| Message ID | 20100115204105.GB5185@outflux.net |
|---|---|
| State | Changes Requested |
| Headers | show |
Hi Kees, > diff --git a/fs/binfmt_elf.c b/fs/binfmt_elf.c > index b10acea..73594b9 100644 > --- a/fs/binfmt_elf.c > +++ b/fs/binfmt_elf.c > @@ -978,6 +978,16 @@ static int load_elf_binary(struct linux_binprm *bprm, > struct pt_regs *regs) #ifdef arch_randomize_brk > if ((current->flags & PF_RANDOMIZE) && (randomize_va_space > 1)) > current->mm->brk = current->mm->start_brk = > +# ifdef CONFIG_X86_32 > + /* in the case of NX emulation, shove the brk > + segment way out of the way of the exec > + randomization area, since it can collide with > + future allocations if not. */ > + ( (current->mm->get_unmapped_exec_area == > + arch_get_unmapped_exec_area) && > + (current->mm->brk < 0x08000000) > + ? (TASK_SIZE/6) : 0) + > +# endif > arch_randomize_brk(current->mm); Seeing as this is arch specific, it might be best to put it in arch_randomize_brk, if possible. Cheers, Jeremy
Hi Jeremy, On Sat, Jan 16, 2010 at 11:04:05AM +1100, Jeremy Kerr wrote: > > diff --git a/fs/binfmt_elf.c b/fs/binfmt_elf.c > > index b10acea..73594b9 100644 > > --- a/fs/binfmt_elf.c > > +++ b/fs/binfmt_elf.c > > @@ -978,6 +978,16 @@ static int load_elf_binary(struct linux_binprm *bprm, > > struct pt_regs *regs) #ifdef arch_randomize_brk > > if ((current->flags & PF_RANDOMIZE) && (randomize_va_space > 1)) > > current->mm->brk = current->mm->start_brk = > > +# ifdef CONFIG_X86_32 > > + /* in the case of NX emulation, shove the brk > > + segment way out of the way of the exec > > + randomization area, since it can collide with > > + future allocations if not. */ > > + ( (current->mm->get_unmapped_exec_area == > > + arch_get_unmapped_exec_area) && > > + (current->mm->brk < 0x08000000) > > + ? (TASK_SIZE/6) : 0) + > > +# endif > > arch_randomize_brk(current->mm); > > Seeing as this is arch specific, it might be best to put it in > arch_randomize_brk, if possible. arch_randomize_brk() is shared by x86 and x86_64, so if I moved it, it would still carry the #ifdef CONFIG_X86_32. I can certainly move it, but I felt it was more readable closer to the brk-calculation logic of the ELF loader. (Nothing in the cs-limit patch[1] currently modifies arch/x86/kernel/process.c, so I was also trying to avoid touching even more files than it already does.) Thanks, -Kees [1] http://kernel.ubuntu.com/git?p=ubuntu/ubuntu-lucid.git;a=commitdiff;h=32a306aa2738c3d0a3f1c451b1a179358f02abf2
diff --git a/fs/binfmt_elf.c b/fs/binfmt_elf.c index b10acea..73594b9 100644 --- a/fs/binfmt_elf.c +++ b/fs/binfmt_elf.c @@ -978,6 +978,16 @@ static int load_elf_binary(struct linux_binprm *bprm, struct pt_regs *regs) #ifdef arch_randomize_brk if ((current->flags & PF_RANDOMIZE) && (randomize_va_space > 1)) current->mm->brk = current->mm->start_brk = +# ifdef CONFIG_X86_32 + /* in the case of NX emulation, shove the brk + segment way out of the way of the exec + randomization area, since it can collide with + future allocations if not. */ + ( (current->mm->get_unmapped_exec_area == + arch_get_unmapped_exec_area) && + (current->mm->brk < 0x08000000) + ? (TASK_SIZE/6) : 0) + +# endif arch_randomize_brk(current->mm); #endif diff --git a/mm/mmap.c b/mm/mmap.c index 814b95f..17ed65d 100644 --- a/mm/mmap.c +++ b/mm/mmap.c @@ -1516,8 +1516,11 @@ arch_get_unmapped_exec_area(struct file *filp, unsigned long addr0, if (flags & MAP_FIXED) return addr; - if (!addr) - addr = randomize_range(SHLIB_BASE, 0x01000000, len); + if (!addr) { + addr = SHLIB_BASE; + if ((current->flags & PF_RANDOMIZE) && randomize_va_space) + addr = randomize_range(addr, 0x01000000, len); + } if (addr) { addr = PAGE_ALIGN(addr); @@ -1545,7 +1548,9 @@ arch_get_unmapped_exec_area(struct file *filp, unsigned long addr0, * Up until the brk area we randomize addresses * as much as possible: */ - if (addr >= 0x01000000) { + if ((current->flags & PF_RANDOMIZE) && + randomize_va_space && + addr >= 0x01000000) { tmp = randomize_range(0x01000000, PAGE_ALIGN(max(mm->start_brk, (unsigned long)0x08000000)), len);
This is a fix for the NX emulation patch to force the brk area well outside of the exec randomization area to avoid future allocation or brk growth collisions. Normally this isn't a problem, except when the text region has been loaded from a PIE binary and the CS limit can't be put just above bss. Additionally, the nx-emulation patch was still randomizing addresses even when randomize_va_space was disabled, which would cause collisions even faster if someone tried to disable randomization. BugLink: http://bugs.launchpad.net/bugs/452175 Signed-off-by: Kees Cook <kees.cook@canonical.com> --- fs/binfmt_elf.c | 10 ++++++++++ mm/mmap.c | 11 ++++++++--- 2 files changed, 18 insertions(+), 3 deletions(-)