ld.bfd doesn't understand elf64-aarch64 but does have
elf64-littleaarch64. Switch to this so we can link kboot with it.
While here switch to the single format version. We are unlikely to
support booting from a big-endian Linux.
Reviewed by: imp, emaste
Sponsored by: Arm Ltd
Differential Revision: https://reviews.freebsd.org/D45258
Move the initialization of hostfs_root to be a bit sooner. While it
doesn't matter for the default case, we may want to use hostfs files
sooner.
Also, while we're here, remove kboot.conf. It duplicates the command
line and has proven difficult to use. It will be replaced by an early
script that can influence the state of the boot loader before we select
a device to boot from (including strongly suggesting which one to boot
from).
Sponsored by: Netflix
Move the console probing to as early as possible. There's no real
support for anything but hostcons, and setting it up early will show
other error messages.
ACPI and SMBIOS probing can be done just after we have the console, so
move it there. This allows other parts of the early code to use info
from that, as well as overriding and env vars set by these things on the
command line (smbios data may be wrong during initial development phases
as the automated way to populate per-board data may not be established,
etc).
Sponsored by: Netflix
If we are able to open /sys/firmware/fdt, but aren't able to read it,
fall back to /proc/device-tree. Remove comment that's not really true,
it turns out.
Sponsored by: Netflix
If we can read the UEFI memory map, go ahead and print the memory map.
While the kernel prints this with bootverbose, having it at this stage
is useful for debugging other problems.
Sponsored by: Netflix
Differential Revision: https://reviews.freebsd.org/D44287
The host_* syscalls are all raw Linux system calls, not the POSIX
wrappers that glibc / musl create. So we have to ranage change the
return value of host_llseek correctly to use the negative value hack
that all Linux system calls use.
This fixes a false positive error detection when we do something like
lseek(fd, 0xf1234567, ...); This returns 0xf1234567, which is a negative
value which used to trigger the error path. Instead, we check using the
is_linux_error() and store the return value in a long. Translate that
errno to a host errno and set the global errno to that and return
-1. lseek can't otherwise return a negative number, since it's the
offset after seeking into the file, which by definition is positive.
This kept the 'read the UEFI memory map out of physical memory' from
working on aarch64 (whose boot loader falls back to reading it since
there are restrictive kernel options that can also prevent it), since
the physical address the memory map was at on my platform was like
0xfa008018.
Sponsored by: Netflix
Differential Revision: https://reviews.freebsd.org/D44286
offset is signed. Copy it to the unsigned res before shifting. This
avoids any possible undefined behavior for right shifting signed
numbers. No functional change intended (and the code generated is the
nearly same for aarch64).
Sponsored by: Netflix
Differential Revision: https://reviews.freebsd.org/D44285
Linux has the convention of returning -ERRNO to flag errors from its
system calls. Sometimes other negative values are returned that are
success... However, only values -1 to -4096 (inclusive) are really
errors. The rest are either truncated values that only look negative (so
use long instead of int), or are things like addresses or legal unsigned
file offsets or similar that are successful returns. Filter out the
latter.
Sponsored by: Netflix
Linker set sections are included by default. No need to do so
explicitly. These were bogusly copied from the efi ldscripts. They were
there due to a workaround introduced in 2002 by peter@ for a gcc
upgrade, but whatever bugs necessitated it were filed by 2010 when
rpaulo@ imported the i386 support (though they were copied even though
the objcopy retained them correctly, the gcc bug having been
fixed). They've never been needed.
Sponsored by: Netflix
Reviewed by: tsoome, kevans
Differential Revision: https://reviews.freebsd.org/D44060
Make the pattern for ldscripts always be
arch/$MACHINE_ARCH/$MACHINE_ARCH.ldscript so we can add it from a
central Makefile. This also moves from ldscript.arch to arch.ldscript to
match the loader's new convention.
Sponsored by: Netflix
Reviewed by: tsoome, kevans
Differential Revision: https://reviews.freebsd.org/D44057
Don't assume that strategy is only called for read. Check the passed
flag for F_READ or F_WRITE and fail if it is neither. Open the disks for
writing and call host_read/host_write depending on that flag.
Sponsored by: Netflix
Reviewed by: kevans, gallatin
Differential Revision: https://reviews.freebsd.org/D44016
This wasn't updated when the other copies were updated. Make it
identical to efi code. We should likely refactor this (with userboot),
but they are all not quite identical.
Sponsored by: Netflix
Create a small wrapper around the new flua hash module so we can use it
here too. There's no 4th bindings, nor will they be created.
Sponsored by: Netflix
Differential Revision: https://reviews.freebsd.org/D43874
Now that we draw in the gfx bindings for all our interpreters only when
graphics support is compiled in, we can eliminate this from all the
loaders that don't have graphics support.
Sponsored by: Netflix
Reviewed by: kevans, jhb
Differential Revision: https://reviews.freebsd.org/D43905
host_syscalls.c: I've written, so put Netflix copyright on. It's
possible in the confusion that Nathan wrote the host_gettimeofday
implementation.
syscall_nr: These files likely can't enjoy copyright protection since
they are just facts (the per-arch Linux system calls), so add a note
they are in the public domain.
Sponsored by: Netflix
According to git blame I've 95%+ rewritten this file. Update copyright to
reflect that, but give nod to Nathan for the original I started with.
Sponsored by: Netflix
For aarch64 and amd64, we don't pull in anything from libkern, so we
don't need it in our path. However, powerpc needs ucmpdi2 from libkern,
so bring it into libkboot's build and omit it from loader.kboot.
Sponsored by: Netflix
Move the startup code from kboot/kboot to kboot/libkboot and add the
necessary infrastructure for it to build. move start_arch.h, a private
header for libkboot, over as well.
