platform/upstream/syslinux.git
10 years agoldlinux/readconfig.c: remove bail: from do_include()
Gene Cumm [Sat, 25 Jan 2014 00:44:51 +0000 (19:44 -0500)]
ldlinux/readconfig.c: remove bail: from do_include()

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
10 years agoAdd const qualifiers to loadleys
Gene Cumm [Sat, 25 Jan 2014 12:37:24 +0000 (07:37 -0500)]
Add const qualifiers to loadleys

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
10 years agolocalboot: Clear register image before using
H. Peter Anvin [Tue, 21 Jan 2014 15:15:52 +0000 (07:15 -0800)]
localboot: Clear register image before using

Using a register image on the stack without initializing it is not a
good idea.

Reported-by: Erwan Velu <erwanaliasr1@gmail.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
10 years agoMerge remote-tracking branch 'origin/syslinux-4.xx' into syslinux-5.xx
H. Peter Anvin [Sun, 19 Jan 2014 00:11:39 +0000 (16:11 -0800)]
Merge remote-tracking branch 'origin/syslinux-4.xx' into syslinux-5.xx

Resolved Conflicts:
version

Signed-off-by: H. Peter Anvin <hpa@zytor.com>
10 years agoMakefiles: don't include *.tmp
H. Peter Anvin [Sun, 19 Jan 2014 00:09:19 +0000 (16:09 -0800)]
Makefiles: don't include *.tmp

Since checkin:

bd09a6d828fa Major Makefile cleanups; gcc 4.3.0 compatiblity

... we include *.tmp into the Makefiles as well as .*.d.  This seems
to have been a mistake in made when adding *.tmp to cleanup rules,
probably using a sed script.

This causes problems, because *.tmp files are generated by the gcc_ok
macro and do not contain Makefile rules at all.

Signed-off-by: H. Peter Anvin <hpa@zytor.com>
10 years agoversion: Update year to 2014
H. Peter Anvin [Fri, 17 Jan 2014 22:26:28 +0000 (14:26 -0800)]
version: Update year to 2014

Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
10 years agoversion: Update to 4.08, update year to 2014
H. Peter Anvin [Fri, 17 Jan 2014 22:25:34 +0000 (14:25 -0800)]
version: Update to 4.08, update year to 2014

Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
10 years agoMerge branch 'master' into elflink
H. Peter Anvin [Fri, 17 Jan 2014 22:21:44 +0000 (14:21 -0800)]
Merge branch 'master' into elflink

Resolved Conflicts:
NEWS
core/fs/fs.c
version

Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
10 years agodiag/geodsp: README fixes
Gene Cumm [Mon, 9 Sep 2013 03:22:22 +0000 (23:22 -0400)]
diag/geodsp: README fixes

Should clarify the situation; also word-wrap & save example

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
10 years agodiag/geodsp: fix Makefile
Gene Cumm [Sun, 8 Sep 2013 23:05:30 +0000 (19:05 -0400)]
diag/geodsp: fix Makefile

Results in null image

Reported-By: ioannis
Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
10 years agocore/legacynet: Enable dot quad resolution
Gene Cumm [Thu, 18 Jul 2013 15:20:49 +0000 (11:20 -0400)]
core/legacynet: Enable dot quad resolution

in dnsresolv.c:dns_resolv()

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
10 years agoRemove some whitespace
Ruben Kerkhof [Mon, 30 Dec 2013 21:53:35 +0000 (16:53 -0500)]
Remove some whitespace

Signed-off-by: Ruben Kerkhof <ruben at rubenkerkhof.com>
Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
10 years agoisolinux: Clear upper half of EDX before using
H. Peter Anvin [Tue, 26 Nov 2013 17:58:17 +0000 (09:58 -0800)]
isolinux: Clear upper half of EDX before using

In checkin:

    cb015497a4e4 isolinux: Update LBA in getlinsec loop

... we use EDX as a sector count, but the sector count is actually in
DX, and the upper half of EDX is uninitialized.  If the BIOS enters
with a nonzero value in the upper half of EDX, this breaks horribly.

At least one set of BIOSes has been identified where if the LBA > 64K
then the upper half of EDX will be nonzero.

Reported-by: Carl Duff <cdrw2400@gmail.com>
Reported-by: Philip Müller <philm@manjaro.org>
Tested-by: Gerardo Exequiel Pozzi <vmlinuz386@yahoo.com.ar>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
10 years agopxe, efi: Bad read of file size over TFTP on EFI64
Celelibi [Sun, 1 Dec 2013 21:06:24 +0000 (22:06 +0100)]
pxe, efi: Bad read of file size over TFTP on EFI64

A fancy pointers logic has been replaced with a plain old if / else
branches. It was assigning only half of a 64 bits integer which is then
assigned to a size_t. Thus leading to a bug on platform where size_t is
64 bits.

Resolves bug #26

Signed-off-by: Celelibi <celelibi@gmail.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
10 years agoisolinux: Clear upper half of EDX before using
H. Peter Anvin [Tue, 26 Nov 2013 17:58:17 +0000 (09:58 -0800)]
isolinux: Clear upper half of EDX before using

In checkin:

    cb015497a4e4 isolinux: Update LBA in getlinsec loop

... we use EDX as a sector count, but the sector count is actually in
DX, and the upper half of EDX is uninitialized.  If the BIOS enters
with a nonzero value in the upper half of EDX, this breaks horribly.

At least one set of BIOSes has been identified where if the LBA > 64K
then the upper half of EDX will be nonzero.

