[PATCH net-next 0/2] bnxt_en: Increase maximum RX ring size when jumbo ring is unused

[PATCH net-next 0/2] bnxt_en: Increase maximum RX ring size when jumbo ring is unused

[Michael Chan]

[-- Attachment #1: Type: text/plain, Size: 831 bytes --]

The RX jumbo ring is automatically enabled when HW GRO/LRO is enabled or
when the MTU exceeds the page size.  The RX jumbo ring provides a lot
more RX buffer space when it is in use.  When the RX jumbo ring is not
in use, some users report that the current maximum of 2K buffers is
too limiting.  This patchset increases the maximum to 8K buffers when
the RX jumbo ring is not used.  The default RX ring size is unchanged
at 511.

Michael Chan (2):
  bnxt_en: Don't use static arrays for completion ring pages.
  bnxt_en: Increase maximum RX ring size if jumbo ring is not used.

 drivers/net/ethernet/broadcom/bnxt/bnxt.c     | 75 ++++++++++++++++++-
 drivers/net/ethernet/broadcom/bnxt/bnxt.h     | 17 +++--
 .../net/ethernet/broadcom/bnxt/bnxt_ethtool.c |  9 ++-
 3 files changed, 90 insertions(+), 11 deletions(-)

-- 
2.18.1


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4209 bytes --]

[PATCH net-next 1/2] bnxt_en: Don't use static arrays for completion ring pages

[Michael Chan]

[-- Attachment #1: Type: text/plain, Size: 4402 bytes --]

We currently store these page addresses and DMA addreses in static
arrays.  On systems with 4K pages, we support up to 64 pages per
completion ring.  The actual number of pages for each completion ring
may be much less than 64.  For example, when the RX ring size is set
to the default 511 entries, only 16 completion ring pages are needed
per ring.

In the next patch, we'll be doubling the maximum number of completion
pages.  So we convert to allocate these arrays as needed instead of
declaring them statically.

Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Reviewed-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 65 +++++++++++++++++++++++
 drivers/net/ethernet/broadcom/bnxt/bnxt.h |  6 +--
 2 files changed, 68 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 03b821897cf3..cc758a66fac0 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -3163,6 +3163,58 @@ static int bnxt_alloc_tx_rings(struct bnxt *bp)
 	return 0;
 }
 
+static void bnxt_free_cp_arrays(struct bnxt_cp_ring_info *cpr)
+{
+	kfree(cpr->cp_desc_ring);
+	cpr->cp_desc_ring = NULL;
+	kfree(cpr->cp_desc_mapping);
+	cpr->cp_desc_mapping = NULL;
+}
+
+static int bnxt_alloc_cp_arrays(struct bnxt_cp_ring_info *cpr, int n)
+{
+	cpr->cp_desc_ring = kcalloc(n, sizeof(*cpr->cp_desc_ring), GFP_KERNEL);
+	if (!cpr->cp_desc_ring)
+		return -ENOMEM;
+	cpr->cp_desc_mapping = kcalloc(n, sizeof(*cpr->cp_desc_mapping),
+				       GFP_KERNEL);
+	if (!cpr->cp_desc_mapping)
+		return -ENOMEM;
+	return 0;
+}
+
+static void bnxt_free_all_cp_arrays(struct bnxt *bp)
+{
+	int i;
+
+	if (!bp->bnapi)
+		return;
+	for (i = 0; i < bp->cp_nr_rings; i++) {
+		struct bnxt_napi *bnapi = bp->bnapi[i];
+
+		if (!bnapi)
+			continue;
+		bnxt_free_cp_arrays(&bnapi->cp_ring);
+	}
+}
+
+static int bnxt_alloc_all_cp_arrays(struct bnxt *bp)
+{
+	int i, n = bp->cp_nr_pages;
+
+	for (i = 0; i < bp->cp_nr_rings; i++) {
+		struct bnxt_napi *bnapi = bp->bnapi[i];
+		int rc;
+
+		if (!bnapi)
+			continue;
+		rc = bnxt_alloc_cp_arrays(&bnapi->cp_ring, n);
+		if (rc)
+			return rc;
+	}
+	return 0;
+}
+
 static void bnxt_free_cp_rings(struct bnxt *bp)
 {
 	int i;
@@ -3190,6 +3242,7 @@ static void bnxt_free_cp_rings(struct bnxt *bp)
 			if (cpr2) {
 				ring = &cpr2->cp_ring_struct;
 				bnxt_free_ring(bp, &ring->ring_mem);
+				bnxt_free_cp_arrays(cpr2);
 				kfree(cpr2);
 				cpr->cp_ring_arr[j] = NULL;
 			}
@@ -3208,6 +3261,12 @@ static struct bnxt_cp_ring_info *bnxt_alloc_cp_sub_ring(struct bnxt *bp)
 	if (!cpr)
 		return NULL;
 
