Expose a new per-queue nest attribute, xsk, which will be present for
queues that are being used for AF_XDP. If the queue is not being used for
AF_XDP, the nest will not be present.
In the future, this attribute can be extended to include more data about
XSK as it is needed.
Signed-off-by: Joe Damato <jdamato@fastly.com>
Suggested-by: Jakub Kicinski <kuba@kernel.org>
---
v2:
- Patch adjusted to include an attribute, xsk, which is an empty nest
and exposed for queues which have a pool.
Documentation/netlink/specs/netdev.yaml | 13 ++++++++++++-
include/uapi/linux/netdev.h | 6 ++++++
net/core/netdev-genl.c | 11 +++++++++++
tools/include/uapi/linux/netdev.h | 6 ++++++
4 files changed, 35 insertions(+), 1 deletion(-)
diff --git a/Documentation/netlink/specs/netdev.yaml b/Documentation/netlink/specs/netdev.yaml
index cbb544bd6c84..4c3eda5ba754 100644
--- a/Documentation/netlink/specs/netdev.yaml
+++ b/Documentation/netlink/specs/netdev.yaml
@@ -268,6 +268,9 @@ attribute-sets:
doc: The timeout, in nanoseconds, of how long to suspend irq
processing, if event polling finds events
type: uint
+ -
+ name: xsk-info
+ attributes: []
-
name: queue
attributes:
@@ -286,6 +289,9 @@ attribute-sets:
-
name: type
doc: Queue type as rx, tx. Each queue type defines a separate ID space.
+ XDP TX queues allocated in the kernel are not linked to NAPIs and
+ thus not listed. AF_XDP queues will have more information set in
+ the xsk attribute.
type: u32
enum: queue-type
-
@@ -296,7 +302,11 @@ attribute-sets:
name: dmabuf
doc: ID of the dmabuf attached to this queue, if any.
type: u32
-
+ -
+ name: xsk
+ doc: XSK information for this queue, if any.
+ type: nest
+ nested-attributes: xsk-info
-
name: qstats
doc: |
@@ -637,6 +647,7 @@ operations:
- napi-id
- ifindex
- dmabuf
+ - xsk
dump:
request:
attributes:
diff --git a/include/uapi/linux/netdev.h b/include/uapi/linux/netdev.h
index e4be227d3ad6..46bdb0b67a39 100644
--- a/include/uapi/linux/netdev.h
+++ b/include/uapi/linux/netdev.h
@@ -130,12 +130,18 @@ enum {
NETDEV_A_NAPI_MAX = (__NETDEV_A_NAPI_MAX - 1)
};
+enum {
+ __NETDEV_A_XSK_INFO_MAX,
+ NETDEV_A_XSK_INFO_MAX = (__NETDEV_A_XSK_INFO_MAX - 1)
+};
+
enum {
NETDEV_A_QUEUE_ID = 1,
NETDEV_A_QUEUE_IFINDEX,
NETDEV_A_QUEUE_TYPE,
NETDEV_A_QUEUE_NAPI_ID,
NETDEV_A_QUEUE_DMABUF,
+ NETDEV_A_QUEUE_XSK,
__NETDEV_A_QUEUE_MAX,
NETDEV_A_QUEUE_MAX = (__NETDEV_A_QUEUE_MAX - 1)
diff --git a/net/core/netdev-genl.c b/net/core/netdev-genl.c
index 715f85c6b62e..efaccfb6438e 100644
--- a/net/core/netdev-genl.c
+++ b/net/core/netdev-genl.c
@@ -371,6 +371,7 @@ netdev_nl_queue_fill_one(struct sk_buff *rsp, struct net_device *netdev,