Reported-by: Carl Duff <cdrw2400@gmail.com>
Reported-by: Philip Müller <philm@manjaro.org>
Tested-by: Gerardo Exequiel Pozzi <vmlinuz386@yahoo.com.ar>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
11 years agocom32: Fix a bug on history of commands.
Raphael S.Carvalho [Tue, 17 Sep 2013 19:52:10 +0000 (16:52 -0300)]
com32: Fix a bug on history of commands.

Previously, even zero-length commands would be added to the history when
they shoudn't, e.g: just typing enter.

For example, if you type: FOO -> (ENTER) -> (ENTER),
then to get FOO from the history you would have to press the UP key
twice. It also saves a bit of memory.

Signed-off-by: Raphael S.Carvalho <raphael.scarv@gmail.com>
Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agocom32: Fix bugs on cmd_reverse_search (Triple fault dimension)
Raphael S.Carvalho [Tue, 17 Sep 2013 19:48:58 +0000 (16:48 -0300)]
com32: Fix bugs on cmd_reverse_search (Triple fault dimension)

cmd_reverse_search has a bug that the variable cursor is updated even if a command
wasn't found. If this happens, and the next key falls into the default case,
memmove's size parameter would be a negative number.

This bug can be reproduced by doing the following:
On cmd_reverse_search (ctrl-r), type multiple keys at the same time.
'Enjoy' the triple fault and a screen of random colors.

There is also a small bug that turns the task of using (ctrl-r) on the first command
impossible. Previously, this command was discarded.

Signed-off-by: Raphael S.Carvalho <raphael.scarv@gmail.com>
Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agocom32/disk: Code cleanup at disk_write_sectors and disk_read_sectors.
Raphael S.Carvalho [Tue, 17 Sep 2013 19:45:34 +0000 (16:45 -0300)]
com32/disk: Code cleanup at disk_write_sectors and disk_read_sectors.

Pulled common code out of these functions into new ones.
The functions chs_setup and ebios_setup were created for this purpose.

Signed-off-by: Raphael S.Carvalho <raphael.scarv@gmail.com>
Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoPXE ISR: Force polling on select hardware WORKAROUND
Gene Cumm [Mon, 23 Sep 2013 21:28:26 +0000 (17:28 -0400)]
PXE ISR: Force polling on select hardware WORKAROUND

By OUI == 00:23:ae and flags == 0xdc1b, detect select hardware.

On select platforms (Dell OptiPlex 760, Dell OptiPlex 960; perhaps
more), the interrupt appears to go "deaf" after a few seconds.  By
matching MAC OUI and flags value, force polling on these select
platforms.  I'm not sure if there's any better data available that
shallow in the core.  I believe PCI IDs can be fetched with functions
from other libraries and the UUID and DMI data (the most likely to be
useful is SYSPRODUCT) is available in ldlinux.c32.

Commit message expanded with Matt Fleming's assistance

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agocore/lwip: Fix NULL pointer check
Gene Cumm [Thu, 22 Aug 2013 02:21:27 +0000 (22:21 -0400)]
core/lwip: Fix NULL pointer check

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agoPXE: use ddprintf macro
Gene Cumm [Fri, 16 Aug 2013 16:55:53 +0000 (12:55 -0400)]
PXE: use ddprintf macro

Messages are needed before ldlinux.c32 is loaded

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agocom32: Define ddprintf() macro
Gene Cumm [Fri, 16 Aug 2013 16:54:39 +0000 (12:54 -0400)]
com32: Define ddprintf() macro

Duplicating Debug Printf;  Certain warning/error printf()
 statements are not visible if they occur before ldlinux.c32
 is loaded

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agoPXELINUX: specify PXE/lwIP
Gene Cumm [Fri, 16 Aug 2013 16:52:58 +0000 (12:52 -0400)]
PXELINUX: specify PXE/lwIP

Styled after SYSLINUX

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agoundiif: show thread of execution on UNDIIF_ID_DEBUG
Gene Cumm [Fri, 16 Aug 2013 16:38:28 +0000 (12:38 -0400)]
undiif: show thread of execution on UNDIIF_ID_DEBUG

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agocore: dprintf() the banner.
Gene Cumm [Fri, 16 Aug 2013 02:28:27 +0000 (22:28 -0400)]
core: dprintf() the banner.

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agoPXELINUX: Use sendto() instead of connect()/send()/disconnect()
Gene Cumm [Mon, 23 Sep 2013 21:13:53 +0000 (17:13 -0400)]
PXELINUX: Use sendto() instead of connect()/send()/disconnect()

This commit prevents a race-condition on systems that have functional
interrupts (observed with iPXE and select other Dell systems).  Without
this, the reply packet could be received by the core prior to the
disconnect() call, see that it doesn't have a matching PCB (protocol
control block, iirc) since the reply has a different far-end UDP port
than the original request, and lwIP will discard the packet before
PXELINUX can see it.

net_core_sendto() instead of
net_core_connect() net_core_send() net_core_disconnect()

Commit message expanded with Matt Fleming's assistance

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agocore: make mbox_post()/__sem_down_slow() check if valid
Gene Cumm [Sat, 3 Aug 2013 20:01:28 +0000 (16:01 -0400)]
core: make mbox_post()/__sem_down_slow() check if valid

If the mailbox/semaphore is invalid, disallow additions.

