H. Peter Anvin [Thu, 10 Sep 2009 21:25:14 +0000 (14:25 -0700)]
gpxecmd: module to send an arbitrary command to gPXE
Simple module to send an arbitrary command to gPXE, assuming the gPXE
CLI API is available.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Sat, 5 Sep 2009 00:26:51 +0000 (17:26 -0700)]
core: clean up SerialIRQPort when uninstalling the interrupt system
Make sure SerialIRQPort is zero when there is no active interrupt
system installed.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Sat, 5 Sep 2009 00:19:09 +0000 (17:19 -0700)]
core: when enabling serial IRQs, unmask interrupts at the PIC
Some BIOSes only enable interrupts in active use by the BIOS at the
PIC, which presumably don't include the serial ports. As such, enable
the interrupt lines at the PIC and hope to $DEITY we don't hang due to
a stuck interrupt line.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Fri, 4 Sep 2009 23:38:21 +0000 (16:38 -0700)]
core: remove overzealous calls to cleanup_hardware
cleanup_hardware is not needed in core/runkernel.inc: cleanup_hardware
is already invoked from replace_bootstrap. Furthermore,
cleanup_hardware is defined to preserve all registers, so there is no
need to push/pop ax around it.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Fri, 4 Sep 2009 22:46:43 +0000 (15:46 -0700)]
core: add missing calls to cleanup_hardware
As we're starting to rely more and more on a modified hardware state,
it is really important to clean up on all exit paths.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Fri, 4 Sep 2009 22:09:50 +0000 (15:09 -0700)]
pxelinux: workaround for EFI CSM bug - BEV stack overwrite on localboot
On at least some machines with an EFI-based BIOS, the Compatibility
Service Module (CSM) puts the BEV stack at an address where the boot
loader is likely to overwrite it. Implement a workaround which forces
INT 18h instead if an EFI signature is detected in memory.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Wed, 26 Aug 2009 04:19:26 +0000 (21:19 -0700)]
NEWS: update for 3.83
Pierre-Alexandre Meyer [Sat, 22 Aug 2009 18:08:04 +0000 (11:08 -0700)]
hdt: enhance error reporting in the CLI (disk)
parse_partition_table can fail for two reasons: disk I/O
error or unrecognized partition layout (non msdos, corrupt, ...).
In the latter case, we were displaying
Error parsing disk 0x80
parse_partition_table: error 0
which is confusing. We now display
Disk 0x80: unrecognized partition layout
or the errno_disk in case of I/O errors.
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
Pierre-Alexandre Meyer [Sat, 22 Aug 2009 17:34:40 +0000 (10:34 -0700)]
hdt: make get_error behaves like perror
This simplifies buffer handling.
Misc.: clean some old error code handling.
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
Pierre-Alexandre Meyer [Sat, 22 Aug 2009 16:42:37 +0000 (09:42 -0700)]
hdt: fix CLI crash when parsing unsupported geometries
Remove extra free() calls: these were needed before the introduction
of the disklib errno. We apparently forgot to remove them.
Reported-by: Gert Hulselmans <hulselmansgert@gmail.com>
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
erwan [Thu, 20 Aug 2009 04:24:16 +0000 (06:24 +0200)]
hdt: 0.3.4
Impact: New release
0.3.4 is now out !
Pierre-Alexandre Meyer [Thu, 20 Aug 2009 04:20:40 +0000 (21:20 -0700)]
hdt: fix disklib display
We currently display the first sector after the end of an extended.
Fix it (-1).
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
Pierre-Alexandre Meyer [Thu, 20 Aug 2009 04:12:57 +0000 (21:12 -0700)]
disklib: fix extended partition code
gert1 reported some issues using disklib with HDT. As it turned out, the
code that iterated through ebrs in an extended partition was broken: ebr offsets
need to be relative to the start of that partition.
Misc.: fix memory leak.
