bnxt_en: Avoid order-5 memory allocation for TPA data

[ Upstream commit accd7e23693aaaa9aa0d3e9eca0ae77d1be80ab3 ]

The driver needs to keep track of all the possible concurrent TPA (GRO/LRO)
completions on the aggregation ring.  On P5 chips, the maximum number
of concurrent TPA is 256 and the amount of memory we allocate is order-5
on systems using 4K pages.  Memory allocation failure has been reported:

NetworkManager: page allocation failure: order:5, mode:0x40dc0(GFP_KERNEL|__GFP_COMP|__GFP_ZERO), nodemask=(null),cpuset=/,mems_allowed=0-1
CPU: 15 PID: 2995 Comm: NetworkManager Kdump: loaded Not tainted 5.10.156 #1
Hardware name: Dell Inc. PowerEdge R660/0M1CC5, BIOS 0.2.25 08/12/2022
Call Trace:
 dump_stack+0x57/0x6e
 warn_alloc.cold.120+0x7b/0xdd
 ? _cond_resched+0x15/0x30
 ? __alloc_pages_direct_compact+0x15f/0x170
 __alloc_pages_slowpath.constprop.108+0xc58/0xc70
 __alloc_pages_nodemask+0x2d0/0x300
 kmalloc_order+0x24/0xe0
 kmalloc_order_trace+0x19/0x80
 bnxt_alloc_mem+0x1150/0x15c0 [bnxt_en]
 ? bnxt_get_func_stat_ctxs+0x13/0x60 [bnxt_en]
 __bnxt_open_nic+0x12e/0x780 [bnxt_en]
 bnxt_open+0x10b/0x240 [bnxt_en]
 __dev_open+0xe9/0x180
 __dev_change_flags+0x1af/0x220
 dev_change_flags+0x21/0x60
 do_setlink+0x35c/0x1100

Instead of allocating this big chunk of memory and dividing it up for the
concurrent TPA instances, allocate each small chunk separately for each
TPA instance.  This will reduce it to order-0 allocations.

Fixes: 79632e9 ("bnxt_en: Expand bnxt_tpa_info struct to support 57500 chips.")
Reviewed-by: Somnath Kotur <[email protected]>
Reviewed-by: Damodharam Ammepalli <[email protected]>
Reviewed-by: Pavan Chebbi <[email protected]>
Signed-off-by: Michael Chan <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
Signed-off-by: Sasha Levin <[email protected]>

drivers/net/ethernet/broadcom/bnxt/bnxt.c

@@ -2999,16 +2999,18 @@ static int bnxt_alloc_ring(struct bnxt *bp, struct bnxt_ring_mem_info *rmem)
 
 static void bnxt_free_tpa_info(struct bnxt *bp)
 {
-	int i;
+	int i, j;
 
 	for (i = 0; i < bp->rx_nr_rings; i++) {
 		struct bnxt_rx_ring_info *rxr = &bp->rx_ring[i];
 
 		kfree(rxr->rx_tpa_idx_map);
 		rxr->rx_tpa_idx_map = NULL;
 		if (rxr->rx_tpa) {
-			kfree(rxr->rx_tpa[0].agg_arr);
-			rxr->rx_tpa[0].agg_arr = NULL;
+			for (j = 0; j < bp->max_tpa; j++) {
+				kfree(rxr->rx_tpa[j].agg_arr);
+				rxr->rx_tpa[j].agg_arr = NULL;
+			}
 		}
 		kfree(rxr->rx_tpa);
 		rxr->rx_tpa = NULL;
@@ -3017,14 +3019,13 @@ static void bnxt_free_tpa_info(struct bnxt *bp)
 
 static int bnxt_alloc_tpa_info(struct bnxt *bp)
 {
-	int i, j, total_aggs = 0;
+	int i, j;
 
 	bp->max_tpa = MAX_TPA;
 	if (bp->flags & BNXT_FLAG_CHIP_P5) {
 		if (!bp->max_tpa_v2)
 			return 0;
 		bp->max_tpa = max_t(u16, bp->max_tpa_v2, MAX_TPA_P5);
-		total_aggs = bp->max_tpa * MAX_SKB_FRAGS;
 	}
 
 	for (i = 0; i < bp->rx_nr_rings; i++) {
@@ -3038,12 +3039,12 @@ static int bnxt_alloc_tpa_info(struct bnxt *bp)
 
 		if (!(bp->flags & BNXT_FLAG_CHIP_P5))
 			continue;
-		agg = kcalloc(total_aggs, sizeof(*agg), GFP_KERNEL);
-		rxr->rx_tpa[0].agg_arr = agg;
-		if (!agg)
-			return -ENOMEM;
-		for (j = 1; j < bp->max_tpa; j++)
-			rxr->rx_tpa[j].agg_arr = agg + j * MAX_SKB_FRAGS;
+		for (j = 0; j < bp->max_tpa; j++) {
+			agg = kcalloc(MAX_SKB_FRAGS, sizeof(*agg), GFP_KERNEL);
+			if (!agg)
+				return -ENOMEM;
+			rxr->rx_tpa[j].agg_arr = agg;
+		}
 		rxr->rx_tpa_idx_map = kzalloc(sizeof(*rxr->rx_tpa_idx_map),
 					      GFP_KERNEL);
 		if (!rxr->rx_tpa_idx_map)