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agocore: mbox/semaphore NULL checks
Gene Cumm [Sat, 3 Aug 2013 19:59:26 +0000 (15:59 -0400)]
core: mbox/semaphore NULL checks

Also set mbox invalid and mbox pointer NULL when free()d

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agocore/lwip/undi: Improve UNDIIF_ID_DEBUG messages
Gene Cumm [Sat, 3 Aug 2013 19:55:54 +0000 (15:55 -0400)]
core/lwip/undi: Improve UNDIIF_ID_DEBUG messages

Remove extra space; add TCP flags

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agoNEWS: document changes in 4.07 syslinux-4.07
Matt Fleming [Thu, 25 Jul 2013 09:28:59 +0000 (10:28 +0100)]
NEWS: document changes in 4.07

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoAllow for UNDIIF_ID_DEBUG in 1 mk line
Gene Cumm [Sat, 20 Jul 2013 13:25:23 +0000 (09:25 -0400)]
Allow for UNDIIF_ID_DEBUG in 1 mk line

Add comments and 1 line to mk/devel.mk to debug headers of all
packets sent/recieved through undiif.c.

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agocore/lwip/undiif: dprint MAC/ARP/IP/ICMP/TCP/UDP headers; poll debug
Gene Cumm [Sat, 20 Jul 2013 13:23:06 +0000 (09:23 -0400)]
core/lwip/undiif: dprint MAC/ARP/IP/ICMP/TCP/UDP headers; poll debug

Encapsulate in #ifdef; use snprintf and 1 dprintf-like statement.
Print debug message that polling has been forced

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agolwip/undiif: split arphdr._hwlen_protolen
Gene Cumm [Sat, 13 Jul 2013 12:16:14 +0000 (08:16 -0400)]
lwip/undiif: split arphdr._hwlen_protolen

lwIP already split this in etharp.h

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agolwip: Allow LWIP_PLATFORM_* output to use dprintf
Gene Cumm [Sun, 2 Oct 2011 16:55:20 +0000 (12:55 -0400)]
lwip: Allow LWIP_PLATFORM_* output to use dprintf

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agolwip: undiif: Fix debug options to unique *_DEBUG macros
Gene Cumm [Fri, 12 Jul 2013 18:28:41 +0000 (14:28 -0400)]
lwip: undiif: Fix debug options to unique *_DEBUG macros

Don't reuse *_DEBUG macros intended for other source files.

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agoversion: bump version and date
Matt Fleming [Fri, 12 Jul 2013 06:33:21 +0000 (07:33 +0100)]
version: bump version and date

Welcome to the 4.07 release cycle

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoMerge branch 'dynamic-debug' into elflink
Matt Fleming [Fri, 5 Jul 2013 12:50:34 +0000 (13:50 +0100)]
Merge branch 'dynamic-debug' into elflink

11 years agomk: enable dynamic debug support
Matt Fleming [Fri, 5 Jul 2013 11:14:11 +0000 (12:14 +0100)]
mk: enable dynamic debug support

This will probably be disabled for releases, but it's definitely
worthwhile having enabled for all prereleases.

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agodebug.c32: Dynamically enable/disable debug code at runtime
Matt Fleming [Fri, 28 Jun 2013 19:53:30 +0000 (20:53 +0100)]
debug.c32: Dynamically enable/disable debug code at runtime

It's useful to be able to enable (and disable) debug code at runtime,
particularly for allowing users that are unable to build their own
Syslinux releases the chance to provide useful debugging output.

For example, say a user reports trouble with their PXE stack but doesn't
have a development environment setup to turn on the debug code
themselves. With this change you can simply request that they do,

    debug.c32 -e pxe_call unload_pxe open_file

to enable the debug in those functions. By only turning on code in
specific functions we reduce the chance of disrupting the buggy
behaviour and improve the signal to noise ratio for print statements.

To disable debug code use the -d flag,

    debug.c32 -d pxe_call

To use this new feature simply do,

    if (syslinux_debug_enabled) {
        debug1();
        debug2();
        ....
    }

from within the function you wish to add debug code. Note that this
feature is not limited to print statements - you can put any code within
the conditional, such as verifying a checksum or checking for memory
leaks.

The plan is to leave the dynamic debug code built in for all prereleases
and to turn it off for final releases.

People may still want to build with all dprintf() statements enabled,
and so a new symbol, DYNAMIC_DEBUG, was introduced rather than
repurposing the old DEBUG, DEBUG_STDIO and DEBUG_PORT symbols.

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agomk: include $(GCCWARN) when building ELF objects
Matt Fleming [Fri, 5 Jul 2013 10:33:26 +0000 (11:33 +0100)]
mk: include $(GCCWARN) when building ELF objects

It's currently impossible to turn on dprintf() statements in
com32/modules because $(GCCWARN) isn't used as part of $(CFLAGS).

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoMerge branch 'idle' into elflink
Matt Fleming [Wed, 3 Jul 2013 21:28:43 +0000 (22:28 +0100)]
Merge branch 'idle' into elflink

11 years agoidle: Consolidate idle implementations
Matt Fleming [Tue, 2 Jul 2013 13:51:26 +0000 (14:51 +0100)]
idle: Consolidate idle implementations

Duplicate code is bad. Move all the idle code to C and delete the old
assembly stuff.

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoidle: enable interrupts before idling
Matt Fleming [Wed, 3 Jul 2013 21:09:15 +0000 (22:09 +0100)]
idle: enable interrupts before idling

Users are *still* reporting executing __idle() with interrupts disabled,
which ultimately leads to a hang. Just enable them explicitly before
idling.

