Age | Commit message (Collapse) | Author | Files | Lines |
|
Signed-off-by: Matt Turner <mattst88@gmail.com>
|
|
Without the checks a malformed PCF file can cause the library to make
atom from random heap memory that was behind the `strings` buffer.
This may crash the process or leak information.
Signed-off-by: Julien Cristau <jcristau@debian.org>
(cherry picked from commit 672bb944311392e2415b39c0d63b1e1902905bcd)
|
|
If a pattern contains '?' character, any character in the string is skipped,
even if it is '\0'. The rest of the matching then reads invalid memory.
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Julien Cristau <jcristau@debian.org>
(cherry picked from commit d1e670a4a8704b8708e493ab6155589bcd570608)
|
|
Signed-off-by: Adam Jackson <ajax@redhat.com>
|
|
https://bugs.freedesktop.org/show_bug.cgi?id=83224
Found by clang's Address Sanitizer
crac.num_auths = set_font_authorizations(&authorizations, &authlen,
client);
/* Work around bug in xfs versions up through modular release 1.0.8
which rejects CreateAC packets with num_auths = 0 & authlen < 4 */
if (crac.num_auths == 0) {
authorizations = padding;
authlen = 4;
} else {
authlen = (authlen + 3) & ~0x3;
}
crac.length = (sizeof (fsCreateACReq) + authlen) >> 2;
crac.acid = cur->acid;
_fs_add_req_log(conn, FS_CreateAC);
_fs_write(conn, (char *) &crac, sizeof (fsCreateACReq));
_fs_write(conn, authorizations, authlen);
In the case in the report, set_font_authorizations setup authorizations as a
34 byte buffer (and authlen set to 34 as one would expect). The following
block changed authlen to 36 to make it 4byte aligned and the final _fs_write()
caused us to read 36 bytes from this 34 byte buffer.
This changes the incorrect size increase to instead use _fs_write_pad which
takes care of the padding for us.
Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
(cherry picked from commit 6972ea08ee5b2ef1cfbdc2fcaf14f06bbd391561)
|
|
src/fc/fserve.c:653:32: warning: format specifies type 'int' but the argument has type 'CARD32' (aka 'unsigned long') [-Wformat]
" from font server\n", rep->length);
^~~~~~~~~~~
1 warning generated.
Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
(cherry picked from commit e6009adbc89ec3e1f924bcb57b333c1c02f5e66d)
|
|
calloc for 0 bytes
Found by clang static analysis:
Call to 'calloc' has an allocation size of 0 bytes
Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
(cherry picked from commit ac559fad20bbae45332c758abb6a790c3fd341a2)
|
|
Found by clang static analysis:
Result of 'calloc' is converted to a pointer of type 'int', which is
incompatible with sizeof operand type 'int *'
This is likely benign because the old size was larger on any platform where
sizeof(int) <= sizeof(void *), which is everywhere.
Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
(cherry picked from commit d0fff111992fed9d9bfbf0c19e136bda9ba1db55)
|
|
Mostly signed vs unsigned comparisons
Signed-off-by: Keith Packard <keithp@keithp.com>
Squashed commit of three cherry-picks from master:
(cherry picked from commit eb67d10ae82b364a4324e96ce53baaa4e5e75f97)
(cherry picked from commit eefc0b0b908eb8533e704d7156ce983ad7891cc5)
(cherry picked from commit d967caa988eaabd9e84c82879e2f21bd33b952a7)
|
|
Lesser of two evil hacks, I suppose...
This reverts commit 0386fa77367a305deea3cc27f8a3865cc3c467c0.
|
|
stubs resolution
Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
|
|
The argument must be an unsigned char or -1; in these cases
we know it's not -1 so cast it to unsigned char.
Fixes
warning: array subscript is of type 'char' [-Wchar-subscripts]
Signed-off-by: Thomas Klausner <wiz@NetBSD.org>
|
|
The fix for CVE-2015-1804 prevent DWIDTH to be negative.
However, the spec states that "DWIDTH [...] is a vector indicating the
position of the next glyph’s origin relative to the origin of this glyph."
So negative values are correct.
Found by trying to compile XTS.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
|
|
Don't seem to have ability to link to BDF or Xserver internals docs yet
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
|
|
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
|
|
We use 32-bit ints to read from the bdf file, but then try to stick
into a 16-bit int in the xCharInfo struct, so make sure they won't
overflow that range.
Found by afl-1.24b.
v2: Verify that additions won't overflow 32-bit int range either.
v3: As Julien correctly observes, the previous check for bh & bw not
being < 0 reduces the number of cases we need to check for overflow.
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Julien Cristau <jcristau@debian.org>
|
|
Previously would charge on ahead with a NULL pointer in ci->bits, and
then crash later in FontCharInkMetrics() trying to access the bits.
Found with afl-1.23b.
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Julien Cristau <jcristau@debian.org>
|
|
Avoid integer overflow or underflow when allocating memory arrays
by multiplying the number of properties reported for a BDF font.