struct net_devmem_dmabuf_binding *binding;
struct netdev_rx_queue *rxq;
struct netdev_queue *txq;
+ struct nlattr *nest;
void *hdr;
hdr = genlmsg_iput(rsp, info);
@@ -394,12 +395,22 @@ netdev_nl_queue_fill_one(struct sk_buff *rsp, struct net_device *netdev,
nla_put_u32(rsp, NETDEV_A_QUEUE_DMABUF, binding->id))
goto nla_put_failure;
+ if (rxq->pool) {
+ nest = nla_nest_start(rsp, NETDEV_A_QUEUE_XSK);
+ nla_nest_end(rsp, nest);
+ }
+
break;
case NETDEV_QUEUE_TYPE_TX:
txq = netdev_get_tx_queue(netdev, q_idx);
if (txq->napi && nla_put_u32(rsp, NETDEV_A_QUEUE_NAPI_ID,
txq->napi->napi_id))
goto nla_put_failure;
+
+ if (txq->pool) {
+ nest = nla_nest_start(rsp, NETDEV_A_QUEUE_XSK);
+ nla_nest_end(rsp, nest);
+ }
}
genlmsg_end(rsp, hdr);
diff --git a/tools/include/uapi/linux/netdev.h b/tools/include/uapi/linux/netdev.h
index e4be227d3ad6..46bdb0b67a39 100644
--- a/tools/include/uapi/linux/netdev.h
+++ b/tools/include/uapi/linux/netdev.h
@@ -130,12 +130,18 @@ enum {
NETDEV_A_NAPI_MAX = (__NETDEV_A_NAPI_MAX - 1)
};
+enum {
+ __NETDEV_A_XSK_INFO_MAX,
+ NETDEV_A_XSK_INFO_MAX = (__NETDEV_A_XSK_INFO_MAX - 1)
+};
+
enum {
NETDEV_A_QUEUE_ID = 1,
NETDEV_A_QUEUE_IFINDEX,
NETDEV_A_QUEUE_TYPE,
NETDEV_A_QUEUE_NAPI_ID,
NETDEV_A_QUEUE_DMABUF,
+ NETDEV_A_QUEUE_XSK,
__NETDEV_A_QUEUE_MAX,
NETDEV_A_QUEUE_MAX = (__NETDEV_A_QUEUE_MAX - 1)
--
2.43.0Hi Joe,
kernel test robot noticed the following build errors:
[auto build test ERROR on c2933b2befe25309f4c5cfbea0ca80909735fd76]
url: https://github.com/intel-lab-lkp/linux/commits/Joe-Damato/netdev-genl-Add-an-XSK-attribute-to-queues/20250205-031236
base: c2933b2befe25309f4c5cfbea0ca80909735fd76
patch link: https://lore.kernel.org/r/20250204191108.161046-2-jdamato%40fastly.com
patch subject: [PATCH net-next v3 1/2] netdev-genl: Add an XSK attribute to queues
config: arm-defconfig (https://download.01.org/0day-ci/archive/20250207/202502071452.B85Lw7aV-lkp@intel.com/config)
compiler: clang version 21.0.0git (https://github.com/llvm/llvm-project ee3bccab34f57387bdf33853cdd5f214fef349a2)
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20250207/202502071452.B85Lw7aV-lkp@intel.com/reproduce)
If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202502071452.B85Lw7aV-lkp@intel.com/
All errors (new ones prefixed by >>):
>> net/core/netdev-genl.c:398:12: error: no member named 'pool' in 'struct netdev_rx_queue'
398 | if (rxq->pool) {
| ~~~ ^
>> net/core/netdev-genl.c:410:12: error: no member named 'pool' in 'struct netdev_queue'
410 | if (txq->pool) {
| ~~~ ^
2 errors generated.