Cc: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoMerge branch 'doc-for-mfleming' of git://github.com/geneC/syslinux into elflink
Matt Fleming [Wed, 3 Jul 2013 11:40:12 +0000 (12:40 +0100)]
Merge branch 'doc-for-mfleming' of git://github.com/geneC/syslinux into elflink

Pull doc/ and txt/ spelling corrections from Gene Cumm,

* 'doc-for-mfleming' of git://github.com/geneC/syslinux:
  doc/ & txt/: Spelling fix

11 years agocore: we only need one reset_idle()
Matt Fleming [Tue, 2 Jul 2013 11:50:54 +0000 (12:50 +0100)]
core: we only need one reset_idle()

We don't need both an assembly version of reset_idle and an
implementation in C. Having these two functions has led to a bug where
we may idle with interrupts disabled because an 'sti' is only performed
in the assembly version. This lead to a hang waiting for user input.

Cc: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agodoc/ & txt/: Spelling fix
Gene Cumm [Sun, 30 Jun 2013 17:59:47 +0000 (13:59 -0400)]
doc/ & txt/: Spelling fix

Also add octal note.

Reported-By: Ady <ady-sf@hotmail.com>
Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agoload_linux: validate sanity of 'initrd_addr_max'
Matt Fleming [Mon, 1 Jul 2013 15:55:20 +0000 (16:55 +0100)]
load_linux: validate sanity of 'initrd_addr_max'

Some kernel headers have bogus 'initrd_addr_max' fields. This field
should never be zero. Set it to the old upper limit if unspecified.
Failure to set a non-zero value for the field results in a bogus
'memlimit' value, thereby unnecessarily reserving part of the memmap.

Cc: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoload_linux: Add dprintf statements to aid debugging
Matt Fleming [Mon, 1 Jul 2013 11:54:06 +0000 (12:54 +0100)]
load_linux: Add dprintf statements to aid debugging

It's useful to know why a kernel is failing to load. Sprinkle some
dprintf() statements with informative messages.

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoconio: Fix pollchar() for serial console
Matt Fleming [Mon, 1 Jul 2013 09:28:15 +0000 (10:28 +0100)]
conio: Fix pollchar() for serial console

The check of the LSR value was inverted. This resulted in pollchar()
always claiming that there was data to be read from the serial console.

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoldlinux: fix INCLUDE regression
Matt Fleming [Thu, 27 Jun 2013 11:54:15 +0000 (12:54 +0100)]
ldlinux: fix INCLUDE regression

We don't want to overwrite 'ConfigName' when parsing a config file via
the INCLUDE directive, which commit 5447ef821 ("ldlinux: Always update
ConfigName when opening a config file") failed to take into account. In
the INCLUDE case we're only parsing config fragments, and not a main
config file.

Rename parse_one_config() to parse_main_config() to more accurately
reflect when it should be invoked (i.e. not for INCLUDE).

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoPXELINUX: implement our own version of local_boot16
Matt Fleming [Wed, 26 Jun 2013 19:35:32 +0000 (20:35 +0100)]
PXELINUX: implement our own version of local_boot16

We need to do different things for PXE, such as reset the PXE
environment when booting from the local disk from PXELINUX.

This fixes a localboot regression.

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoMerge branch 'ipappend-fix-for-mfleming' of git://github.com/geneC/syslinux into...
Matt Fleming [Wed, 26 Jun 2013 19:56:09 +0000 (20:56 +0100)]
Merge branch 'ipappend-fix-for-mfleming' of git://github.com/geneC/syslinux into elflink

Pull patches to standardize and document IPAPPEND/SYSAPPEND from Gene Cumm,

* 'ipappend-fix-for-mfleming' of git://github.com/geneC/syslinux:
  doc/ & txt/: Document IPAPPEND/SYSAPPEND format
  core & menu: fix IPAPPEND/SYSAPPEND conversion

11 years agodoc/ & txt/: Document IPAPPEND/SYSAPPEND format
Gene Cumm [Wed, 26 Jun 2013 16:01:56 +0000 (12:01 -0400)]
doc/ & txt/: Document IPAPPEND/SYSAPPEND format

Previously the maximum value was 7 which didn't require
differentiation between decimal and hexadecimal.

11 years agocore & menu: fix IPAPPEND/SYSAPPEND conversion
Gene Cumm [Wed, 26 Jun 2013 15:55:24 +0000 (11:55 -0400)]
core & menu: fix IPAPPEND/SYSAPPEND conversion

Make both use the same functions.

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agocore: remove unused header file vkernel.inc
H. Peter Anvin [Mon, 10 Jun 2013 21:09:24 +0000 (14:09 -0700)]
core: remove unused header file vkernel.inc

vkernel.inc is related to the old assembly-based config file parser;
it is no longer relevant.

Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
11 years agoMerge branch 'menu-ipappend-1-for-mfleming' of git://github.com/geneC/syslinux into...
Matt Fleming [Mon, 24 Jun 2013 11:03:05 +0000 (12:03 +0100)]
Merge branch 'menu-ipappend-1-for-mfleming' of git://github.com/geneC/syslinux into elflink

Pull a fix for the SYSAPPEND directive from Gene Cumm,

* 'menu-ipappend-1-for-mfleming' of git://github.com/geneC/syslinux:
  menu.c32: Fix SYSAPPEND

11 years agoMerge branch 'doc-for-mfleming' of git://github.com/geneC/syslinux into elflink
Matt Fleming [Mon, 24 Jun 2013 09:28:58 +0000 (10:28 +0100)]
Merge branch 'doc-for-mfleming' of git://github.com/geneC/syslinux into elflink

