Re: [kerneloops] regression in 2.6.27 wrt "lock_page" and the "hwclock" program

[Linus Torvalds <torvalds@linux-foundation.org>]

On Sat, 4 Oct 2008, Andrew Morton wrote:
> > Pid: 9591, comm: hwclock Tainted: G        W 2.6.27-0.372.rc8.fc10.i686 #1 
> >  [<c0427a53>] __might_sleep+0xd1/0xd6 
> >  [<c0479a8b>] lock_page+0x1a/0x34
> >  [<c0479cfa>] find_lock_page+0x23/0x48
> >  [<c047a215>] filemap_fault+0x9b/0x330
> >  [<c0486493>] __do_fault+0x40/0x2e6
> >  [<c0487d63>] handle_mm_fault+0x2ec/0x6d2
> >  [<c06e8260>] do_page_fault+0x2e5/0x693
> 
> Looks like `hwclock' disabled interrupts in userspace with sys_iopl()?

We probably should enable interrupts in the page fault code. We already do

	#ifdef CONFIG_X86_32
	        /* It's safe to allow irq's after cr2 has been saved and the vmalloc
	           fault has been handled. */
	        if (regs->flags & (X86_EFLAGS_IF | X86_VM_MASK))
	                local_irq_enable();

	        /*
	         * If we're in an interrupt, have no user context or are running in an
	         * atomic region then we must not take the fault.
	         */
	        if (in_atomic() || !mm)
	                goto bad_area_nosemaphore;
	#endif
	...

so we have code to do so, it's just that we don't do it if the page fault 
happened in an interrupt. But that's for the _kernel_ having interrupts 
disabled and us needing to fix up the vmalloc area lazily (do we ever even 
do that any more.. I dunno).

So we could easily add a check for 'user_space_vm(regs)' instead of 
checking the VM_MASK, and fix it that way. Hmm?

		Linus