This is correct fix for c12128ce ("mt76: use a per rx queue page
fragment cache"). We use wrong length when we allocate segments for
MCU transmissions, which require bigger segment size than e->buf_size.

Commit 481bb043 ("mt76: usb: make rx page_frag_cache access atomic")
partially solved the problem or actually mask it by changing
mt76u_mcu_init_rx() and mt76u_alloc_queues() sequence, so e->buf_size
become non zero any longer, but still not big enough to handle MCU data.

Patch fixes memory corruption which can manifest itself as random,
not easy to reproduce crashes, during mt76 driver load or unload.

Fixes: c12128ce ("mt76: use a per rx queue page fragment cache")
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Lorenzo Bianconi <lorenzo.bianconi@redhat.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>