Pull txt/ and doc/ updates from Gene Cumm,

* 'doc-for-mfleming' of git://github.com/geneC/syslinux:
  txt/syslinux.txt: rewrap long command
  txt/Makefile: add isolinux.txt, pxelinux.txt
  txt/: Add isolinux.txt, pxelinux.txt
  txt/syslinux-cli.txt: Version on Ctrl-N
  txt/Makefile: order-only prerequisite
  txt/: Add common file for derivatives
  txt/syslinux.cfg.txt: Updates
  txt/syslinux.txt: synopsis, extlinux.sys, wrap long command
  txt/syslinux-cli.txt: Path rules
  txt/syslinux.cfg.txt: Add SENDCOOKIES, example config
  txt/syslinux.cfg.txt: Add SYSAPPEND
  doc/syslinux.txt: grammar
  Fix SERIAL directive in docs

11 years agomenu.c32: Fix SYSAPPEND
Gene Cumm [Mon, 24 Jun 2013 03:30:51 +0000 (23:30 -0400)]
menu.c32: Fix SYSAPPEND

Using the SYSAPPEND directive would examine "D" instead of the
value after the directive.  Also allows a 1-step switch to strtol()

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agotxt/syslinux.txt: rewrap long command
Gene Cumm [Sun, 23 Jun 2013 16:19:08 +0000 (12:19 -0400)]
txt/syslinux.txt: rewrap long command

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agotxt/Makefile: add isolinux.txt, pxelinux.txt
Gene Cumm [Sun, 23 Jun 2013 12:41:32 +0000 (08:41 -0400)]
txt/Makefile: add isolinux.txt, pxelinux.txt

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agotxt/: Add isolinux.txt, pxelinux.txt
Gene Cumm [Sun, 23 Jun 2013 12:41:09 +0000 (08:41 -0400)]
txt/: Add isolinux.txt, pxelinux.txt

Reformatted from previous in doc/.  Reflow to try to be more
manpage-like and put more commonly used information on the top.

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agotxt/syslinux-cli.txt: Version on Ctrl-N
Gene Cumm [Sun, 23 Jun 2013 12:39:25 +0000 (08:39 -0400)]
txt/syslinux-cli.txt: Version on Ctrl-N

Only on some versions.

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agocli: don't disable linewrap when redrawing the cmdline
Matt Fleming [Fri, 14 Jun 2013 07:48:11 +0000 (08:48 +0100)]
cli: don't disable linewrap when redrawing the cmdline

It messes things up when we want to use printf(), for example when
printing labels after the user has hit the TAB key.

This fixes a bug where labels would be printed on a single line on the
screen (with some being displayed off screen) when they should have been
printed across multiple lines.

I can't think of a reason to disable linerap when redrawing the command
line - so let's not.

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoldlinux: fix loading of kbdmap files
Matt Fleming [Thu, 13 Jun 2013 13:22:26 +0000 (14:22 +0100)]
ldlinux: fix loading of kbdmap files

Use the canonical code sequence for opening the kbdmap file instead of
referencing KernelName, which presumably is a remanent of the old
assembly code.

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agotxt/Makefile: order-only prerequisite
Gene Cumm [Thu, 13 Jun 2013 11:19:47 +0000 (07:19 -0400)]
txt/Makefile: order-only prerequisite

No need to rebuild when the directory's timestamp is updated.

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agoPATH: use a linked list internally
Matt Fleming [Thu, 13 Jun 2013 10:46:40 +0000 (11:46 +0100)]
PATH: use a linked list internally

In retrospect, choosing the colon character as the entry separator for
the PATH directive was not a smart move, as that character is also used
in TFTP-style paths. This conflict manifests as PXELINUX being unable to
find and load files.

An example dnsmasq log looks like,

dnsmasq-tftp: sent /arch/boot/syslinux/lpxelinux.0 to 192.168.0.90
dnsmasq-tftp: file /arch/ldlinux.c32 not found
dnsmasq-tftp: file /arch//ldlinux.c32 not found
dnsmasq-tftp: file /arch//boot/isolinux/ldlinux.c32 not found
dnsmasq-tftp: file /arch//isolinux/ldlinux.c32 not found
dnsmasq-tftp: file /arch//boot/syslinuxldlinux.c32 not found
dnsmasq-tftp: sent /arch//boot/syslinux/ldlinux.c32 to 192.168.0.90
dnsmasq-tftp: error 0 No error, file close received from 192.168.0.90
dnsmasq-tftp: failed sending /arch//boot/syslinux/ldlinux.c32 to 192.168.0.90
dnsmasq-tftp: sent /arch/boot/syslinux/archiso.cfg to 192.168.0.90
dnsmasq-tftp: sent /arch/boot/syslinux/whichsys.c32 to 192.168.0.90
dnsmasq-tftp: file /arch/libcom32.c32 not found
dnsmasq-tftp: file /arch//libcom32.c32 not found
dnsmasq-tftp: file /arch/libcom32.c32 not found
dnsmasq-tftp: file /arch//arch//boot/syslinux/libcom32.c32 not found

The last line of the log is the indication that there's a problem.
Internally, Syslinux adds the location of ldlinux.c32 to PATH by
querying the current working directory once ldlinux.c32 is successfully
loaded. Under PXELINUX that means the initial PATH string will be,

   "::/arch/boot/syslinux/"

The PATH parsing code doesn't know how to correctly parse the "::"
string and hence, the file is searched for relative to the 210 dhcp
option directory - /arch/.

Implement PATH with a linked list which *greatly* simplifies the path
code, and means we no longer have to parse strings backwards and
forwards.