+	rc = bnxt_alloc_cp_arrays(cpr, bp->cp_nr_pages);
+	if (rc) {
+		bnxt_free_cp_arrays(cpr);
+		kfree(cpr);
+		return NULL;
+	}
 	ring = &cpr->cp_ring_struct;
 	rmem = &ring->ring_mem;
 	rmem->nr_pages = bp->cp_nr_pages;
@@ -3218,6 +3277,7 @@ static struct bnxt_cp_ring_info *bnxt_alloc_cp_sub_ring(struct bnxt *bp)
 	rc = bnxt_alloc_ring(bp, rmem);
 	if (rc) {
 		bnxt_free_ring(bp, rmem);
+		bnxt_free_cp_arrays(cpr);
 		kfree(cpr);
 		cpr = NULL;
 	}
@@ -4253,6 +4313,7 @@ static void bnxt_free_mem(struct bnxt *bp, bool irq_re_init)
 	bnxt_free_tx_rings(bp);
 	bnxt_free_rx_rings(bp);
 	bnxt_free_cp_rings(bp);
+	bnxt_free_all_cp_arrays(bp);
 	bnxt_free_ntp_fltrs(bp, irq_re_init);
 	if (irq_re_init) {
 		bnxt_free_ring_stats(bp);
@@ -4373,6 +4434,10 @@ static int bnxt_alloc_mem(struct bnxt *bp, bool irq_re_init)
 			goto alloc_mem_err;
 	}
 
+	rc = bnxt_alloc_all_cp_arrays(bp);
+	if (rc)
+		goto alloc_mem_err;
+
 	bnxt_init_ring_struct(bp);
 
 	rc = bnxt_alloc_rx_rings(bp);
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.h b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
index e379c48c1df9..eba8d8f0ac81 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.h
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
@@ -972,11 +972,11 @@ struct bnxt_cp_ring_info {
 	struct dim		dim;
 
 	union {
-		struct tx_cmp	*cp_desc_ring[MAX_CP_PAGES];
-		struct nqe_cn	*nq_desc_ring[MAX_CP_PAGES];
+		struct tx_cmp	**cp_desc_ring;
+		struct nqe_cn	**nq_desc_ring;
 	};
 
-	dma_addr_t		cp_desc_mapping[MAX_CP_PAGES];
+	dma_addr_t		*cp_desc_mapping;
 
 	struct bnxt_stats_mem	stats;
 	u32			hw_stats_ctx_id;
-- 
2.18.1


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4209 bytes --]

[PATCH net-next 2/2] bnxt_en: Increase maximum RX ring size if jumbo ring is not used

[Michael Chan]

[-- Attachment #1: Type: text/plain, Size: 4276 bytes --]

The current maximum RX ring size is defined assuming the RX jumbo ring
(aka aggregation ring) is used.  The RX jumbo ring is automicatically used
when the MTU exceeds a threshold or when rx-gro-hw/lro is enabled.  The RX
jumbo ring is automatically sized up to 4 times the size of the RX ring
size.

The BNXT_MAX_RX_DESC_CNT constant is the upper limit on the size of the
RX ring whether or not the RX jumbo ring is used.  Obviously, the
maximum amount of RX buffer space is significantly less when the RX jumbo
ring is not used.

To increase flexibility for the user who does not use the RX jumbo ring,
we now define a bigger maximum RX ring size when the RX jumbo ring is not
used.  The maximum RX ring size is now up to 8K when the RX jumbo ring
is not used.  The maximum completion ring size also needs to be scaled
up to accomodate the larger maximum RX ring size.

Note that when the RX jumbo ring is re-enabled, the RX ring size will
automatically drop if it exceeds the maximum.

Reviewed-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c         | 10 ++++++++--
 drivers/net/ethernet/broadcom/bnxt/bnxt.h         | 11 +++++++----
 drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c |  9 +++++++--
 3 files changed, 22 insertions(+), 8 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index cc758a66fac0..865fcb8cf29f 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -3710,9 +3710,15 @@ void bnxt_set_ring_params(struct bnxt *bp)
 		if (jumbo_factor > agg_factor)
 			agg_factor = jumbo_factor;
 	}
-	agg_ring_size = ring_size * agg_factor;
+	if (agg_factor) {
+		if (ring_size > BNXT_MAX_RX_DESC_CNT_JUM_ENA) {
+			ring_size = BNXT_MAX_RX_DESC_CNT_JUM_ENA;
+			netdev_warn(bp->dev, "RX ring size reduced from %d to %d because the jumbo ring is now enabled\n",
+				    bp->rx_ring_size, ring_size);
+			bp->rx_ring_size = ring_size;
+		}
+		agg_ring_size = ring_size * agg_factor;
 
