On 7/8/24 14:40, Simon Horman wrote:
On Wed, Jul 03, 2024 at 08:59:19AM -0400, Mateusz Polchlopek wrote:
From: Przemek Kitszel <przemyslaw.kits...@intel.com>
Add Tx hang devlink health reporter, see struct ice_tx_hang_event to see
what is reported.
Subsequent commits will extend it by more info, for now it dumps
descriptors with little metadata.
Signed-off-by: Przemek Kitszel <przemyslaw.kits...@intel.com>
Reviewed-by: Igor Bagnucki <igor.bagnu...@intel.com>
Reviewed-by: Wojciech Drewek <wojciech.dre...@intel.com>
Signed-off-by: Mateusz Polchlopek <mateusz.polchlo...@intel.com>
...
+/**
+ * ice_fmsg_put_ptr - put hex value of pointer into fmsg
+ *
+ * @fmsg: devlink fmsg under construction
+ * @name: name to pass
+ * @ptr: 64 bit value to print as hex and put into fmsg
+ */
+static void ice_fmsg_put_ptr(struct devlink_fmsg *fmsg, const char *name,
+ void *ptr)
+{
+ char buf[sizeof(ptr) * 3];
+
+ sprintf(buf, "%p", ptr);
+ devlink_fmsg_put(fmsg, name, buf);
+}
...
+static int ice_tx_hang_reporter_dump(struct devlink_health_reporter *reporter,
+ struct devlink_fmsg *fmsg, void *priv_ctx,
+ struct netlink_ext_ack *extack)
+{
+ struct ice_tx_hang_event *event = priv_ctx;
+
+ devlink_fmsg_obj_nest_start(fmsg);
+ ICE_DEVLINK_FMSG_PUT_FIELD(fmsg, event, head);
+ ICE_DEVLINK_FMSG_PUT_FIELD(fmsg, event, intr);
+ ICE_DEVLINK_FMSG_PUT_FIELD(fmsg, event, vsi_num);
+ ICE_DEVLINK_FMSG_PUT_FIELD(fmsg, event, queue);
+ ICE_DEVLINK_FMSG_PUT_FIELD(fmsg, event, next_to_clean);
+ ICE_DEVLINK_FMSG_PUT_FIELD(fmsg, event, next_to_use);
+ devlink_fmsg_put(fmsg, "irq-mapping", event->tx_ring->q_vector->name);
+ ice_fmsg_put_ptr(fmsg, "desc-ptr", event->tx_ring->desc);
+ ice_fmsg_put_ptr(fmsg, "dma-ptr", (void *)event->tx_ring->dma);
As reported by the kernel test robot, GCC 13 complains about this cast:
.../devlink_health.c: In function 'ice_tx_hang_reporter_dump':
.../devlink_health.c:76:43: warning: cast to pointer from integer of
different size [-Wint-to-pointer-cast]
76 | ice_fmsg_put_ptr(fmsg, "dma-ptr", (void
*)event->tx_ring->dma);
|
Perhaps a good solution is to add a helper similar to ice_fmsg_put_ptr,
but which takes a dma_buf_t rather than a void * as it's last argument.
instead of duplicating the function for just one call, I will simply
resolve the warning by yet another cast:
ice_fmsg_put_ptr(fmsg, "dma-ptr", (void *)(long)event->tx_ring->dma);
^^^^^^ // cast to long added
+ devlink_fmsg_binary_pair_put(fmsg, "desc", event->tx_ring->desc,
+ size_mul(event->tx_ring->count,
+ sizeof(struct ice_tx_desc)));
Here I would drop size_mul(), as any wrong ::count value could easily
extent the dump past tx_ring memory, resulting in attempt at reading
past their page
And we are not really protecting against "too big" fmsg, as it is capped
anyway to 4-8K.
Perhaps fmsg-put also ::count to aid spotting such cases, but only if it
is not the default 256.
--
not a change request, just digression:
it would be nice for devlink_fmsg_binary_pair_put() to compress
"repeated same value", like hexdump(1) does.
+ devlink_fmsg_obj_nest_end(fmsg);
+
+ return 0;
+}
