[PATCH net] eth: bnxt: fix counting packets discarded due to OOM and netpoll

[PATCH net] eth: bnxt: fix counting packets discarded due to OOM and netpoll

[Jakub Kicinski]

I added OOM and netpoll discard counters, naively assuming that
the cpr pointer is pointing to a common completion ring.
Turns out that is usually *a* completion ring but not *the*
completion ring which bnapi->cp_ring points to. bnapi->cp_ring
is where the stats are read from, so we end up reporting 0
thru ethtool -S and qstat even though the drop events have happened.
Make 100% sure we're recording statistics in the correct structure.

Fixes: 907fd4a294db ("bnxt: count discards due to memory allocation errors")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
CC: michael.chan@broadcom.com
CC: edwin.peer@broadcom.com
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 44 ++++++++++-------------
 1 file changed, 18 insertions(+), 26 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index a30df865be7b..88cba1e52903 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -1811,7 +1811,7 @@ static inline struct sk_buff *bnxt_tpa_end(struct bnxt *bp,
 		skb = bnxt_copy_skb(bnapi, data_ptr, len, mapping);
 		if (!skb) {
 			bnxt_abort_tpa(cpr, idx, agg_bufs);
-			cpr->sw_stats.rx.rx_oom_discards += 1;
+			cpr->bnapi->cp_ring.sw_stats.rx.rx_oom_discards += 1;
 			return NULL;
 		}
 	} else {
@@ -1821,7 +1821,7 @@ static inline struct sk_buff *bnxt_tpa_end(struct bnxt *bp,
 		new_data = __bnxt_alloc_rx_frag(bp, &new_mapping, GFP_ATOMIC);
 		if (!new_data) {
 			bnxt_abort_tpa(cpr, idx, agg_bufs);
-			cpr->sw_stats.rx.rx_oom_discards += 1;
+			cpr->bnapi->cp_ring.sw_stats.rx.rx_oom_discards += 1;
 			return NULL;
 		}
 
@@ -1837,7 +1837,7 @@ static inline struct sk_buff *bnxt_tpa_end(struct bnxt *bp,
 		if (!skb) {
 			skb_free_frag(data);
 			bnxt_abort_tpa(cpr, idx, agg_bufs);
-			cpr->sw_stats.rx.rx_oom_discards += 1;
+			cpr->bnapi->cp_ring.sw_stats.rx.rx_oom_discards += 1;
 			return NULL;
 		}
 		skb_reserve(skb, bp->rx_offset);
@@ -1848,7 +1848,7 @@ static inline struct sk_buff *bnxt_tpa_end(struct bnxt *bp,
 		skb = bnxt_rx_agg_pages_skb(bp, cpr, skb, idx, agg_bufs, true);
 		if (!skb) {
 			/* Page reuse already handled by bnxt_rx_pages(). */
-			cpr->sw_stats.rx.rx_oom_discards += 1;
+			cpr->bnapi->cp_ring.sw_stats.rx.rx_oom_discards += 1;
 			return NULL;
 		}
 	}
@@ -2127,11 +2127,8 @@ static int bnxt_rx_pkt(struct bnxt *bp, struct bnxt_cp_ring_info *cpr,
 			u32 frag_len = bnxt_rx_agg_pages_xdp(bp, cpr, &xdp,
 							     cp_cons, agg_bufs,
 							     false);
-			if (!frag_len) {
-				cpr->sw_stats.rx.rx_oom_discards += 1;
-				rc = -ENOMEM;
-				goto next_rx;
-			}
+			if (!frag_len)
+				goto oom_next_rx;
 		}
 		xdp_active = true;
 	}
@@ -2157,9 +2154,7 @@ static int bnxt_rx_pkt(struct bnxt *bp, struct bnxt_cp_ring_info *cpr,
 				else
 					bnxt_xdp_buff_frags_free(rxr, &xdp);
 			}
-			cpr->sw_stats.rx.rx_oom_discards += 1;
-			rc = -ENOMEM;
-			goto next_rx;
+			goto oom_next_rx;
 		}
 	} else {
 		u32 payload;
@@ -2170,29 +2165,21 @@ static int bnxt_rx_pkt(struct bnxt *bp, struct bnxt_cp_ring_info *cpr,
 			payload = 0;
 		skb = bp->rx_skb_func(bp, rxr, cons, data, data_ptr, dma_addr,
 				      payload | len);
-		if (!skb) {
-			cpr->sw_stats.rx.rx_oom_discards += 1;
-			rc = -ENOMEM;
-			goto next_rx;
-		}
+		if (!skb)
+			goto oom_next_rx;
 	}
 