Reported-by: Ilja Van Sprundel <ivansprundel@ioactive.com>
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Julien Cristau <jcristau@debian.org>
|
|
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Signed-off-by: Thomas Klausner <wiz@NetBSD.org>
|
|
Parts were indented, others weren't, now is more consistent.
'git diff -w' shows no non-whitespace changes in this commit
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
|
|
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
|
|
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
|
|
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
|
|
Weak symbols on PE platforms do not work the same way as on ELF
platforms, hence we have been unable to have a fully functional shared
libXfont until now. This patch works around these issues so that we
can fix that.
In summary, only when compiling shared libraries on NO_WEAK_SYMBOLS
platforms, when the first stub is called, the invoking program is first
checked to determine if it exports the stubbed functions. Then, for
every stub call, if the function is exported by the loader, it is called
instead of the stub code.
serverClient and serverGeneration are data pointers, and therefore are
replaced by getter functions. ErrorF is variadic, so the override is
routed through VErrorF instead. FatalError has no va_list equivalent,
but it is not actually used in libXfont and therefore should be safe to
remove.
This requires all X servers to export their symbols, which requires
forthcoming patches for hw/xwin and xfs; the other xservers (including
tigervnc) already do this via LD_EXPORT_SYMBOLS_FLAG.
Signed-off-by: Yaakov Selkowitz <yselkowitz@users.sourceforge.net>
Reviewed-by: Colin Harrison <colin.harrison@virgin.net>
Acked-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Tested-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
|
|
When accessing a 16-bit font with firstRow > 0 with 8-bit text, check
to see if the font has a default character and return that for every
incoming character.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
|
|
If the only bitmaps we support are builtins, don't need the code to
register all the bitmap font file handlers.
Fixes gcc warnings:
bitmapfunc.c:110:1: warning: 'BitmapOpenBitmap' defined but not used [-Wunused-function]
BitmapOpenBitmap (FontPathElementPtr fpe, FontPtr *ppFont, int flags,
^
bitmapfunc.c:155:1: warning: 'BitmapGetInfoBitmap' defined but not used [-Wunused-function]
BitmapGetInfoBitmap (FontPathElementPtr fpe, FontInfoPtr pFontInfo,
^
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Rémi Cardona <remi@gentoo.org>
|
|
pcfread.c is a special case - it's needed for either reading pcf files
from disk (--enable-pcfformat) or from the builtin fonts in memory
(--enable-builtins), so needed a new AM_CONDITIONAL case.
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Rémi Cardona <remi@gentoo.org>
|
|
Require the #defines from configure.ac now that we're not sharing source
with the imake builds any longer.
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Rémi Cardona <remi@gentoo.org>
|
|
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Rémi Cardona <remi@gentoo.org>
|
|
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Rémi Cardona <remi@gentoo.org>
|
|
fs_read_list_info() parses a reply from the font server. The reply
contains a number of additional data items with embedded length or
count fields, none of which are validated. This can cause out of
bound reads when looping over these items in the reply.
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
|
|
fs_read_list() parses a reply from the font server. The reply
contains a list of strings with embedded length fields, none of
which are validated. This can cause out of bound reads when looping
over the strings in the reply.
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
|
|
fs_read_glyphs() parses a reply from the font server. The reply
contains embedded length fields, none of which are validated.
This can cause out of bound reads when looping over the glyph
bitmaps in the reply.
Reported-by: Ilja Van Sprundel <ivansprundel@ioactive.com>
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Matthieu Herrb <matthieu@herrb.eu>
|
|
Looping over the extents in the reply could go past the end of the
reply buffer if the reply indicated more extents than could fit in
the specified reply length.
Reported-by: Ilja Van Sprundel <ivansprundel@ioactive.com>
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Matthieu Herrb <matthieu@herrb.eu>
|
|
fs_alloc_glyphs() is a malloc wrapper used by the font code.
It contains a classic integer overflow in the malloc() call,
which can cause memory corruption.
Reported-by: Ilja Van Sprundel <ivansprundel@ioactive.com>
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Matthieu Herrb <matthieu@herrb.eu>
|
|
fs_read_extent_info() parses a reply from the font server.
The reply contains a 32bit number of elements field which is used
to calculate a buffer length. There is an integer overflow in this
calculation which can lead to memory corruption.
Reported-by: Ilja Van Sprundel <ivansprundel@ioactive.com>
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Matthieu Herrb <matthieu@herrb.eu>
|
|
fs_read_query_info() parses a reply from the font server. The reply
contains embedded length fields, none of which are validated. This
can cause out of bound reads in either fs_read_query_info() or in
_fs_convert_props() which it calls to parse the fsPropInfo in the reply.
Reported-by: Ilja Van Sprundel <ivansprundel@ioactive.com>
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Matthieu Herrb <matthieu@herrb.eu>
|
|
fs_get_reply() would take any reply size, multiply it by 4 and pass to
_fs_start_read. If that size was bigger than the current reply buffer
size, _fs_start_read would add it to the existing buffer size plus the
buffer size increment constant and realloc the buffer to that result.