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agotxt/: Add common file for derivatives
Gene Cumm [Thu, 13 Jun 2013 03:59:09 +0000 (23:59 -0400)]
txt/: Add common file for derivatives

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agotxt/syslinux.cfg.txt: Updates
Gene Cumm [Thu, 13 Jun 2013 03:56:49 +0000 (23:56 -0400)]
txt/syslinux.cfg.txt: Updates

DEFAULT often references a LABEL; TIMEOUT action affected by UI.
IPAPPEND/SYSAPPEND: wrap SYSAPPEND; notes on mask values 2,1.
TIMEOUT-related clarifications
SERIAL is sticky
PATH updated for 5.10-next with notes

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agotxt/syslinux.txt: synopsis, extlinux.sys, wrap long command
Gene Cumm [Thu, 13 Jun 2013 02:56:36 +0000 (22:56 -0400)]
txt/syslinux.txt: synopsis, extlinux.sys, wrap long command

Add -v/--version to synopsis; add note on extlinux.sys from old
versions; Wrap altmbr command for cleaner manpage

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agotxt/syslinux-cli.txt: Path rules
Gene Cumm [Thu, 13 Jun 2013 02:53:30 +0000 (22:53 -0400)]
txt/syslinux-cli.txt: Path rules

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agoversion: Welcome to the 5.11 release cycle
Matt Fleming [Wed, 12 Jun 2013 07:25:14 +0000 (08:25 +0100)]
version: Welcome to the 5.11 release cycle

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoldlinux: fix serial port handling regression
Matt Fleming [Wed, 12 Jun 2013 06:52:29 +0000 (07:52 +0100)]
ldlinux: fix serial port handling regression

commit c4fa33189f1d ("serial: Clean up and abstract handling of serial
ports") removed an assignment to 'port' in the config parsing code which
is used to setup the serial ports. Restore the assignment.

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agotxt/syslinux.cfg.txt: Add SENDCOOKIES, example config
Gene Cumm [Mon, 10 Jun 2013 19:55:47 +0000 (15:55 -0400)]
txt/syslinux.cfg.txt: Add SENDCOOKIES, example config

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agotxt/syslinux.cfg.txt: Add SYSAPPEND
Gene Cumm [Mon, 10 Jun 2013 19:43:35 +0000 (15:43 -0400)]
txt/syslinux.cfg.txt: Add SYSAPPEND

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agodoc/syslinux.txt: grammar
Gene Cumm [Mon, 10 Jun 2013 19:32:58 +0000 (15:32 -0400)]
doc/syslinux.txt: grammar

Reported-by: Ady <ady-sf@hotmail.com>
Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agoFix SERIAL directive in docs
Gene Cumm [Mon, 10 Jun 2013 14:59:38 +0000 (10:59 -0400)]
Fix SERIAL directive in docs

Signed-off-by: Gene Cumm <gene.cumm@gmail.com>
11 years agoNEWS: Mention noteworthy changes in 5.10
Matt Fleming [Tue, 4 Jun 2013 12:56:05 +0000 (13:56 +0100)]
NEWS: Mention noteworthy changes in 5.10

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoNEWS: Clarify network stack files
Matt Fleming [Tue, 4 Jun 2013 11:11:26 +0000 (12:11 +0100)]
NEWS: Clarify network stack files

Point users in the right direction when looking for the new and legacy
network stacks.

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agofont: write to 'fontbuf', not random memory
Matt Fleming [Tue, 4 Jun 2013 10:50:17 +0000 (11:50 +0100)]
font: write to 'fontbuf', not random memory

commit be5a345d385d ("font: load data as a single block") cleaned up the
loop for reading into the font buffer but made the mistake of dropping
the assignment to 'p', which means that the _fread() call is writing
through a garbage pointer. Since 'p' was only used to keep track while
writing into the font buffer in a loop, we can now delete it and
reference 'fontbuf' directly.

Also delete the unused variable 'i'.

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoserial: Clean up and abstract handling of serial ports
H. Peter Anvin [Fri, 31 May 2013 15:05:57 +0000 (08:05 -0700)]
serial: Clean up and abstract handling of serial ports

The special handling of serial ports 0-3 meaning "look in a BIOS
table" is at least officially BIOS-specific, so create an inline
function and move it to bios.h.

While we are at it, make the function look slightly less like
converted assembly.

Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
11 years agofont: load data as a single block
H. Peter Anvin [Fri, 31 May 2013 14:52:55 +0000 (07:52 -0700)]
font: load data as a single block

The only reason for a loop is to interleave data, which we don't need
anymore; instead the data is stored in packed form.

Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
11 years agoserial: fix SERIAL directive for ports > 0
Matt Fleming [Wed, 29 May 2013 14:01:12 +0000 (15:01 +0100)]
serial: fix SERIAL directive for ports > 0

There is a bug in the SERIAL directive parsing code that means that
using a port other than the first (SERIAL 0) results in a non-working
serial console. We need to use the serial port number from the config
file to offset the BIOS I/O port base address, otherwise we always
lookup the I/O port address for the first serial port instead of the one
the user specified.

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agofont: increment pointer to fontbuf by font height
Matt Fleming [Wed, 29 May 2013 12:56:02 +0000 (13:56 +0100)]
font: increment pointer to fontbuf by font height

Instead of always assuming that the font height is 32 bytes, use
hdr.height to increment our fontbuf pointer. hdr.height is the size of
the chunks we read from the .psf file.

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agolinux: Clear up error message
Matt Fleming [Thu, 25 Apr 2013 14:14:18 +0000 (15:14 +0100)]
linux: Clear up error message

Passing opt.device to perror() in do_open_file() results in confusing
error messages, as Jack reports,

  Last night, I was trying to install syslinux (5.01) to a SD card of
  mine, writing to a subdirectory to keep the layout neat:

  syslinux-5.01/linux# ./syslinux -i -d /boot /dev/mmcblk0p1
  /dev/mmcblk0p1: No such file or directory

  The message is coming from do_open_file() in syslinux.c: the call to
  open() fails because the containing directory isn't present, and then it
  calls perror(opt.device).

