The default implementation was completely broken. Use the GCC specific
__alignof__ if compiling for C99 or C++03. If not C++11, C11, or
GCC, then it is an error.
When the MII bus is unterminated on unused interfaces, it results in PHY
read timeouts which manifest as spurious PHYs during the attach call.
Detect these timeouts during the probe so the device can be ignored.
This alters the selection of the 4 Cadence GEM interfaces on the Zynq
Ultrascale+ MPSoC BSP to be provided by config.inc instead of being
provided by options in the RTEMS BSP itself since those options appear
to be dead code when not used in conjunction with LibBSD.
Added a new test for the TTCP command. Modified default-network-init.h
to conditionally build the shell with TTCP. Modified libbsd.py to build
the new TTCP test.
Updated ttcp.c to build clean for RTEMS 6 and the machines it
originally built for. Also fixed ttcp.c to close network
sockets after completion. Defined a shell command for TTCP in
rtems-bsd-shell-ttcp.c. Added TTCP to the list of RTEMS network
commands in netcmds-config.h. Added declaration of the TTCP shell
command to rtems-bsd-commands.h. Modified libbsd.py to make waf
build TTCP and its shell command.
Added the original Test TCP (TTCP) program in unmodified form.
Also added the original README for TTCP. Both the README and the
TTCP program were sourced from the first commit in the RTEMS
network demos repository.
Since the epoch call callbacks may lock/unlock a mutex the callbacks must be
called from within thread context with thread dispatching enabled. Use thread
pinning to ensure that the calling thread stays on the right processor. Use
the interrupt server to ensure the thread context for the epoch drain.
Changed nfs_utime() to nfs_utimens(), changed the arguments to use
a timespec array instead of individual variables for access and
modified time.
Updates #4400
setkey uses pfkey_open to open a socket. But setkey doesn't close the
socket.
The libipsec functions are used only by user space applications (setkey
and racoon). Adding the wrapper for socket makes sure that the opened
socket is registered and closes if the application exits.
Fixes#4405
Possible data sources for SD driver:
- mmc_sd_switch():
- length: 64 byte;
- buffer on stack
- mmc_test_bus_width():
- length: 4 or 8 byte
- buffer in program memory or on stack
- mmc_app_send_scr():
- length: 8 byte
- buffer from device ivar structure
- mmc_app_sd_status():
- length: 64 byte
- buffer from device ivar structure
- mmc_send_ext_csd():
- length: MMC_EXTCSD_SIZE = 512
- buffer from device ivar structure
- rtems_bsd_mmcsd_disk_read_write():
- length: depends on read
- buffer from rtems_blkdev buffer -> already aligned
- mmcsd_ioctl_cmd():
- length: depends on call
- buffer malloced, not aligned -> patched in RTEMS
So the problematic buffers are only the ones up to 512 bytes. Copy these data
into a buffer to avoid that problem.
ZynqMP hardware comes with many different Ethernet PHYs depending on
which board is used. Add the UKPHY driver to handle basic PHY
interaction for any unrecognized PHYs.
According to commentary on GCC bug
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99143, the alignment
behavior of linker sections on which RTEMS has relied was never
guaranteed to be consistent across platforms and any alignment
requirements for linker sections needs to be enforced explicitly.
This adds those explicit alignment requirements.
Add a System Level Control Register driver for the Xilinx Zynq
Ultrascale+ MPSoC with basic clock control functionality for use with
the Cadence GEM. This also removes the Zynq-7000 clock control weakref
from compilation depending on the BSP in use.
Similar to the UARTs, ZynqMP hardware platforms use the highest
memory-mapped CGEM peripheral as the primary instance of that
peripheral. This change allows operation on hardware as well as QEMU.
This is a port of the latest patch in FreeBSD that adds support for
64bit CGEMs as used in ZynqMP. This does not include the work necessary
for support on RISC-V systems.
Alignment on 64bit hardware is strict in comparison to running in an
emulator. This resolves an alignment exception when allocating memory on
real hardware.
With the previous fix, it could happen that the end of the packet hasn't
been flushed. For example assume the following addresses:
ds_addr: 0x81c804A
ds_len: 0x57
In that case the data ends at 0x81c80a1. But due to the rounding the
area from 0x81c8040 to 0x81c80a0 would have been flushed.
This fix now first calculates the start and end address, aligns these
addresses and then recalculates the len that has to be flushed.
Update #4180
Update the host frame interval after a device connection. Select also
the FS/LS PHY clock. It is not clear if this works on all platforms.
Update #3910.
