, 1 reply; 22+ messages in thread
From: Jakub Kicinski @ 2017-11-26 1:43 UTC (permalink / raw)
To: Alexei Starovoitov; +Cc: netdev, oss-drivers, daniel, jiri
On Sat, 25 Nov 2017 17:02:09 -0800, Alexei Starovoitov wrote:
> On Thu, Nov 23, 2017 at 06:36:07PM -0800, Jakub Kicinski wrote:
> > Hi!
> >
> > The purpose of this series is to add a software model of BPF offloads
> > to make it easier for everyone to test them and make some of the more
> > arcane rules and assumptions more clear.
> >
> > The series starts with 3 patches aiming to make XDP handling in the
> > drivers less error prone. Currently driver authors have to remember
> > to free XDP programs if XDP is active during unregister. With this
> > series the core will disable XDP on its own. It will take place
> > after close, drivers are not expected to perform reconfiguration
> > when disabling XDP on a downed device.
> >
> > Next two patches add the software netdev driver. Last but not least
> > there is a python test which exercises all the corner cases which
> > came to my mind.
> >
> > Test needs to be run as root. It will print basic information to
> > stdout, but can also create a more detailed log of all commands
> > when --log option is passed. Log is in Emacs Org-mode format.
> >
> > ./tools/testing/selftests/bpf/test_offload.py --log /tmp/log
> >
> > Something I'm still battling with, and would appreciate help of
> > wiser people is that occasionally during the test something makes
> > the refcount of init_net drop to 0 :S I tried to create a simple
> > reproducer, but seems like just running the script in the loop is
> > the easiest way to go... Could it have something to do with the
> > recent TC work? The driver is pretty simple and never touches
> > ref counts. The only slightly unusual thing is that the BPF code
> > sleeps for a bit on remove in the netdev notifier.
>
> I like the direction.
> Patch 3 in particular with auto prog_put is a great idea!
> Patch 4 - if you want to do dual gpl+bsd. It's ok and really your call as an author.
> Patch 5 - I'm only not excited about debugfs. Can you do the same with tracepoints?
I would prefer to stick to DebugFS, if you don't mind. Some of the
knobs are writeable. I look at DebugFS as a back door. We throw
standard API commands at the device and want to inspect its internal
state. And cross check the standard reporting APIs. As such DebugFS
is completely inappropriate but for API-test-drivers.
On that topic - is there a way to trigger and wait for an rcu barrier
from user space? Quentin pointed out that selftest is shaky unless run
with RCU debug enabled. To wait for deletion of programs I need a
barrier :S
> Patch 6 - python 3 is a requirement? or can work with 2.7 ? I don't mind if it's 3+ only.
> Just trying to understand the dependencies.
Am a python noob so I defaulted to version 3 and proceeded to look
things I needed up in the python 3 docs. Which perhaps resulted in
using more python3 only features than necessary...
Upon a quick inspection, the only python3 feature I seem to be using
is the Popen.args member.
> and thanks for adding it to selftests.
Thanks for the comments!
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [RFC net-next 0/6] xdp: make stack perform remove and tests
2017-11-26 1:43 ` Jakub Kicinski
@ 2017-11-26 7:43 ` Jiri Pirko
0 siblings, 0 replies; 22+ messages in thread
From: Jiri Pirko @ 2017-11-26 7:43 UTC (permalink / raw)
To: Jakub Kicinski; +Cc: Alexei Starovoitov, netdev, oss-drivers, daniel
Sun, Nov 26, 2017 at 02:43:04AM CET, jakub.kicinski@netronome.com wrote:
>On Sat, 25 Nov 2017 17:02:09 -0800, Alexei Starovoitov wrote:
>> On Thu, Nov 23, 2017 at 06:36:07PM -0800, Jakub Kicinski wrote:
>> > Hi!
>> >
>> > The purpose of this series is to add a software model of BPF offloads
>> > to make it easier for everyone to test them and make some of the more
>> > arcane rules and assumptions more clear.
>> >
>> > The series starts with 3 patches aiming to make XDP handling in the
>> > drivers less error prone. Currently driver authors have to remember
>> > to free XDP programs if XDP is active during unregister. With this
>> > series the core will disable XDP on its own. It will take place
>> > after close, drivers are not expected to perform reconfiguration
>> > when disabling XDP on a downed device.
>> >
>> > Next two patches add the software netdev driver. Last but not least
>> > there is a python test which exercises all the corner cases which
>> > came to my mind.
>> >
>> > Test needs to be run as root. It will print basic information to
>> > stdout, but can also create a more detailed log of all commands
>> > when --log option is passed. Log is in Emacs Org-mode format.
>> >
>> > ./tools/testing/selftests/bpf/test_offload.py --log /tmp/log
>> >
>> > Something I'm still battling with, and would appreciate help of
>> > wiser people is that occasionally during the test something makes
>> > the refcount of init_net drop to 0 :S I tried to create a simple
>> > reproducer, but seems like just running the script in the loop is
>> > the easiest way to go... Could it have something to do with the
>> > recent TC work? The driver is pretty simple and never touches
>> > ref counts. The only slightly unusual thing is that the BPF code
>> > sleeps for a bit on remove in the netdev notifier.
>>
>> I like the direction.
>> Patch 3 in particular with auto prog_put is a great idea!
>> Patch 4 - if you want to do dual gpl+bsd. It's ok and really your call as an author.
Sure. I was just curious if it is really necessary. That's all.
>> Patch 5 - I'm only not excited about debugfs. Can you do the same with tracepoints?
>
>I would prefer to stick to DebugFS, if you don't mind. Some of the
>knobs are writeable. I look at DebugFS as a back door. We throw
>standard API commands at the device and want to inspect its internal
>state. And cross check the standard reporting APIs. As such DebugFS
>is completely inappropriate but for API-test-drivers.
I agree. This is one of few exceptions where debugfs is good to be used.
>
>On that topic - is there a way to trigger and wait for an rcu barrier
>from user space? Quentin pointed out that selftest is shaky unless run
>with RCU debug enabled. To wait for deletion of programs I need a
>barrier :S
>
>> Patch 6 - python 3 is a requirement? or can work with 2.7 ? I don't mind if it's 3+ only.
>> Just trying to understand the dependencies.
>
>Am a python noob so I defaulted to version 3 and proceeded to look
>things I needed up in the python 3 docs. Which perhaps resulted in
>using more python3 only features than necessary...
>
>Upon a quick inspection, the only python3 feature I seem to be using
>is the Popen.args member.
>
>> and thanks for adding it to selftests.
>
>Thanks for the comments!
^ permalink raw reply [flat|nested] 22+ messages in thread