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
Signed-off-by: Erwan Velu <erwan.velu@free.fr>
erwan [Wed, 19 Aug 2009 22:19:44 +0000 (00:19 +0200)]
hdt: Fixing memory corruption
Impact: Improve stability
In some case, sectors_to_size can return a 8 char long string like "1000
MiB". In such case, some buffer were corrupted.
This is now fixed by a size[9] instead of size[8].
erwan [Wed, 19 Aug 2009 21:50:50 +0000 (23:50 +0200)]
Merge branch 'master' of git://git./boot/syslinux/syslinux into origin/master
Erwan Velu [Wed, 19 Aug 2009 19:17:49 +0000 (21:17 +0200)]
Merge commit 'pam/disklib-fixes'
H. Peter Anvin [Tue, 18 Aug 2009 02:06:01 +0000 (19:06 -0700)]
MEMDISK: generate map files
Generate map files to make debugging less painful.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Mon, 17 Aug 2009 02:51:38 +0000 (19:51 -0700)]
memdisk: make the stack size configurable
Make it possible to configure the stack size on the command line,
so that we can rule that out in case of problems.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Mon, 17 Aug 2009 19:26:08 +0000 (12:26 -0700)]
menu: initialize cm (current menu) before empty check
The empty check in the menu system used "cm", but before "cm" had
actually been initialized. Move the "cm" initialization earlier so we
do the right thing here.
Reported-by: CKSoon <cksoon79@yahoo.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Sat, 15 Aug 2009 00:22:01 +0000 (17:22 -0700)]
Merge commit 'origin/master'
H. Peter Anvin [Fri, 14 Aug 2009 05:31:13 +0000 (22:31 -0700)]
menu: fix Ctrl-W (word erase) when editing a command line
The Ctrl-W (word erase) key was broken: it would think the command
line got longer, not shorter, resulting in havoc.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Thu, 13 Aug 2009 22:27:46 +0000 (15:27 -0700)]
Revert "Run Nindent on com32/include/pngconf.h"
This reverts commit
6361b71f87466877f8fd1273240e8fa4460f8d05.
<pngconf.h> is part of libpng, it should never have been Nindented.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Thu, 13 Aug 2009 22:26:49 +0000 (15:26 -0700)]
Revert "Run Nindent on com32/include/png.h"
This reverts commit
ab7b6c3380cd9188edbc99693d5562dcca055317.
<png.h> is part of libpng and hence externally maintained; it should
not have been Nindented.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Mon, 10 Aug 2009 18:18:12 +0000 (11:18 -0700)]
zlib: remove unused sample programs
Remove unused sample programs (example.c, minigzip.c); we don' tuse
them and they are just noise.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Sun, 9 Aug 2009 22:52:26 +0000 (15:52 -0700)]
zlib: kill gzio.c
Remove gzio.c, which is a zlib-specific stdio-like interface. We
implement this in the actual filesystem layer, which is much cleaner.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Pierre-Alexandre Meyer [Thu, 6 Aug 2009 04:31:13 +0000 (21:31 -0700)]
hdt: Remove extra more_printf calls
Let's rather use string concatenation.
Reported-by: Shao Miller <Shao.Miller@yrdsb.edu.on.ca>
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
Pierre-Alexandre Meyer [Thu, 6 Aug 2009 04:03:04 +0000 (21:03 -0700)]
gpllib: Introduce typedefs for callbacks
typedefs are evil but useful for function pointers as it makes them more
readable and maintainable.
This fixes a bug by the way: we had
void *callback(struct driveinfo *, struct part_entry *, int, int)
where we should have had
void (*callback)(struct driveinfo *, struct part_entry *, int, int)
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
Pierre-Alexandre Meyer [Thu, 6 Aug 2009 02:19:59 +0000 (19:19 -0700)]
hdt: Don't show partition header if not needed
Don't blindly print the partition header: check if we are able to iterate
through them first.
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
Pierre-Alexandre Meyer [Wed, 5 Aug 2009 22:59:57 +0000 (15:59 -0700)]
hdt: Fix disk menu to use the new disklib interfaces
Adapt the disk menu to use the new disklib interfaces. This fixes bugs related
to the display of partition numbering and swsusp signature.
