bnxt_en: Avoid order-5 memory allocation for TPA data

In the Linux kernel, the following vulnerability has been resolved:

bnxt_en: Avoid order-5 memory allocation for TPA data

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.