Comment 10 for bug 1823753

The culprit for the 256 33 page allocations (that causes the fragmentation mentioned in comment #8) is the hisi_sas_v3_hw driver:

[ 21.220867] Call trace:
[ 21.223301] dump_backtrace+0x0/0x1b0
[ 21.226948] show_stack+0x24/0x30
[ 21.230251] dump_stack+0x90/0xb4
[ 21.233554] cma_alloc+0x3f4/0x430
[ 21.236942] dma_alloc_from_contiguous+0x70/0x80
[ 21.241544] __dma_direct_alloc_pages+0x14c/0x228
[ 21.246234] dma_direct_alloc_pages+0x48/0xc0
[ 21.250576] dma_direct_alloc+0x50/0x80
[ 21.254397] dma_alloc_attrs+0x94/0x128
[ 21.258218] dmam_alloc_attrs+0x68/0xb8
[ 21.262043] hisi_sas_alloc+0x360/0x538 [hisi_sas_main]
[ 21.267257] hisi_sas_shost_alloc_pci+0xfc/0x170 [hisi_sas_v3_hw]
[ 21.273337] hisi_sas_v3_probe+0xd8/0x360 [hisi_sas_v3_hw]
[ 21.278810] local_pci_probe+0x44/0xa8
[ 21.282544] work_for_cpu_fn+0x20/0x30
[ 21.286279] process_one_work+0x1f0/0x430
[ 21.290274] worker_thread+0x248/0x488
[ 21.294009] kthread+0x134/0x138
[ 21.297223] ret_from_fork+0x10/0x18

I wonder if it'd be possible to adjust this allocation somehow.