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
Pierre-Alexandre Meyer [Wed, 5 Aug 2009 22:11:01 +0000 (15:11 -0700)]
hdt: Make the menu and cli use the new disk interfaces
The disklib interfaces have changed (errno_disk introduced). Update
these changes in hdt (nobody else use them).
Misc: fix get_error declaration.
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
Pierre-Alexandre Meyer [Wed, 5 Aug 2009 22:09:47 +0000 (15:09 -0700)]
gpllib: More errno fixes
* Fix header files due to previous commit
* Prefix disk errnos with ED to avoid conflicts with classic errno
* Fix swsusp code to use errno_disk
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
Pierre-Alexandre Meyer [Wed, 5 Aug 2009 20:49:37 +0000 (13:49 -0700)]
gpllib: Enhance Makefile
Leverage Make built-ins rather than listing manually all objects files to
compile.
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
Pierre-Alexandre Meyer [Wed, 5 Aug 2009 20:48:54 +0000 (13:48 -0700)]
gpllib: Fix compilation
ata.c was missing the inttypes.h header.
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
Pierre-Alexandre Meyer [Wed, 5 Aug 2009 19:42:37 +0000 (12:42 -0700)]
disklib: Change read/write interface
Get rid of the error pointer and use a errno-like error reporting mechanism.
Intent is to make these more like the standard read/write system calls.
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
Pierre-Alexandre Meyer [Wed, 5 Aug 2009 19:39:21 +0000 (12:39 -0700)]
disklib: Fix write_verify_sectors
Compare the entire chunk (all sectors), not only the first one.
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
Pierre-Alexandre Meyer [Wed, 5 Aug 2009 03:23:16 +0000 (20:23 -0700)]
Merge commit 'hpa/master' into for-erwan
Conflicts:
com32/cmenu/libmenu/com32io.h
com32/gplinclude/dmi/dmi.h
com32/gplinclude/dmi/dmi_base_board.h
com32/gplinclude/dmi/dmi_memory.h
com32/gplinclude/dmi/dmi_system.h
com32/gpllib/dmi/dmi.c
com32/gpllib/dmi/dmi_memory.c
com32/include/sys/pci.h
com32/lib/pci/scan.c
com32/modules/Makefile
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
Pierre-Alexandre Meyer [Wed, 5 Aug 2009 00:35:32 +0000 (17:35 -0700)]
gpllib: refactor partitions detection
Re-implement the logic to parse extended partitions to correctly go through the
linked list of partitions.
Updated the CLI in hdt and tested it both on Qemu (Debian Lenny installation)
and on DELL E6400 (6 partitions). The output seemed sane.
Signed-off-by: Pierre-Alexandre Meyer <pierre@mouraf.org>
H. Peter Anvin [Mon, 3 Aug 2009 21:14:40 +0000 (14:14 -0700)]
NEWS: document isohybrid -id fixed.
Steffen Winterfeldt [Mon, 3 Aug 2009 13:21:02 +0000 (15:21 +0200)]
isohybrid: convert -id to a binary string
Shouldn't 'isohybrid -id ID' write ID as 32bit number? Currently it ends up
as string messing up the mbr.
H. Peter Anvin [Sun, 2 Aug 2009 18:59:03 +0000 (11:59 -0700)]
memdisk: remove unnecessary %ifdef
Remove an %ifdef that the macros take care of anyway.
H. Peter Anvin [Sun, 2 Aug 2009 18:56:38 +0000 (11:56 -0700)]
memdisk: don't set CX:DX for INT 13h AH=15h for floppies
Win98SE DOS apparently hangs if INT 13h, AH=15h returns the capacity
in CX:DX for floppies... *unless* EDD is enabled. Confused yet?