vim +398 net/core/netdev-genl.c
366
367 static int
368 netdev_nl_queue_fill_one(struct sk_buff *rsp, struct net_device *netdev,
369 u32 q_idx, u32 q_type, const struct genl_info *info)
370 {
371 struct net_devmem_dmabuf_binding *binding;
372 struct netdev_rx_queue *rxq;
373 struct netdev_queue *txq;
374 struct nlattr *nest;
375 void *hdr;
376
377 hdr = genlmsg_iput(rsp, info);
378 if (!hdr)
379 return -EMSGSIZE;
380
381 if (nla_put_u32(rsp, NETDEV_A_QUEUE_ID, q_idx) ||
382 nla_put_u32(rsp, NETDEV_A_QUEUE_TYPE, q_type) ||
383 nla_put_u32(rsp, NETDEV_A_QUEUE_IFINDEX, netdev->ifindex))
384 goto nla_put_failure;
385
386 switch (q_type) {
387 case NETDEV_QUEUE_TYPE_RX:
388 rxq = __netif_get_rx_queue(netdev, q_idx);
389 if (rxq->napi && nla_put_u32(rsp, NETDEV_A_QUEUE_NAPI_ID,
390 rxq->napi->napi_id))
391 goto nla_put_failure;
392
393 binding = rxq->mp_params.mp_priv;
394 if (binding &&
395 nla_put_u32(rsp, NETDEV_A_QUEUE_DMABUF, binding->id))
396 goto nla_put_failure;
397
> 398 if (rxq->pool) {
399 nest = nla_nest_start(rsp, NETDEV_A_QUEUE_XSK);
400 nla_nest_end(rsp, nest);
401 }
402
403 break;
404 case NETDEV_QUEUE_TYPE_TX:
405 txq = netdev_get_tx_queue(netdev, q_idx);
406 if (txq->napi && nla_put_u32(rsp, NETDEV_A_QUEUE_NAPI_ID,
407 txq->napi->napi_id))
408 goto nla_put_failure;
409
> 410 if (txq->pool) {
411 nest = nla_nest_start(rsp, NETDEV_A_QUEUE_XSK);
412 nla_nest_end(rsp, nest);
413 }
414 }
415
416 genlmsg_end(rsp, hdr);
417
418 return 0;
419
420 nla_put_failure:
421 genlmsg_cancel(rsp, hdr);
422 return -EMSGSIZE;
423 }
424
--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
On Tue, 4 Feb 2025 19:10:47 +0000 Joe Damato wrote:
> + if (rxq->pool) {
> + nest = nla_nest_start(rsp, NETDEV_A_QUEUE_XSK);
> + nla_nest_end(rsp, nest);
> + }
nla_nest_start() can fail, you gotta nul-check the return value.
You could possibly add an nla_put_empty_nest() helper in netlink.h
to make this less awkward? I think the iouring guys had the same bug
On Thu, Feb 06, 2025 at 04:57:46PM -0800, Jakub Kicinski wrote:
> On Tue, 4 Feb 2025 19:10:47 +0000 Joe Damato wrote:
> > + if (rxq->pool) {
> > + nest = nla_nest_start(rsp, NETDEV_A_QUEUE_XSK);
> > + nla_nest_end(rsp, nest);
> > + }
>
> nla_nest_start() can fail, you gotta nul-check the return value.
> You could possibly add an nla_put_empty_nest() helper in netlink.h
> to make this less awkward? I think the iouring guys had the same bug
Ah, right.
I'll see what a helper looks like. Feels like maybe overkill?
Thanks for the review.
On Thu, 6 Feb 2025 17:31:47 -0800 Joe Damato wrote:
> > nla_nest_start() can fail, you gotta nul-check the return value.
> > You could possibly add an nla_put_empty_nest() helper in netlink.h
> > to make this less awkward? I think the iouring guys had the same bug
>
> Ah, right.
>
> I'll see what a helper looks like. Feels like maybe overkill?
Yeah, not sure either. Technically nla_nest_end() isn't required here,
but that's not very obvious to a casual reader. So a helper that hides
that fact could be useful:
static inline int nla_put_empty_nest(struct sk_buff *skb, int attrtype)
{
return nla_nest_start(skb, attrtype) ? 0 : -EMSGSIZE;
}
But totally unsure whether it's worthwhile. Just don't want for someone
to suggest this on v4 and make you respin once again.
On Thu, Feb 06, 2025 at 05:41:38PM -0800, Jakub Kicinski wrote:
> On Thu, 6 Feb 2025 17:31:47 -0800 Joe Damato wrote:
> > > nla_nest_start() can fail, you gotta nul-check the return value.
> > > You could possibly add an nla_put_empty_nest() helper in netlink.h
> > > to make this less awkward? I think the iouring guys had the same bug
> >
> > Ah, right.
> >
> > I'll see what a helper looks like. Feels like maybe overkill?
>
> Yeah, not sure either. Technically nla_nest_end() isn't required here,
> but that's not very obvious to a casual reader. So a helper that hides
> that fact could be useful:
>
> static inline int nla_put_empty_nest(struct sk_buff *skb, int attrtype)
> {
> return nla_nest_start(skb, attrtype) ? 0 : -EMSGSIZE;
> }
Yea after reading the code it makes sense that nla_nest_end is not
needed. I wrote a small thing similar to what you proposed above,
but yours is more succinct.
I'll go with that and see how it looks.
> But totally unsure whether it's worthwhile. Just don't want for someone
> to suggest this on v4 and make you respin once again.
No worries; respinning is not the end of the world.
© 2016 - 2025 Red Hat, Inc.