The following commit has been merged in the linux branch: commit 1d2127123db9b1821959c2b8b7473dd7ffcdf527 Author: Imre Deak imre.deak@nokia.com Date: Mon Oct 5 13:40:44 2009 +0100

ARM: 5742/1: ARM: add debug check for invalid kernel page faults

According to the following in arch/arm/mm/fault.c page faults from kernel mode are invalid if mmap_sem is already held and there is no exception handler defined for the faulting instruction:

/* * As per x86, we may deadlock here. However, since the kernel only * validly references user space from well defined areas of the code, * we can bug out early if this is from code which shouldn't. */ if (!down_read_trylock(&mm->mmap_sem)) { if (!user_mode(regs) && !search_exception_tables(regs->ARM_pc)) goto no_context;

Since mmap_sem can be held at arbitrary times by another thread this also means that any page faults from kernel mode are invalid if no exception handler is defined for them, regardless whether mmap_sem is held at the time of fault.

To easier detect code that can trigger the above error, add a check also for the case where mmap_sem is acquired. As this has an overhead make it a VM debug check.

Signed-off-by: Imre Deak imre.deak@nokia.com Signed-off-by: Russell King rmk+kernel@arm.linux.org.uk

diff --git a/arch/arm/mm/fault.c b/arch/arm/mm/fault.c index ae0e25f..10e0680 100644 --- a/arch/arm/mm/fault.c +++ b/arch/arm/mm/fault.c @@ -292,6 +292,11 @@ do_page_fault(unsigned long addr, unsigned int fsr, struct pt_regs *regs) * down_read() */ might_sleep(); +#ifdef CONFIG_DEBUG_VM + if (!user_mode(regs) && + !search_exception_tables(regs->ARM_pc)) + goto no_context; +#endif }

fault = __do_page_fault(mm, addr, fsr, tsk);