We need at least one per IN/IRQ endpoint, as well as at least 3 for
every OUT endpoint. Plus some more depending on the application,
in case of cardem there could be one in uart_tx and one in uart_rx.
So for cardem, it should be 7 per slot, resulting 14 for dual-slot
devices like OWHW and QMOD.
Given that we have plenty of RAM available (utilization up from 49%
to 56% with this change), we can be on the comfortable size and
easily have 20 buffers for now.
Future MITM / triple-play applications likely need more.
Change-Id: I022ba3bb3e1f5fd0c364aafe1a23f8fd9b2d8499
Related: OS#4251
This helps when debugging the firmware, as it shows the current
utliization of the 10-msgb-talloc pool.
Change-Id: Ib10c4396cd4c9c4a6257cf45886e367214787927
Related: OS#4251
There have been tons of format-string related bugs in our code which
we never discovered due to disabling -Wformat. Let's fix that.
Change-Id: I5ec466361bcc526fac1f4897673264ee5af3458b
the curent local copies of libosmocore headers + source is a temporary
hack anyway. We should instead rely on a system-wide install of
libosmocore cross-compiled for arm-none-eabi. But leave that as a
second (later) step beyond this patch.
Change-Id: Ia63fd842d45a2b404233b4326050e7eda0604cf0
We now generalize the USB communiction and abandon the 'req_ctx'
structure inherited from openpcd. Instead we use the libosmocore 'msgb'
structure to handle incoming and outgoing USB tranfers. We also use
linuxlist-based msgb-queues for each endpoint.