The symptom was an assembler warning
"GOT relocation burb: `___EXCEPTION_TABLE__' should be global"
followed (sometimes) by a core dump. The fix makes the compiler
generate the correct GOTOFF addressing for that symbol, rather than the
GOT addressing it was emitting before.
Warning: There is still at least one serious bug in the i386 exception
code for PIC. The exception code that is generated clobbers the GOT
register (%ebx) and then tries to use it later. That leads to core
dumps at program execution time. I know where the problem is, but I do
not have a fix for it at this time. Until it is fixed, exceptions will
not work in PIC code. This is a general problem for all i386 platforms;
it is not specific to FreeBSD.
libg++'s exception code causes gcc to generate (ahem!) non-conventional
assembler code in -fpic mode that gas and ld choke on. Basically, gas
and ld require than symbols referenced in the GOT (global offset table)
are actually global (as the name implies). It attempted to work around
it before, but didn't quite go far enough to prevent a core dump in ld.
This hack causes GOT referenced symbols to be forced global. This
probably breaks the __EXCEPTION_TABLE__ stuff in pic mode, but heck, it
wasn't even possible to compile with a shared library before at all.
I'm not 100% sure what the bug is. There's two possibilities:
1: gcc/cp/exception.c has to be fixed to stop doing GOT references to
local symbols, or
2: as/ld/symorder/ld.so etc need to be taught about how to keep local
symbols around so that they can be dealt with in GOT references.
John Polstra's elfkit stuff seems to deal with this fine though, which is
why I think it's a "missing feature" in our hacked gas and ld..
kernel" mechanism. This is just the foundation - more work follows
and will be committed over the next few hours.
Submitted-by: "Eric L. Hernes" <erich@lodgenet.com> & jkh
certain error conditions, it is possible for pages to be left allocated
in the object beyond it's end. It is generally bad practice to allocate
pages beyond the end of an object.
Turn OFF the "small servers" by default. FreeBSD systems should only
serve actively used programs. Jewels like chargen and echo are too
useful in attack scenarios.
possibility of security holes allowing root penetration.
Inspired by: Mark Handley <M.Handley@cs.ucl.ac.uk> and
Theo de Raadt <deraadt@theos.com> independently
Submitted by: Theo de Raadt <deraadt@theos.com>
since rt_readenv() already takes care of not setting unsafe variables.
This was part of the changes I submitted to Peter and John during the
review which must have gotten missed.
how I managed to get this out of sync, but I did. I guess that's what I
get for directly committing from different machines that I was testing on.
Pointed out by: Paul Traina <pst@freebsd.org>
known to printf(3) and then used printf() to format it... The only
problem what the #define printf out1fmt. The code was behaving differently
when run as a shell builtin since out1fmt() isn't printf(3).
Simple hack. Print to a buffer and fputs (also #defined for sh) the
result. This should fix the printf builtin problem in PR#1673, rather
than leaving the call commented out. (printf.o was being statically linked
in anyway, we might as well use it)
