The indenter seems to have gotten confused by initializing arrays of
structs with the struct defined inline - for predefined structs it did
a better job, so match that.
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
LoaderSymbol calls dlsym with RTLD_DEFAULT pseudo handle making it search in
every loaded library. In addition glibc adds NODELETE flag to the library
containing the symbol.
It's used in doLoadModule to locate <modulename>ModuleData symbol, the
module's library gets the flag and is kept in memory even after it is
unloaded.
This patch adds LoaderSymbolFromModule function that looks for symbol only in
library specified by handle. That way the NODELETE flag isn't added.
This glibc behavior doesn't seem to be documented, but even if other
implementations differ, there is no reason to search ModuleData symbol outside
the module's library.
Signed-off-by: Michal Srb <msrb@suse.com>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
v2: Switch LoaderSymbolFromModule arguments order.
Correct description.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The function may be called from a fatal signal handler.
Signed-off-by: Chase Douglas <chase.douglas@canonical.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This is strictly the application of the script 'x-indent-all.sh'
from util/modular. Compared to the patch that Daniel posted in
January, I've added a few indent flags:
-bap
-psl
-T PrivatePtr
-T pmWait
-T _XFUNCPROTOBEGIN
-T _XFUNCPROTOEND
-T _X_EXPORT
The typedefs were needed to make the output of sdksyms.sh match the
previous output, otherwise, the code is formatted badly enough that
sdksyms.sh generates incorrect output.
The generated code was compared with the previous version and found to
be essentially identical -- "assert" line numbers and BUILD_TIME were
the only differences found.
The comparison was done with this script:
dir1=$1
dir2=$2
for dir in $dir1 $dir2; do
(cd $dir && find . -name '*.o' | while read file; do
dir=`dirname $file`
base=`basename $file .o`
dump=$dir/$base.dump
objdump -d $file > $dump
done)
done
find $dir1 -name '*.dump' | while read dump; do
otherdump=`echo $dump | sed "s;$dir1;$dir2;"`
diff -u $dump $otherdump
done
Signed-off-by: Keith Packard <keithp@keithp.com>
Acked-by: Daniel Stone <daniel@fooishbar.org>
Acked-by: Alan Coopersmith <alan.coopersmith@oracle.com>
This has never been buildable in any modular server release.
Reviewed-by: Julien Cristau <jcristau@debian.org>
Reviewed-by: Mark Kettenis <kettenis@openbsd.org>
Signed-off-by: Adam Jackson <ajax@redhat.com>
During unwind following an error when attempting to a load a module, we
attempt to call dlclose on a potentially NULL handle. This is a
side-effect of removing the abstraction layer in ab7f057.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Adam Jackson <ajax@redhat.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
libdl will refcount objects for us just fine, thanks.
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Julien Cristau <jcristau@debian.org>
Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Julien Cristau <jcristau@debian.org>
Signed-off-by: Adam Jackson <ajax@redhat.com>
This was always 0 from all the callers.
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Julien Cristau <jcristau@debian.org>
Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Julien Cristau <jcristau@debian.org>
Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Julien Cristau <jcristau@debian.org>
Signed-off-by: Adam Jackson <ajax@redhat.com>
This lets us drop some double-tracking of loaded modules too. If your
OS is too lame to have libdl, fix that first.
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Julien Cristau <jcristau@debian.org>
Signed-off-by: Adam Jackson <ajax@redhat.com>
Error: Write outside array bounds at Xext/geext.c:406
in function 'GEWindowSetMask' [Symbolic analysis]
In array dereference of cli->nextSib[extension] with index 'extension'
Array size is 128 elements (of 4 bytes each), index <= 128
Error: Buffer overflow at dix/events.c:592
in function 'SetMaskForEvent' [Symbolic analysis]
In array dereference of filters[deviceid] with index 'deviceid'
Array size is 20 elements (of 512 bytes each), index >= 0 and index <= 20
Error: Read buffer overflow at hw/xfree86/loader/loader.c:226
in function 'LoaderOpen' [Symbolic analysis]
In array dereference of refCount[new_handle] with index 'new_handle'
Array size is 256 elements (of 4 bytes each), index >= 1 and index <= 256
These bugs were found using the Parfait source code analysis tool.
For more information see http://research.sun.com/projects/parfait
Signed-off-by: Alan Coopersmith <alan.coopersmith@sun.com>
Signed-off-by: Adam Jackson <ajax@redhat.com>
Acked-by: Peter Hutterer <peter.hutterer@who-t.net>
All .a libraries were converted to .la, and instead of linking the
Xorg binary with a mix of .a and .la, and adding some libraries more
then once in the command line, etc, now it generates a single libxorg.la
from all the required convenience libraries, and links with a dummy
xorg.c (that should usually be the file with the main function...).
This removes the requirement of some things like libosandcommon and
libinit, that existed to circumvent problems when linking multiple
.a and .la in the final Xorg binary.
The "symbol table" is now generated dynamically, by a shell script,
with an embedded gawk parser that parses cpp output. The new file
sdksyms.sh is generated by hand by analyzing all Makefile.am's and
making it create a sdksyms.c file, that includes all sdk headers that
will add symbols for the Xorg binary. Module headers aren't read, and
a in 2 files it was required to add a "<hash>ifndef XorgLoader" around
declarations shared between the Xorg binary and libextmod. A few
other changes were added to other sdk headers, like preventing
multiple inclusion, or including other headers to satisfy dependencies.
This should be a lot more portable, and better (hopefully properly)
using libtool to generate convenience libraries.
If not taking the symbol addresses, linkage will break badly, as not
all symbols will be present, and it also requires changing library order,
and/or making some changes like the "libosandcommon".
This table should be modified to be generated automatically, as
it is required to "fool" the compiler/loader into adding all required
symbols to the X Server.
Save in a few special cases, _X_EXPORT should not be used in C source
files. Instead, it should be used in headers, and the proper C source
include that header. Some special cases are symbols that need to be
shared between modules, but not expected to be used by external drivers,
and symbols that are accessible via LoaderSymbol/dlopen.
This patch also adds conditionally some new sdk header files, depending
on extensions enabled. These files were added to match pattern for
other extensions/modules, that is, have the headers "deciding" symbol
visibility in the sdk. These headers are:
o Xext/panoramiXsrv.h, Xext/panoramiX.h
o fbpict.h (unconditionally)
o vidmodeproc.h
o mioverlay.h (unconditionally, used only by xaa)
o xfixes.h (unconditionally, symbols required by dri2)
LoaderSymbol and similar functions now don't have different prototypes,
in loaderProcs.h and xf86Module.h, so that both headers can be included,
without the need of defining IN_LOADER.
xf86NewInputDevice() device prototype readded to xf86Xinput.h, but
not exported (and with a comment about it).
A couple #if defined(Lynx) && defined(sun) had become just if defined(sun),
resulting in wrong settings for Solaris builds, so they're now just deleted.
Spiritual revert of 1fa4de80fc. Intel's C
compiler claims to be gcc-compatible; if they're not defining the same
macros as gcc then that's their bug, not ours. Even if we were to do
this aliasing we should do it once and for all in servermd.h.
lists, a really bad hash table, the last vestiges of the other backends,
and some miscellaneous cleanups. Good for dropping 300k from the size of
the built server on x86.