| Commit message (Collapse) | Author | Age |
|---|
| |
|
|
|
|
| |
This lets other bits of the code know when USB has been connected.
Signed-off-by: Keith Packard <keithp@keithp.com>
|
| |
|
|
|
|
| |
Unused variables, functions and labels, missing enums in switch.
Signed-off-by: Keith Packard <keithp@keithp.com>
|
| |
|
|
|
|
|
| |
The USB enable register wasn't actually getting rewritten with the
enable bit turned off, so the USB device was still powered on in flight.
Signed-off-by: Keith Packard <keithp@keithp.com>
|
| |
|
|
|
|
|
|
|
| |
We were only using the low 8 bits of the setup packet reply max len,
which meant that if the other side sent a weird max len (as Windows 7
does), then we'd truncate our setup reply to whatever was in the low 8
bits of that value.
Signed-off-by: Keith Packard <keithp@keithp.com>
|
| |
|
|
|
|
|
| |
This lets the self flashing loader be linked without any of the stdio
code, which saves a bunch of memory.
Signed-off-by: Keith Packard <keithp@keithp.com>
|
| |
|
|
|
|
|
| |
Handl EP0 actions from interrupt handler. This allows USB to be used
in a taskless environment, like the STM flash loader
Signed-off-by: Keith Packard <keithp@keithp.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Unlike the AVR and CC1111 USB drivers, the STM usb driver queues IN
bytes in a local buffer instead of in the driver; this means that the
driver is queuing bytes while the previous IN packet is queued for the
host, which allows for overlapping execution.
It also means that when the local buffer is full, we must check to see
if the host has picked up the previous IN packet before trying to
queue another IN packet for transmission. This is done by always
waiting for the IN buffer to be ready before sending data.
Signed-off-by: Keith Packard <keithp@keithp.com>
|
| |
|
|
|
|
|
|
| |
This adds a pile of debugging hooks to the USB driver to try and
isolate various lockup-related issues. It's all disabled by default,
of course.
Signed-off-by: Keith Packard <keithp@keithp.com>
|
| |
|
|
|
|
|
|
|
|
|
| |
getchar relies on interrupts being blocked across the pollchar calls
and into the sleep call or it may go to sleep with data pending.
This prefixes all pollchar functions with _ to indicate that they are
to be called with interrupts blocked and eliminates all interrupt
manipulation calls from within the pollchar functions.
Signed-off-by: Keith Packard <keithp@keithp.com>
|
| |
|
|
|
|
|
|
|
|
| |
We were stealing one value (0xff) in the return value from pollchar to
indicate 'not ready yet'. Instead of doing that, use the integer value
-1 and have pollchar return an int instead of a char. That
necessitated cleaning a few other bits to make sure that 0xff wouldn't
get promoted to -1 on accident.
Signed-off-by: Keith Packard <keithp@keithp.com>
|
| |
|
|
|
|
| |
Stop using cli/sei, which are avr-specific
Signed-off-by: Keith Packard <keithp@keithp.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
sdcc offers __critical as a machine-independent way to block
interrupts, but as gcc doesn't, we need to use a compiler-independent
construct instead. ao_arch_critical has been around since the AVR
port, but some old __critical usages remained.
This fixes a bunch of random hangs when communicating with MM over USB
or the radio as the various stdio loops were running without
interrupts blocked between the test and the sleep.
Signed-off-by: Keith Packard <keithp@keithp.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Must set ao_usb_in_pending before telling USB about new data or an
interrupt could arrive at the wrong time to clear it.
Same for ao_usb_in_flushed.
Without these changes, I've seen the USB bus lock up on occasion,
waiting for an IN packet to consume data, but with no IN data pending
in the hardware.
Signed-off-by: Keith Packard <keithp@keithp.com>
|
| |
|
|
|
|
|
| |
This makes sure that a reboot will reliably cause the device to
disconnect from the USB bus.
Signed-off-by: Keith Packard <keithp@keithp.com>
|
| |
|
|
|
|
|
| |
This does "something" in radio test mode, appearing to generate a
730MHz signal.
Signed-off-by: Keith Packard <keithp@keithp.com>
|
| |
|
|
|
|
|
| |
Configure endpoint registers correctly now.
Restructure code to make sure we send the right IN packets.
Signed-off-by: Keith Packard <keithp@keithp.com>
|
| |
|
|
|
|
|
|
| |
Support multiple serial ports more cleanly
Split out parts of ao.h into separate feature header files
Signed-off-by: Keith Packard <keithp@keithp.com>
|
|
|
Emulates the usual CDC-ACM device
Signed-off-by: Keith Packard <keithp@keithp.com>
|