This patch also adds some more tracing framework.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Miller, Shao [Sun, 2 Aug 2009 05:33:36 +0000 (01:33 -0400)]
MEMDISK: additional EDD DPT Fields
We incorrectly used "equ" instead of "dd" (or "db") for strings in
memdisk.inc. Oops! My mistake in my 'freedos' branch. Silly
mistake. - Shao
H. Peter Anvin [Fri, 31 Jul 2009 15:27:58 +0000 (08:27 -0700)]
memdisk: correct the types for the EDD DPT, and write a real one
Fix the sizes of some of the fields in the EDD DPT. Write a real one
to the best of our ability.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Fri, 31 Jul 2009 15:24:27 +0000 (08:24 -0700)]
memdisk: don't write the EDD DPT unless we are in EDD mode...
Don't write the EDD DPT unless we're actually in EDD mode! Otherwise
we end up corrupting memory beyond the defined code. This was the
reason
5ca2764460883f5e9a5bd83db00f242ea1272e33 worked, by moving
variables out of the corruption zone.
Debugged-by: Shao Miller <Shao.Miller@yrdsb.edu.on.ca>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Fri, 31 Jul 2009 05:40:06 +0000 (22:40 -0700)]
Merge commit 'andytim/freedos'
H. Peter Anvin [Fri, 31 Jul 2009 05:27:16 +0000 (22:27 -0700)]
graphics: actually do 32-bit writes to VGA
Use "rep movsd" to copy data to VGA memory, not "rep movsb".
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Shao Miller [Fri, 31 Jul 2009 05:27:05 +0000 (01:27 -0400)]
[memdisk] Additional EDD Device Parameter Table fields
Some additional fields from the EDD-4 spec. draft for the Device Parameter
Table have been added into the structure in setup.c and memdisk.inc. These
were added in the hopes of resolving a FreeDOS MEMDISK bug on IBM ThinkPads.
H. Peter Anvin [Fri, 31 Jul 2009 00:44:12 +0000 (17:44 -0700)]
graphics: trivial optimization of old-style graphic mode
Trivially optimize the old-style graphics by using another level of
temporary buffering in bitplane format, thereby enabling the use of
rep movsd to write the contents to the screen. This can speed things
up substantially in virtualization environments.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Thu, 30 Jul 2009 19:31:10 +0000 (12:31 -0700)]
NEWS: document Thinkpad fix
H. Peter Anvin [Thu, 30 Jul 2009 19:28:45 +0000 (12:28 -0700)]
memdisk: move temporary variables away from the EDD data block
Move a couple of variables away from the EDD data block. Without this
change, FreeDOS doesn't boot on Thinkpads. Unfortunately, this should
have been a null change, so it is really not clear why this happens.
Stack overflows have been ruled out. Probably need to examine the
FreeDOS kernel in some detail.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Thu, 30 Jul 2009 18:24:25 +0000 (11:24 -0700)]
memdisk: handle recursive INT 13h invocation
Some systems, e.g. a lot of Thinkpads, invoke INT 13h recursively from
within the stack itself. This is generally a bad idea as it re-runs
any INT 13h hooks (including the ones that DOS installs!), but it
should work. Keep a counter of recursive invocations and simply
bypass the whole system including the DL shift if the invocation is
recursive.
With this, MS-DOS boots on a Thinkpad T61. FreeDOS is still broken,
though, but apparently in a different manner...
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Wed, 29 Jul 2009 03:19:14 +0000 (20:19 -0700)]
PXELINUX: allow filenames up to 251 characters long
At least one users really need this, so be nice and provide it...
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Mon, 27 Jul 2009 18:15:43 +0000 (11:15 -0700)]
Merge commit 'syslinux-3.83-pre1'
Conflicts:
com32/modules/chain.c
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Fri, 24 Jul 2009 00:24:13 +0000 (17:24 -0700)]
NEWS: document chain.c32 ISOLINUX fix.
H. Peter Anvin [Fri, 24 Jul 2009 00:23:04 +0000 (17:23 -0700)]
chain.c32: set DL properly when loaded from ISOLINUX
When loaded from ISOLINUX, we still have a valid boot drive. Set it!