-	if (agg_ring_size) {
 		bp->rx_agg_nr_pages = bnxt_calc_nr_ring_pages(agg_ring_size,
 							RX_DESC_CNT);
 		if (bp->rx_agg_nr_pages > MAX_RX_AGG_PAGES) {
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.h b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
index eba8d8f0ac81..9c3324e76ff7 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.h
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
@@ -596,15 +596,17 @@ struct nqe_cn {
 #define MAX_TPA_SEGS_P5	0x3f
 
 #if (BNXT_PAGE_SHIFT == 16)
-#define MAX_RX_PAGES	1
+#define MAX_RX_PAGES_AGG_ENA	1
+#define MAX_RX_PAGES	4
 #define MAX_RX_AGG_PAGES	4
 #define MAX_TX_PAGES	1
-#define MAX_CP_PAGES	8
+#define MAX_CP_PAGES	16
 #else
-#define MAX_RX_PAGES	8
+#define MAX_RX_PAGES_AGG_ENA	8
+#define MAX_RX_PAGES	32
 #define MAX_RX_AGG_PAGES	32
 #define MAX_TX_PAGES	8
-#define MAX_CP_PAGES	64
+#define MAX_CP_PAGES	128
 #endif
 
 #define RX_DESC_CNT (BNXT_PAGE_SIZE / sizeof(struct rx_bd))
@@ -622,6 +624,7 @@ struct nqe_cn {
 #define HW_CMPD_RING_SIZE (sizeof(struct tx_cmp) * CP_DESC_CNT)
 
 #define BNXT_MAX_RX_DESC_CNT		(RX_DESC_CNT * MAX_RX_PAGES - 1)
+#define BNXT_MAX_RX_DESC_CNT_JUM_ENA	(RX_DESC_CNT * MAX_RX_PAGES_AGG_ENA - 1)
 #define BNXT_MAX_RX_JUM_DESC_CNT	(RX_DESC_CNT * MAX_RX_AGG_PAGES - 1)
 #define BNXT_MAX_TX_DESC_CNT		(TX_DESC_CNT * MAX_TX_PAGES - 1)
 
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
index 786ca51e669b..485252d12245 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
@@ -768,8 +768,13 @@ static void bnxt_get_ringparam(struct net_device *dev,
 {
 	struct bnxt *bp = netdev_priv(dev);
 
-	ering->rx_max_pending = BNXT_MAX_RX_DESC_CNT;
-	ering->rx_jumbo_max_pending = BNXT_MAX_RX_JUM_DESC_CNT;
+	if (bp->flags & BNXT_FLAG_AGG_RINGS) {
+		ering->rx_max_pending = BNXT_MAX_RX_DESC_CNT_JUM_ENA;
+		ering->rx_jumbo_max_pending = BNXT_MAX_RX_JUM_DESC_CNT;
+	} else {
+		ering->rx_max_pending = BNXT_MAX_RX_DESC_CNT;
+		ering->rx_jumbo_max_pending = 0;
+	}
 	ering->tx_max_pending = BNXT_MAX_TX_DESC_CNT;
 
 	ering->rx_pending = bp->rx_ring_size;
-- 
2.18.1


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4209 bytes --]

Re: [PATCH net-next 0/2] bnxt_en: Increase maximum RX ring size when jumbo ring is unused

[patchwork-bot+netdevbpf]

Hello:

This series was applied to netdev/net-next.git (refs/heads/master):

On Mon,  2 Aug 2021 10:52:37 -0400 you wrote:
> The RX jumbo ring is automatically enabled when HW GRO/LRO is enabled or
> when the MTU exceeds the page size.  The RX jumbo ring provides a lot
> more RX buffer space when it is in use.  When the RX jumbo ring is not
> in use, some users report that the current maximum of 2K buffers is
> too limiting.  This patchset increases the maximum to 8K buffers when
> the RX jumbo ring is not used.  The default RX ring size is unchanged
> at 511.
> 
> [...]

Here is the summary with links:
  - [net-next,1/2] bnxt_en: Don't use static arrays for completion ring pages
    https://git.kernel.org/netdev/net-next/c/03c7448790b8
  - [net-next,2/2] bnxt_en: Increase maximum RX ring size if jumbo ring is not used
    https://git.kernel.org/netdev/net-next/c/c1129b51ca0e

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