This math could overflow, causing the code to allocate a smaller
buffer than the amount it was about to read into that buffer from
the network. It could also succeed, allowing the remote font server
to cause massive allocations in the X server, possibly using up all
the address space in a 32-bit X server, allowing the triggering of
other bugs in code that fails to handle malloc failure properly.
This patch protects against both problems, by disconnecting any
font server trying to feed us more than (the somewhat arbitrary)
64 mb in a single reply.
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Matthieu Herrb <matthieu@herrb.eu>
|
|
Functions to handle replies to font server requests were casting replies
from the generic form to reply specific structs without first checking
that the reply was at least as long as the struct being cast to.
Reported-by: Ilja Van Sprundel <ivansprundel@ioactive.com>
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Matthieu Herrb <matthieu@herrb.eu>
|
|
The connection setup reply from the font server can include a list
of alternate servers to contact if this font server stops working.
The reply specifies a total size of all the font server names, and
then provides a list of names. _fs_recv_conn_setup() allocated the
specified total size for copying the names to, but didn't check to
make sure it wasn't copying more data to that buffer than the size
it had allocated.
Reported-by: Ilja Van Sprundel <ivansprundel@ioactive.com>
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Matthieu Herrb <matthieu@herrb.eu>
|
|
lexAlias() reads from a file in a loop. It does this by starting with a
64 byte buffer. If that size limit is hit, it does a realloc of the
buffer size << 1, basically doubling the needed length every time the
length limit is hit.
Eventually, this will shift out to 0 (for a length of ~4gig), and that
length will be passed on to realloc(). A length of 0 (with a valid
pointer) causes realloc to free the buffer on most POSIX platforms,
but the caller will still have a pointer to it, leading to use after
free issues.
Reported-by: Ilja Van Sprundel <ivansprundel@ioactive.com>
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Matthieu Herrb <matthieu@herrb.eu>
|
|
FontFileReadDirectory() opens a fonts.dir file, and reads over every
line in an fscanf loop. For each successful entry read (font name,
file name) a call is made to FontFileAddFontFile().
FontFileAddFontFile() will add a font file entry (for the font name
and file) each time it’s called, by calling FontFileAddEntry().
FontFileAddEntry() will do the actual adding. If the table it has
to add to is full, it will do a realloc, adding 100 more entries
to the table size without checking to see if that will overflow the
int used to store the size.
Reported-by: Ilja Van Sprundel <ivansprundel@ioactive.com>
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Matthieu Herrb <matthieu@herrb.eu>
|
|
When _fs_load_glyphs calls fs_send_open_font with FontReopen set, it
passes a NULL name and namelen of 0, since fs_send_open_font is going
to reuse the previous name.
This overly restrictive check was added in XFree86 4.3.99.12:
http://cvsweb.xfree86.org/cvsweb/xc/lib/font/fc/fserve.c.diff?r1=3.23&r2=3.24
http://cvsweb.xfree86.org/cvsweb/xc/lib/font/fc/fserve.c?rev=3.24&content-type=text/vnd.viewcvs-markup
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
|
|
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
|
|
Instead of editing fsio.h to turn on debugging logs, just add
-DDEBUG to CPPFLAGS when building.
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
|
|
libXfont 1.5.0 will be synchronized with the fontsproto 2.1.3 API
changes needed for xorg-server 1.16 branch.
libXfont 1.4.x will be left for stable release branch for older
Xserver releases.
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
|
|
Building current libXfont git against fontsproto 2.1.2 causes clang
complaints of:
patcache.c:130:1: error: conflicting types for 'CacheFontPattern'
CacheFontPattern (FontPatternCachePtr cache,
^
patcache.c:176:1: error: conflicting types for 'FindCachedFontPattern'
FindCachedFontPattern (FontPatternCachePtr cache,
^
due to the constification of arguments not matching.
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Thomas Klausner <wiz@NetBSD.org>
|
|
Fixes clang analyzer warning:
bufio.c:165:13: warning: Access to field 'bufp' results in a dereference
of a null pointer (loaded from variable 'f')
f->bufp = f->buffer;
~ ^
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Thomas Klausner <wiz@NetBSD.org>
|
|
"FreeType" is only eight bytes long. The atom "FreeType\x00\x??" is
probably not what the author intended.
Signed-off-by: Peter Harris <pharris@opentext.com>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
|
|
Many const char issues.
One extra 'i' declared in ScaleFont; we can just use the same 'i' as
exists at the top level scope.
Also ignore bad-function-cast in ftfuncs.c and bitscale.c because
we're casting the return value from floor or ceil from double to
int. As floor and ceil are kinda designed to generate integer results,
it's pretty clear that we're doing what we want and that the compiler
is generating noise. I'm not sure why bad-function-cast is ever a good
warning to turn on, but I'll leave that for another day.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Gaetan Nadon <memsize@videotron.ca>
|