This permits the use of chainloaded boot loaders.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Sebastian Herbszt [Sat, 18 Jul 2009 15:06:32 +0000 (17:06 +0200)]
gfxboot: remove obsolete message regarding pxelinux support
PXELINUX is supported by the gfxboot module.
Signed-off-by: Sebastian Herbszt <herbszt@gmx.de>
Sebastian Herbszt [Sat, 18 Jul 2009 15:05:22 +0000 (17:05 +0200)]
gfxboot: parse TIMEOUT keyword
Parse the TIMEOUT keyword from the config file and pass it to gfxboot core.
Signed-off-by: Sebastian Herbszt <herbszt@gmx.de>
Sebastian Herbszt [Sat, 18 Jul 2009 15:03:29 +0000 (17:03 +0200)]
gfxboot: only call gfxboot core if labels are specified
Only call gfxboot core if labels are specified in the config file.
Signed-off-by: Sebastian Herbszt <herbszt@gmx.de>
Sebastian Herbszt [Sat, 18 Jul 2009 15:01:19 +0000 (17:01 +0200)]
gfxboot: display error message if bootlogo file not found
Display an error message if the user specified bootlogo file is not found.
Signed-off-by: Sebastian Herbszt <herbszt@gmx.de>
Sebastian Herbszt [Sat, 18 Jul 2009 14:53:09 +0000 (16:53 +0200)]
gfxboot: change handling of keywords
Makes handling of keywords more flexible.
Signed-off-by: Sebastian Herbszt <herbszt@gmx.de>
H. Peter Anvin [Sun, 5 Jul 2009 21:15:12 +0000 (14:15 -0700)]
dosutil/Makefile: more complete structure
More complete and consistent structure for dosutil/Makefile.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Sun, 5 Jul 2009 08:04:50 +0000 (01:04 -0700)]
dosutil: update mdiskchk, add Makefile
Update mdiskchk and add a Makefile. Note that since this needs
OpenWatcom to compile, we still check in the binary file (which is
quite small, anyway.)
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Sat, 27 Jun 2009 00:44:22 +0000 (17:44 -0700)]
Create a dosutil subdirectory and move mdiskchk there
DOS utilities aren't really "samples". Create a dedicated "dosutil"
directory, and move mdiskchk there for now.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Gilles Espinasse [Sat, 20 Jun 2009 00:21:19 +0000 (17:21 -0700)]
extlinux: remove duplicate "const char *program;"
Remove redundant declaration.
H. Peter Anvin [Sun, 14 Jun 2009 23:21:10 +0000 (16:21 -0700)]
Merge branch 'syslinux-3.8x'
H. Peter Anvin [Sun, 14 Jun 2009 23:03:54 +0000 (16:03 -0700)]
NEWS: document chainloading fix
H. Peter Anvin [Sun, 14 Jun 2009 23:01:30 +0000 (16:01 -0700)]
PXELINUX: when chaining an NBP, restore *all* registers
On at least ASUS A8N-E, unless *all* registers are restored to pre-PXE
status, localboot doesn't work. We were clobbering DS, ESI and EDX
when chainloading an NBP, which meant that localboot wouldn't work in
the chainloaded NBP. Fix this.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Sun, 14 Jun 2009 21:32:42 +0000 (14:32 -0700)]
bootsect: zero memory above a chainloaded boot sector/NBP
Windows RIS has been reported to make inappropriate use of data found
in uninitialized memory. To avoid that, clear memory between
a chainloaded boot sector or NBP and Free Base Memory.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Wed, 10 Jun 2009 05:50:55 +0000 (22:50 -0700)]
com32.h: cleanups, make OFFS_VALID() take the pointer size into account
Clean up some of the constructs in com32.h, and make OFFS_VALID() take
the pointer type into account for the range check -- the whole object
needs to fit within the segment.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Tue, 9 Jun 2009 18:34:00 +0000 (11:34 -0700)]
memdisk: don't use the stack before bootstrap
Under some certain circumstances, it might not be safe to use the
stack as it was set up, so use a jump instead of a call to invoke the
bootstrap.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Tue, 9 Jun 2009 17:49:31 +0000 (10:49 -0700)]
Merge commit 'syslinux-3.82'
Conflicts:
com32/menu/menumain.c
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Tue, 9 Jun 2009 17:19:25 +0000 (10:19 -0700)]
pxelinux: actually stop parsing on a null option
Stop the OACK parsing on a null option, even if it isn't the very
first option.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Tue, 9 Jun 2009 15:51:45 +0000 (08:51 -0700)]
pxelinux: revert to using TFTP error 8
The main reason for ERROR is OACK parsing failure, so revert to using
error code 8 (but a fixed text error.)