Sponsored by: Netflix
In anticipation of separating the library elements from the kboot
prorgam elements, move kboot down a level into a subdirectory. There
will be libkboot and include directories in subsequent commits,
mirroring other subsystems like i386 and efi.
Sponsored by: Netflix
Apply the following automated changes to try to eliminate
no-longer-needed sys/cdefs.h includes as well as now-empty
blank lines in a row.
Remove /^#if.*\n#endif.*\n#include\s+<sys/cdefs.h>.*\n/
Remove /\n+#include\s+<sys/cdefs.h>.*\n+#if.*\n#endif.*\n+/
Remove /\n+#if.*\n#endif.*\n+/
Remove /^#if.*\n#endif.*\n/
Remove /\n+#include\s+<sys/cdefs.h>\n#include\s+<sys/types.h>/
Remove /\n+#include\s+<sys/cdefs.h>\n#include\s+<sys/param.h>/
Remove /\n+#include\s+<sys/cdefs.h>\n#include\s+<sys/capsicum.h>/
Sponsored by: Netflix
Preemptively address a collision with LIBFDT (to be added in the future)
from src.libnames.mk, which gets included via bsd.progs.mk. No
functional change intended.
Reviewed by: imp
MFC after: 1 week
Sponsored by: Innovate UK
Differential Revision: https://reviews.freebsd.org/D42486
Add support for getting smbios from /sys/firmware/efi/systab, if
any. Add ptov mapping that uses mmap on /dev/mem to do the mapping with
64k pages (usually we only need 1 or two mappings).
Sponsored by: Netflix
Differential Revision: https://reviews.freebsd.org/D39792
Fix an off-by-one error that would mean we'd get stuck on the newline if
ACPI= wasn't first.
Sponsored by: Netflix
Differential Revision: https://reviews.freebsd.org/D39817
There's plenty of stack in kboot, so use it here rather than the
malloc/free dance.
Sponsored by: Netflix
Reviewed by: tsoome, kevans
Differential Revision: https://reviews.freebsd.org/D39416
If there's a kboot.conf, prase it after the command line args are
parsed. It's not always easy to get all the right command line args
depending on the environment. Allow an escape hatch. While we can't do
everything one might like in this file, we can do enough.
Sponsored by: Netflix
We were assuming that hostdisk_override was both a directory and a
file, which is not going to work very well. It's supposed to be a
single file, so recode it as such. Simplify erorr handling a little as
well and fix a return type-mismatch that doesn't matter for the
generated code (return NULL is the same as return false in this
context)
Sponsored by: Netflix
MIN works for any type, while min() is only for integers. So we were
rounding down to 0 since that's the size of 4GB truncated to an int.
Sponsored by: Netflix
When we're running on a FreeBSD host, we can't open /proc/iomem. So,
for now, just assume that we have 32GB of ram starting at 4GB.
Sponsored by: Netflix
Address two issues with current help file logic:
The existing condition prevents the common help file from being
installed when there are no additional help files defined. This results
in no loader.help on EFI platforms, for example.
Second, due to the fact that we build and install multiple loader types,
each successive install will clobber the previous loader.help. The
result is that we could lose type-specific commands, or possibly list
them in loaders that do not have such commands.
Instead, give each loader type a uniquely named help file. The EFI
loader will look for /boot/loader.help.efi, userboot will look for
/boot/loader.help.userboot, etc. The interpreter variant has no effect
on which help file is loaded.
This leaves the old /boot/loader.help unused.
Some credit for the final approach goes to Mathieu <sigsys@gmail.com>
for their version of the fix in https://reviews.freebsd.org/D22951.
PR: 267134
Reported by: Daniel O'Connor <darius@dons.net.au>
Reviewed by: imp
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D28591
Keep track of how much is used in the segment as we allocate it to the
application. Set memsz to 0 first, and increment it as used. Adjust the
bufsz before we call kexec so the kernel copies the right amount (it's
an error for bufsz to be bigger than memsz, so we set them == when we
retrieve the segment). Make sure we round to the page size, otherwise
kexec_load gets cranky.
Sponsored by: Netflix
Reviewed by: tsoome
Differential Revision: https://reviews.freebsd.org/D38315
Allocate a huge segment for the first kexec_load segments. We limit the
lessor of:
allocation to the size of the remaining memory segment
45% of available memory
95% of the memory we can allocate
This allows us to have really large RAM disks. We likely need to limit
this to the amount we actually used, though, since this can be a lot of
memory.
We have to do this complicated calculation for a few reasons: First, we
need 2 copies of the loaded kernel in the memory: The kernel can copy
everything to a temporary buffer. Next, malloc (via mmap) is limited to
a certain amount due to over commit, so we have to not allocate all we
can (only most of what we can).
Sponsored by: Netflix
Reviewed by: tsoome
Differential Revision: https://reviews.freebsd.org/D38314
kboot_get_phys_load_segment is defined in kboot.h, so remove them from
the .c files.
Sponsored by: Netflix
Reviewed by: tsoome
Differential Revision: https://reviews.freebsd.org/D38310
Try to open /dev/mem to read in the UEFI memory map. If we can't, then
we'll read it in the trampoline.
Retain reading in /proc/iomem to find reserved areas in Linux. We need
to know them for good places to put the kernel. These are not reflected
in the UEFI memory map. However, we should not adjust the UEFI memory
map since these reserved areas of the Linux kernel are free to be used
once we enter the kexec trampoline...
Sponsored by: Netflix
Reviewed by: tsoome, kevans, andrew
Differential Revision: https://reviews.freebsd.org/D38264
