| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
| |
Fix missing -M from NASM dependency generation calls; adopt the
uniform stanza "-M -DDEPEND" even for sites that were already correct.
Also, use ">" not ">>" for obvious reasons...
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
|
|
|
|
|
|
|
| |
Unify dependency generation and move common rules into
MCONFIG.embedded.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
|
|
|
|
| |
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
|
|
|
|
|
|
| |
Print the source of the geometry determination.
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
|
|
|
|
|
|
|
| |
Fix the MBR geometry extraction; there was an off-by-2 error in the
MBR signature offset.
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
|
|
|
|
| |
Fix build failure due to incorrect comment syntax in memdisk.ld
|
|
|
|
|
|
|
| |
Update copyright notices. Per agreement with my new employer, Intel
Corporation, add Intel copyright notices where appropriate.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
|
|
|
|
|
|
|
|
|
| |
Impact: aestetics
Change "and contributors" to "et al" so that the message actually fits
on one line. This is unfortunate, but what can you do...
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
|
|
|
|
|
|
|
| |
Impact: bug fix
Int13FuncsCnt is a compile-time constant, not a memory location. This
is a rather important distinction.
|
|
|
|
|
| |
Clarify that there are nowadays more copyright holders than just
myself.
|
|
|
|
| |
Just a sprinkling of comments adding further description to the code.
|
|\ |
|
| |
| |
| |
| |
| |
| |
| | |
Deal with issues that made older versions of nasm:
a) prevent "make spotless" from running
b) generate bad dependencies (no final newline)
|
| |
| |
| |
| | |
This seems to have better behavior for "make spotless".
|
| |
| |
| |
| |
| | |
This reverts commit 8f3919c8f5dd37ccd3b7565d32ef1dac8b2522d5.
Wrong fix.
|
| |
| |
| |
| |
| | |
Older versions of NASM puke if there is an %include of a non-existent
file during dependency generation. Make it more forgiving.
|
| |
| |
| |
| |
| | |
Some CPUs really want a reasonable alignment for a rep;stosl, so give
it to them...
|
| |
| |
| |
| |
| | |
It is just too confusing to have a .bss section that isn't zeroed, to
explicitly zero it.
|
| |
| |
| |
| |
| | |
Move the floppy disk geometry generation code into parallel with the
MBR code, which pretty much mimics the previous behavior.
|
| |
| |
| |
| |
| |
| |
| |
| | |
We can't rely on the disk type to say if this is FAT vs MBR, since we
haven't determined that yet. Thus, move the FAT test up before the
MBR test; both are heuristic, but the FAT test has much less
probability of false positives, and can be further tightened if
needed.
|
| |
| |
| |
| |
| |
| | |
If a "large floppy" image is formatted with a FAT filesystem, we can
use the headers in the FAT image to derive the geometry. This is nice
and user-friendly, so do it that way.
|
| |
| |
| |
| |
| |
| | |
The memdisk install code previously controlled EDD features by controlling
how many INT 13 functions were available. Since we have EDD and non-EDD
versions of the memdisk hooking code now, this is no longer needed.
|
| |
| |
| |
| |
| | |
The descriptor table limit was off by one. Not that it matters, but
be picky about these sort of things.
|
| |
| |
| |
| |
| |
| | |
With the creation of wrapper assembly scripts, there is no reason for
the indirection of WITH_EDD -> EDD; we can just set the EDD variable
directly.
|
|/
|
|
|
|
|
|
|
|
|
|
| |
This patch "sets the stage" for the memdisk kernel to include different
"installables": versions of the memdisk hooks.
This patch splits the case of the EDD-compiled hook from the non-EDD-compiled
hook. Both are available inside the kernel, and the kernel chooses which one
to install based on the command-line.
It's possible that other, slightly modified versions of the memdisk hook could
be useful to have...
|
|
|
|
|
|
|
|
|
|
|
| |
In safeint mode, don't bother testing for unchanged INT 15h. It
points back at ourselves, anyway, and we end up doing, in the end, the
same thing as the safeint stub would do.
Furthermore, the test was broken for years, and so this is actually
closer to what we originally tested.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
|
|
|
|
|
| |
Signed-off-by: Sergey Vlasov <vsu@altlinux.ru>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
|
|
|
|
|
|
| |
Drop a newline we didn't need after all.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
|
|
|
|
|
|
|
|
|
|
| |
There is enough evidence that "safeint" is a decent compromise between
The Right Thing[TM] and dealing with the reality of ultra-broken
DOSen. Make it the default.
Clean up some of the command-line parsing code and status printing.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
|
|
|
|
|
|
|
|
| |
Pre-initialize the e820 buffer to all zero, except for the active bit;
we have found that BIOSes will make assumptions about the pre-existing
value of the buffer and this seems the most conservative value.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
|
|
|
|
|
|
|
| |
Makefile rule for e820test that actually works. It's suboptimal, and
that is just fine.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
|
|
|
|
|
|
| |
Make it possible to build e820test again, to debug the e820 code.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
|
|
|
|
|
|
|
| |
Skip non-active ranges when looking for a suitable chunk of memory for
decompression.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
|
|
|
|
|
|
|
| |
Correct the assembly portion of the new e820 code, which had been
left, in effect, half finished.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Some blithering idiot thought it was a good idea to introduce
"extended attributes" for INT 15h, AX=E820h, and in doing so, breaking
compatibility with ALL E820 users out there. F*cking morons.
Implement handling of extended attributes in:
- e820 parsing in the core
- e820 parsing in libcom32
- e820 parsing *and proxying* in memdisk
The latter is the really painful one.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
|
|
|
|
|
|
| |
Unify common pieces to "embedded" targets (those that produce code
that runs neither in the host nor in a com32 environment); this fixes
the broken sample/ directory Makefile.
|
|
|
|
|
| |
Cleanup and centralize the Makefile system even more.
Fix a gcc 4.3 incompatibility in memdisk (definition of strlen).
|
|
|
|
|
| |
If git-describe comes up with a tag not starting with
syslinux-<version>- then leave it as-is.
|
|
|
|
| |
For cleaning commands, we need version.mk to be optional.
|
|
|
|
| |
Both core and memdisk wants to use the gen-id stuff, so let it.
|
|
|
|
| |
Centralize more (most) of the version number machinery to version.pl.
|
|
|
|
|
|
|
|
|
| |
Per the Interrupt list, treat INT 13 08 returning with CL=0 as a
failure, meaning single drive only.
If we find ourselves the only floppy drive, install a DPT into INT
1Eh. This appears to be needed for PC-DOS 7.0 to boot. This can be
overridden with the "nodpt" option, and forced with the "dpt" option.
|
|
|
|
| |
2008, not 2007
|
|
|
|
|
| |
UHCI supposedly wants a null command (0FFh) sent to the KBC as part of
the A20-flipping sequence. See if that works for us.
|
|
|
|
|
| |
Return exit values to go with the error messages; don't divide by zero
in case of range error.
|
| |
|
|
|
|
|
| |
Consider all disks < 4 MB to be floppies, and guess their geometry
algorithmically.
|
|
|
|
|
| |
Use the aligning memcpy/memset from libcom32. Switch memdisk to
compiling with -mregparm=3, except for the interfaces to assembly.
|
|
|
|
|
|
|
|
| |
Move source files out of the root directory; the root is a mess and
has become virtually unmaintainable. The Syslinux core now lives in
core/; the Linux and generic utilities has moved into utils/, and
copybs.com has moved into dos/; it had to go somewhere, and it seemed
as good a place as any.
|
|
|
|
| |
-fstack-protector flag enabled by default.
|