If the remote end just closes an endpoint and no longer handles it,
the system is prone to intensive cpu polling via mpfs_write_tx_fifo()
especially if the device side doesn't know a thing about what the
remote did.
Fix this by marking the EP as dead, which will skip all writes causing
unnecessary polling. The EP is back in business if the remote end
sends some data (rx) or the connection is re-established.
Signed-off-by: Eero Nurkkala <eero.nurkkala@offcode.fi>
Signed-off-by: Ville Juven <ville.juven@unikie.com>
95163f504c