We should instead be passing 'name' to perror() so that we know _which_
file doesn't exist.

Reported-by: Jack Kelly <jack@jackkelly.name>
Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoload_linux: update errno before returning
Matt Fleming [Thu, 25 Apr 2013 12:17:06 +0000 (13:17 +0100)]
load_linux: update errno before returning

syslinux_boot_linux() only returns control to the caller when something
has gone wrong, under normal circumstances a kernel is booted.

Update errno accordingly in the error paths, so the user no longer sees
inaccurate errno values which result in the following kinds of messages,

    "Booting kernel failed: Bad file descriptor"

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoFix support for Linux kernel images with no protected mode code
Josh Triplett [Wed, 27 Mar 2013 21:50:37 +0000 (14:50 -0700)]
Fix support for Linux kernel images with no protected mode code

Some kernel images use the Linux kernel boot protocol and header
structure, but do not actually have any protected-mode code.  For
instance, grub's 1024-byte lnxboot.img consists of 1024 real-mode bytes
and 0 protected-mode bytes; you can concatenate it with a full grub
core.img to produce a self-contained bootable kernel, but you can also
use it standalone as the kernel with the core.img loaded as an initrd.

syslinux 4 supports this, but it no longer works with syslinux 5: the
memmap functions do not correctly handle a request to work with a 0-byte
region.  With lnxboot.img, this would cause syslinux to bail because it
thinks it has no space at the 1M load location but cannot relocate
lnxboot.img.  (In bailing, it gives the confusing error message "Bad
file descriptor", not because that error actually occurred when
attempting to boot the kernel, but because errno has that value on entry
to syslinux_boot_linux and nothing clears or sets it.)

Fix the regression by handling the corner case of no protected-mode code
explicitly.

Signed-off-by: Josh Triplett <josh@joshtriplett.org>
Signed-off-by: Burt Triplett <burt@pbjtriplett.org>
Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agopxe: disentangle the legacy and lwip stacks without #ifdef
H. Peter Anvin [Thu, 28 Mar 2013 21:14:20 +0000 (14:14 -0700)]
pxe: disentangle the legacy and lwip stacks without #ifdef

Disentangle the dependencies via the legacy and the lwip stacks.  Now
there are no more #ifdefs (except for a single one in pxelinux.asm)
and the only difference between the pxelinux and lpxelinux archives
are which files are included.

To make this happen:

1. Replace the private part of the socket structure with a union.
2. Move the list of URL methods into a stack-dependent file.

Reported-by: Josh Triplett <josh@joshtriplett.org>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
11 years agopxe: Fix compiler warnings and real bug syslinux-5.10-pre2
Matt Fleming [Wed, 20 Mar 2013 15:22:40 +0000 (15:22 +0000)]
pxe: Fix compiler warnings and real bug

commit 00a6f13139 ("Partially revert "pxe: Pass absolute path to
pxe_chdir()"") introduced a bug by comparing the function 'url_type'
with URL_SUFFIX. This resulted in the following warning,

fs/pxe/pxe.c: In function ‘pxe_chdir’:
fs/pxe/pxe.c:460:18: warning: comparison between pointer and integer [enabled by default]
fs/pxe/pxe.c:458:19: warning: unused variable ‘path_type’ [-Wunused-variable]
fs/pxe/pxe.c: In function ‘pxe_chdir’:
fs/pxe/pxe.c:460:18: warning: comparison between pointer and integer [enabled by default]
fs/pxe/pxe.c:458:19: warning: unused variable ‘path_type’ [-Wunused-variable]

which unfortunately went unnoticed.

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoldlinux: Fixup lwip merge botch
Matt Fleming [Fri, 15 Mar 2013 16:32:52 +0000 (16:32 +0000)]
ldlinux: Fixup lwip merge botch

The lwip merge changed the code for copying cmdline arguments from,

    strcpy(q, p);

to this,

    do {
*q++ = ch = *p++;
    } while (ch);

which means 'p' no longer points at 'args'. Use 'args' explicitly for
the case where we need to apply a filename extension, otherwise users
end up seeing errors like,

    Failed to load COM32 file .c32

Reported-by: Mattias Schlenker <ms@mattiasschlenker.de>
Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoPartially revert "pxe: Pass absolute path to pxe_chdir()"
Matt Fleming [Mon, 11 Mar 2013 13:57:56 +0000 (13:57 +0000)]
Partially revert "pxe: Pass absolute path to pxe_chdir()"

This partially reverts commit b208ba467f678ed8e73f8d11fc0609634120cb83.

It isn't correct to always pass URL_OLD_TFTP to chdir(), since the
path prefix option may contain a full url, e.g. http://

Specialise the one case where we need to build a ::-style TFTP path
inside of get_prefix() to maintain backwards compatability, since from
Syslinux 5.00 onwards we may move out of the root directory ("::")
while searching for ldlinux.c32.

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoreadconfig: Don't display "(null)" labels when hitting TAB
Matt Fleming [Thu, 28 Feb 2013 15:07:19 +0000 (15:07 +0000)]
readconfig: Don't display "(null)" labels when hitting TAB

It is entirely legitimate for a menu entry to have a NULL label or
displayname. Since these entries are not used for execution, skip
displaying them in print_labels() when hitting the TAB key.