 	if (agg_bufs) {
 		if (!xdp_active) {
 			skb = bnxt_rx_agg_pages_skb(bp, cpr, skb, cp_cons, agg_bufs, false);
-			if (!skb) {
-				cpr->sw_stats.rx.rx_oom_discards += 1;
-				rc = -ENOMEM;
-				goto next_rx;
-			}
+			if (!skb)
+				goto oom_next_rx;
 		} else {
 			skb = bnxt_xdp_build_skb(bp, skb, agg_bufs, rxr->page_pool, &xdp, rxcmp1);
 			if (!skb) {
 				/* we should be able to free the old skb here */
 				bnxt_xdp_buff_frags_free(rxr, &xdp);
-				cpr->sw_stats.rx.rx_oom_discards += 1;
-				rc = -ENOMEM;
-				goto next_rx;
+				goto oom_next_rx;
 			}
 		}
 	}
@@ -2270,6 +2257,11 @@ static int bnxt_rx_pkt(struct bnxt *bp, struct bnxt_cp_ring_info *cpr,
 	*raw_cons = tmp_raw_cons;
 
 	return rc;
+
+oom_next_rx:
+	cpr->bnapi->cp_ring.sw_stats.rx.rx_oom_discards += 1;
+	rc = -ENOMEM;
+	goto next_rx;
 }
 
 /* In netpoll mode, if we are using a combined completion ring, we need to
@@ -2316,7 +2308,7 @@ static int bnxt_force_rx_discard(struct bnxt *bp,
 	}
 	rc = bnxt_rx_pkt(bp, cpr, raw_cons, event);
 	if (rc && rc != -EBUSY)
-		cpr->sw_stats.rx.rx_netpoll_discards += 1;
+		cpr->bnapi->cp_ring.sw_stats.rx.rx_netpoll_discards += 1;
 	return rc;
 }
 
-- 
2.44.0

Re: [PATCH net] eth: bnxt: fix counting packets discarded due to OOM and netpoll

[Michael Chan]

On Tue, Apr 23, 2024 at 5:21 PM Jakub Kicinski <kuba@kernel.org> wrote:
>
> I added OOM and netpoll discard counters, naively assuming that
> the cpr pointer is pointing to a common completion ring.
> Turns out that is usually *a* completion ring but not *the*
> completion ring which bnapi->cp_ring points to. bnapi->cp_ring
> is where the stats are read from, so we end up reporting 0
> thru ethtool -S and qstat even though the drop events have happened.
> Make 100% sure we're recording statistics in the correct structure.
>
> Fixes: 907fd4a294db ("bnxt: count discards due to memory allocation errors")
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Yes, on the newer chips, we have 2 cpr pointers and the correct one
must be used for the sw_stats.  We actually have a patch that changes
cpr->sw_stats to a pointer so that both cpr can share the same
sw_stats structure.  This reminds me to post that patch soon to avoid
this type of confusion in the future.  Thanks.

Reviewed-by: Michael Chan <michael.chan@broadcom.com>

Re: [PATCH net] eth: bnxt: fix counting packets discarded due to OOM and netpoll

[patchwork-bot+netdevbpf]

Hello:

This patch was applied to netdev/net.git (main)
by Jakub Kicinski <kuba@kernel.org>:

On Tue, 23 Apr 2024 17:21:48 -0700 you wrote:
> I added OOM and netpoll discard counters, naively assuming that
> the cpr pointer is pointing to a common completion ring.
> Turns out that is usually *a* completion ring but not *the*
> completion ring which bnapi->cp_ring points to. bnapi->cp_ring
> is where the stats are read from, so we end up reporting 0
> thru ethtool -S and qstat even though the drop events have happened.
> Make 100% sure we're recording statistics in the correct structure.
> 
> [...]

Here is the summary with links:
  - [net] eth: bnxt: fix counting packets discarded due to OOM and netpoll
    https://git.kernel.org/netdev/net/c/730117730709

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html