Re: [patch] x86: some lock annotations for user copy paths

[Ingo Molnar <mingo@elte.hu>]

* Ingo Molnar <mingo@elte.hu> wrote:

> > shouldn't some of that be conditional on pagefault_disable() 'n 
> > such?
> 
> i dont think so - those have their own special __atomic user-copy 
> primitives which Nick didnt touch.

hm, but the prefetch bits were touched, and -tip testing stumbled upon 
this false positive:

[   66.187448] PM: Adding info for No Bus:vcsa12
[   67.138685] evbug.c: Event. Dev: input0, Type: 20, Code: 0, Value: 500
[   80.149087] BUG: sleeping function called from invalid context at arch/x86/mm/fault.c:122
[   80.157344] in_atomic(): 1, irqs_disabled(): 0, pid: 6811, name: gam_server
[   80.164315] no locks held by gam_server/6811.
[   80.168685] Pid: 6811, comm: gam_server Not tainted 2.6.27-rc6-tip-00187-gbfd4ed7-dirty #2
[   80.177013] Call Trace:
[   80.179454]  [<ffffffff8026c1e8>] ? __debug_show_held_locks+0x22/0x24
[   80.185876]  [<ffffffff80242c4d>] __might_sleep+0x104/0x109
[   80.191430]  [<ffffffff8022fd25>] is_prefetch+0xe8/0x228
[   80.196760]  [<ffffffff8023050b>] do_page_fault+0x5d8/0x9a9
[   80.202361]  [<ffffffff802c8218>] ? sys_newstat+0x36/0x41
[   80.207774]  [<ffffffff80c3b2fa>] error_exit+0x0/0xb9

so i've excluded these commits for now. (once there's a fix i can 
continue testing it)

	Ingo