Reported-by: Bernd Blaauw <bblaauw@home.nl>
Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoldlinux: Always update ConfigName when opening a config file
Matt Fleming [Thu, 28 Feb 2013 17:44:54 +0000 (17:44 +0000)]
ldlinux: Always update ConfigName when opening a config file

With the introduction of commit aa7dd29db684 ("ldlinux: Pass config
filename as argv[1] to ldlinux.c32") we stopped using ConfigName to
store the filename of the config file we intend to open in execute()
since it interfered with the filesystem-specific open_config()
implementations. But a side-effect of this change is that
syslinux_config_file() no longer works if we explicitly pass a
filename to parse_config_file().

The logical place to keep ConfigName in sync is in parse_config_file()
since it's expected that all the filesystem-specific open_config()
functions will fill out ConfigName for us. In the case where the
filename is supplied by the caller, we need to fill it out ourselves.

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoMerge branch 'lwip-merge' into elflink syslinux-5.10-pre1
Matt Fleming [Wed, 6 Mar 2013 17:10:32 +0000 (17:10 +0000)]
Merge branch 'lwip-merge' into elflink

11 years agonet: Return of the legacy network stack
Matt Fleming [Tue, 26 Feb 2013 16:24:56 +0000 (16:24 +0000)]
net: Return of the legacy network stack

While it's nice having the shiny new lwIP stack, there is definitely
merit in being able to choose between two different network stacks. We
want to keep the legacy network stack around as it is known to handle
funky BIOS implementations and provides a good reference point when
bugs are suspected in the lwIP code.

Users now have a choice of .0 files. pxelinux.0 uses the legacy
network stack, while lpxelinux.0 uses lwIP.

Note that not every protocol is converted to using this new API. The
http, ftp and tcp code is still inherently tied to the netconn API,
and is only available with lpxelinux.0 and the lwIP stack. It's
unlikely that this code will ever be fixed up to work with the legacy
network stack.

Network stack operations are abstracted behind the net_core_*
interface, and each network stack has private data fields contained
within a struct net_private.

Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: Gene Cumm <gene.cumm@gmail.com>
Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agomenugen: Make it compatible with Py3k
Paulo Alcantara [Sun, 6 Jan 2013 20:36:17 +0000 (18:36 -0200)]
menugen: Make it compatible with Py3k

Signed-off-by: Paulo Alcantara <pcacjr@zytor.com>
Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoshuffle: syslinux_dump_movelist() only takes 1 argument
Matt Fleming [Tue, 26 Feb 2013 14:25:00 +0000 (14:25 +0000)]
shuffle: syslinux_dump_movelist() only takes 1 argument

Fix up the following compiler error that is triggered with -DDEBUG=2,

syslinux/shuffle.c:135:2: error: too many arguments to function ‘syslinux_dump_movelist’
In file included from syslinux/shuffle.c:44:0:
../include/syslinux/movebits.h:87:6: note: declared here
syslinux/shuffle.c:155:5: warning: passing argument 1 of ‘syslinux_dump_movelist’ from incompatible pointer type [enabled by default]

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agolwip: Fix use-after-free memory corruption
Matt Fleming [Tue, 19 Feb 2013 12:18:19 +0000 (12:18 +0000)]
lwip: Fix use-after-free memory corruption

Set *sem to NULL after free() otherwise calling sys_sem_set_invalid()
will cause us to write into a memory location that has potentially
either been reused for another allocation or contains freelist
metadata.

This manifested as malloc() corruption, because we
sys_sem_set_invalid() was overwriting malloc metadata used for
maintaining the freelist.

Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: Gene Cumm <gene.cumm@gmail.com>
Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agopxe: Handle closing file with NULL ->inode
Matt Fleming [Mon, 18 Feb 2013 20:25:27 +0000 (20:25 +0000)]
pxe: Handle closing file with NULL ->inode

It's possible that pxe_close_file() will be called with a NULL
file->inode from searchdir() if pxe_searchdir() fails to lookup a
file.

Signed-off-by: Matt Fleming <matt.fleming@intel.com>
11 years agoMerge branch 'lwip' into elflink
Matt Fleming [Mon, 25 Feb 2013 15:25:16 +0000 (15:25 +0000)]
Merge branch 'lwip' into elflink

Welcome to Syslinux 5.10.

Conflicts:
NEWS
com32/lib/Makefile
com32/lib/sys/open.c
com32/lib/syslinux/ipappend.c
com32/modules/Makefile
com32/modules/prdhcp.c
core/Makefile
core/cmdline.inc
core/com32.inc
core/comboot.inc
core/configinit.inc
core/fs/chdir.c
core/fs/fs.c
core/fs/pxe/dnsresolv.c
core/fs/pxe/pxe.c
core/fs/pxe/pxe.h
core/idle.c
core/include/ctype.h
core/init.inc
core/mem/init.c
core/parseconfig.inc
core/runkernel.inc
core/syslinux.ld
core/ui.inc
doc/comboot.txt
version

11 years agocom32: Add com32_cmdline() which returns argv[1..argc-1] syslinux-5.02-pre3
Matt Fleming [Fri, 22 Feb 2013 15:13:38 +0000 (15:13 +0000)]
com32: Add com32_cmdline() which returns argv[1..argc-1]

cmd.c32 needs an equivalent of __com32.cs_cmdline now that the COMBOOT
code is dead. Introducing com32_cmdline(), which returns a string
consisting of module arguments from argv[1] to argv[argc-1], separated
by spaces.

Reported-by: Gene Cumm <gene.cumm@gmail.com>
Signed-off-by: Matt Fleming <matt.fleming@intel.com>