Users are reporting font issues under VirtualBox. Try this as a speculative fix since the old 5.x behaviour was to pass in a partially zero'd register set. Incidentally, the input regs are also used for output, so at least we now avoid potentially constructing a pointer to the font buffer using garbage register values. Cc: H. Peter Anvin <hpa@zytor.com> Signed-off-by: Matt Fleming <matt.fleming@intel.com>
Since the UEFI specification doesn't state that the firmware is responsible for filling out this field if it's unset (though some will set a default value) we need to do it ourselves. Signed-off-by: Matt Fleming <matt.fleming@intel.com>
Some firmware implementations fill out a default value for the Time To Live field when none is set, others do not. Since the UEFI specification doesn't mandate that the firmware set this field, we need to do it ourselves. Pick the recommended value of 64. Signed-off-by: Matt Fleming <matt.fleming@intel.com>
The TFTP protocol uses the local port as an idenitifer during a transfer (TID), which means that once we've established a TFTP connection, we must ensure we reuse the same local port number in each packet. Failure to do so is an error, which causes the TFTP server to send an error packet. From RFC 1350 - THE TFTP PROTOCOL (REVISION 2), Section 4, In the next step, and in all succeeding steps, the hosts should make sure that the source TID matches the value that was agreed on in steps 1 and 2. If a source TID does not match, the packet should be discarded as erroneously sent from somewhere else. An error packet should be sent to the source of the incorrect packet, while not disturbing the transfer. Once the UDPv4 protocol driver has been assigned a local port number (which happens on the first core_udp_connect()) reuse that number until core_udp_close() time. Signed-off-by: Matt Fleming <matt.fleming@intel.com>
Users are reporting hitting the following error when typing 'make installer', make[3]: *** No rule to make target `efi32/core/codepage.o', needed by `syslinux.so'. Stop. But there's no actual need to build any installers for EFI (none exist), especially not since the stuff in utils/ is already built for BIOS and doesn't need building again. Signed-off-by: Matt Fleming <matt.fleming@intel.com>
Welcome to the 6.01 release cycle. Signed-off-by: Matt Fleming <matt.fleming@intel.com>
The BIOS firmware backend is missing a .load_linux pointer, and so anyone trying to boot a Linux kernel under BIOS is hitting the following error message, "No linux boot function registered for firmware" The usual way to handle this kind of abstraction would be to move bios_load_linux() to core/bios.c and assign it to .load_linux, but that would necessitate pulling the movebits and shuffler code into the core. For now, leave the BIOS loader where it is and use it as the default. In future we will want to move this to BIOS-specific code (though not necessarily in the core) because, by having it in the generic loader code, it is currently being built for the EFI backends even though it is never used. Signed-off-by: Matt Fleming <matt.fleming@intel.com>
I missed some symbols previously that are required to be exported when loading vesamenu.c32. Signed-off-by: Matt Fleming <matt.fleming@intel.com>
We need to provide a __bcopyxx_len symbol for EFI because it's referenced in generic code in libcom32.c32. Without this change, libcom32.c32 will fail to load under EFI. Signed-off-by: Matt Fleming <matt.fleming@intel.com>
efi/main.c: In function ‘find_addr’: efi/main.c:527:9: warning: unused variable ‘addr’ [-Wunused-variable] efi/main.c:524:13: warning: unused variable ‘status’ [-Wunused-variable] Signed-off-by: Matt Fleming <matt.fleming@intel.com>