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Tue, 9 Jun 2009 14:29:13 +0000 (07:29 -0700)]
pxelinux: cleaner test for OACK trailing null bytes
Instead of looking for a string of null bytes at the end of the OACK
string, simply abort parsing if we run into a null byte where an
option is expected; either we are seeing junk at the end of the
packet, or we are hopelessly confused about how to make sense of the
rest of the packet -- in either case, ignoring is the "liberal"
option.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Tue, 9 Jun 2009 05:19:41 +0000 (22:19 -0700)]
PXELINUX: handle OACK packets with extra NULs, cleaner TFTP error
There are apparently TFTP servers in the field which will send OACK
packets with extra NUL bytes appended at the end. If we find an OACK
packet where the only thing left at some point during processing is
NULs, then just consider the packet processed.
We have reported all TFTP protocol errors as "tsize required", which
is definitely not true anymore. Change error code to 0 (undefined)
and the error string to "TFTP error". When this code gets converted
to C we'll do better.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Mon, 8 Jun 2009 03:59:52 +0000 (20:59 -0700)]
memdisk: additional cleanups
Additional stylistic cleanups. Rename "syscall" to "intcall" (we
can't call it __intcall without clashing with com32.h); use macros
instead of copying variables to different places with only the type
being different.
Also, only change the rm/pm jump instructions when actually relocating
the code.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Mon, 8 Jun 2009 01:47:35 +0000 (18:47 -0700)]
memdisk: add missing memmove.S
We need memmove.S in the memdisk directory...
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Mon, 8 Jun 2009 01:46:59 +0000 (18:46 -0700)]
memdisk: minor cleanups
Minor cleanups to the source base; make die() also print an error
message and move die() into conio.c.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Mon, 8 Jun 2009 01:25:09 +0000 (18:25 -0700)]
memdisk: relocate real-mode code before booting
Relocate the real-mode code before booting. This allows the target
bootstrap to be loaded at an arbitrary address, not necessarily
0x7c00, and to be almost arbitrarily long.
Add some initial infrastructure for other bootstrap addresses, too.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Mon, 8 Jun 2009 00:13:58 +0000 (17:13 -0700)]
memdisk: move rm_args initialization to C code; export rm_size
Move the rm_args initialization (and the associated sti) to C code.
