Now that we actually have the real backup header, we can safely write it
out without having the wrong LBA pointers and other such issues that
would come from writing the primary header to the backup header (ask me
how I know)
This commit introduced a lot of changes and seems to have caused a few
issues, including breaking crc32 generation. Revert it for now with the
intention to reimplement some of the improvements.
This reverts commit 9d7600df51.
If we are using C++, we might as well use scope-based memory management
for these things, cleaner and less error prone. Also removed
unnecessary gotos.
The lines:
if (instance == PRIMARY_GPT)
hdr_offset = block_size;
else {
hdr_offset = lseek64(fd, 0, SEEK_END) - block_size;
}
if (hdr_offset < 0) {
fprintf(stderr, "%s: Failed to get gpt header offset\n",
__func__);
goto error;
}
are error checked. But previously hdr_offset could never be less than
zero because it was unsigned.
getting attributes work, setting them (e.g. with mark_boot_successful)
doesn't seem to work properly.
The IOCTL required to set the boot LUN doesn't work, some kernel driver
is missing for it.
This is a port of the Android bootctrl HAL and Qualcomms GPT based
implementation for musl/glibc based systems running mainline linux
Currently it doesn't work because the gpt hdr which is read doesn't
seem to have valid data.
Caleb Connolly
Eric Curtin
alefnode
Richard Acayan