Export the total size of the real-mode code to the protected-mode
code.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Mon, 8 Jun 2009 00:06:59 +0000 (17:06 -0700)]
memdisk: set up a limited-size real-mode stack
Limit the size of the real-mode stack. This should limit the amount
of memory needed for a relocated real-mode code.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Mon, 8 Jun 2009 00:04:03 +0000 (17:04 -0700)]
memdisk: remove hard-coded assumptions for a specific RM segment
Remove some hard-coded assumptions about having an RM segment at a
specific location. This should allow us to relocate the RM code once
the PM code knows where DOS free memory ends. This in turn allows the
PM code to load the boot sector almost anywhere it wants to go in low
memory, as is required for 100% compliant El Torito support.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Sun, 7 Jun 2009 22:16:24 +0000 (15:16 -0700)]
memdisk: remove the requirement that CS_BASE < 64K
Remove the requirement that CS_BASE is below 64K, based on its use in
the A20 test code. This means it can now be located anywhere in high
memory, but it is still a constant.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Sun, 7 Jun 2009 22:12:39 +0000 (15:12 -0700)]
memdisk: move the heap to high memory
Instead of using a heap arbitrarily allocated in low memory, move it
to a chunk of bss in high memory.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Sun, 7 Jun 2009 21:44:30 +0000 (14:44 -0700)]
memdisk: don't hard-code 0:7C00 as the entry point
Make it possible for the setup code to override SS:SP and CS:IP.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Sun, 7 Jun 2009 21:36:07 +0000 (14:36 -0700)]
memdisk: pass in the address of the real-mode code
Pass in the address of the real-mode code instead of hard-coding it in
two separate places.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Fri, 5 Jun 2009 00:24:58 +0000 (17:24 -0700)]
core: prevent buggy INT 13h from leaving IF=0
If it can happen for CD-ROM BIOSes, it can probably happen elsewhere,
too; make sure we don't leave interrupts disabled after broken INT 13h
calls.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Sergey Vlasov [Thu, 4 Jun 2009 15:14:48 +0000 (19:14 +0400)]
simple menu: make ONTIMEOUT work with MENU HIDDEN
The command invoked when no keys were pressed for the specified time
(ONTIMEOUT) may be different from the menu item initially selected
when the menu is displayed (DEFAULT or MENU DEFAULT). Unfortunately,
this did not work together with MENU HIDDEN (which is exactly the case
when having a separate ONTIMEOUT command makes the most sense).
Signed-off-by: Sergey Vlasov <vsu@altlinux.ru>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Mon, 1 Jun 2009 04:53:53 +0000 (21:53 -0700)]
Merge branch 'syslinux-3.8x'
Conflicts:
NEWS
H. Peter Anvin [Fri, 29 May 2009 22:47:20 +0000 (15:47 -0700)]
bin2hex: actually account for the number of bytes written
Unlike C, in Perl printf() doesn't return the number of bytes written.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Mon, 1 Jun 2009 04:29:59 +0000 (21:29 -0700)]
isohdpfx: actually generate the stack frame isolinux expects
Actually generate the stack frame isolinux expects. This is not the
optimal order, but it appears to be what both earlier and later
versions of isolinux expects...
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Mon, 1 Jun 2009 04:29:12 +0000 (21:29 -0700)]
isolinux: can't clobber edx when dx contains the drive number...
Use ebx to hold the upper half of the partition offset, not edx.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Mon, 1 Jun 2009 03:25:43 +0000 (20:25 -0700)]
NEWS: add changes since 3.81
H. Peter Anvin [Mon, 1 Jun 2009 03:23:58 +0000 (20:23 -0700)]
core: when calling abort_check, we're not idle
When we call abort_check, we're not idle even though we are polling
the keyboard. Handle that by calling reset_idle near the top.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Mon, 1 Jun 2009 03:19:33 +0000 (20:19 -0700)]
isolinux: bsHidden can't be in data; it is set before checksumming
We cannot put bsHidden in initializated data, as it is set before we
run the global checksumming.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Mon, 1 Jun 2009 03:19:17 +0000 (20:19 -0700)]
version: next version will be 3.82
H. Peter Anvin [Mon, 1 Jun 2009 03:13:08 +0000 (20:13 -0700)]
isolinux: handle systems which disables interrupts in El Torito
At least one system has been identified which disables interrupts when
El Torito INT 13h is executed. Thus, save/restore IF around INT 13h.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Fri, 29 May 2009 22:47:20 +0000 (15:47 -0700)]
bin2hex: actually account for the number of bytes written
Unlike C, in Perl printf() doesn't return the number of bytes written.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
H. Peter Anvin [Fri, 29 May 2009 22:41:43 +0000 (15:41 -0700)]
Assume the next version will be 3.82
H. Peter Anvin [Fri, 29 May 2009 22:25:40 +0000 (15:25 -0700)]
Add back spaces around ellipses in case statements
GNU C allows the ellipsis (...) to be used in case statements to
indicate a range. However, it requires spaces around it to be valid.
Nindent stripped those off, put them back.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>