From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755532AbZBMMzc (ORCPT ); Fri, 13 Feb 2009 07:55:32 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1759578AbZBMMzL (ORCPT ); Fri, 13 Feb 2009 07:55:11 -0500 Received: from ns.suse.de ([195.135.220.2]:35888 "EHLO mx1.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755189AbZBMMzJ (ORCPT ); Fri, 13 Feb 2009 07:55:09 -0500 Date: Fri, 13 Feb 2009 13:55:05 +0100 From: Nick Piggin To: Peter Zijlstra Cc: akpm , Mathieu Desnoyers , linux-kernel , Ingo Molnar Subject: Re: irq-disabled vs vmap vs text_poke Message-ID: <20090213125505.GE18462@wotan.suse.de> References: <1234529407.6519.28.camel@twins> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1234529407.6519.28.camel@twins> User-Agent: Mutt/1.5.9i Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Fri, Feb 13, 2009 at 01:50:07PM +0100, Peter Zijlstra wrote: > Hi, > > Ingo got the following splat: > > [ 5.101748] ------------[ cut here ]------------ > [ 5.104305] WARNING: at kernel/smp.c:329 smp_call_function_many+0x34/0x1ea() > [ 5.104305] Hardware name: P4DC6 > [ 5.104305] Modules linked in: > [ 5.104305] Pid: 1, comm: swapper Not tainted 2.6.29-rc4-tip-01766-g1757c19-dirty #2 > [ 5.104305] Call Trace: > [ 5.104305] [] warn_slowpath+0x79/0x8f > [ 5.104305] [] ? dump_trace+0x7d/0xac > [ 5.104305] [] ? __lock_acquire+0x319/0x382 > [ 5.104305] [] smp_call_function_many+0x34/0x1ea > [ 5.104305] [] ? do_flush_tlb_all+0x0/0x48 > [ 5.104305] [] ? do_flush_tlb_all+0x0/0x48 > [ 5.104305] [] smp_call_function+0x21/0x28 > [ 5.104305] [] on_each_cpu+0x14/0x23 > [ 5.104305] [] flush_tlb_all+0x19/0x1b > [ 5.104305] [] flush_tlb_kernel_range+0xd/0xf > [ 5.104305] [] vmap_debug_free_range+0x1c/0x20 > [ 5.104305] [] remove_vm_area+0x28/0x67 > [ 5.104305] [] __vunmap+0x30/0xab > [ 5.104305] [] vunmap+0x27/0x29 > [ 5.104305] [] text_poke+0xd6/0x104 > [ 5.104305] [] ? kprobe_target+0x0/0x15 > [ 5.104305] [] arch_disarm_kprobe+0x13/0x15 > [ 5.104305] [] __unregister_kprobe_top+0x68/0xe8 > [ 5.104305] [] unregister_kretprobes+0x2c/0xb9 > [ 5.104305] [] unregister_kretprobe+0x16/0x18 > [ 5.104305] [] init_test_probes+0x2ed/0x40c > [ 5.104305] [] init_kprobes+0x127/0x131 > [ 5.104305] [] ? lock_release_holdtime+0x43/0x48 > [ 5.104305] [] ? __slab_alloc+0x5d/0x27a > [ 5.104305] [] ? lockdep_init_map+0x80/0xe7 > [ 5.104305] [] ? clocksource_read+0xd/0xf > [ 5.104305] [] ? getnstimeofday+0x5e/0xe9 > [ 5.104305] [] ? timespec_to_ktime+0xe/0x11 > [ 5.104305] [] ? init_kprobes+0x0/0x131 > [ 5.104305] [] do_one_initcall+0x6a/0x169 > [ 5.104305] [] ? number+0x10d/0x1cf > [ 5.104305] [] ? register_lock_class+0x17/0x228 > [ 5.104305] [] ? __lock_acquire+0x319/0x382 > [ 5.104305] [] ? debug_check_no_obj_freed+0xda/0x126 > [ 5.104305] [] ? __lock_acquire+0x319/0x382 > [ 5.104305] [] ? lock_release_holdtime+0x43/0x48 > [ 5.104305] [] ? _spin_unlock+0x22/0x25 > [ 5.104305] [] ? proc_register+0x14b/0x15c > [ 5.104305] [] ? create_proc_entry+0x76/0x8c > [ 5.104305] [] ? default_affinity_write+0x3f/0x8a > [ 5.104305] [] ? init_irq_proc+0x58/0x65 > [ 5.104305] [] kernel_init+0x118/0x169 > [ 5.104305] [] ? kernel_init+0x0/0x169 > [ 5.104305] [] kernel_thread_helper+0x7/0x10 > [ 5.104305] ---[ end trace e93713a9d40cd06c ]--- > > Which points to vunmap() being called with interrupts disabled. > > Which made me look at the vmap/vunmap calls, and they appear to not be > irq-safe, therefore this would be a bug in text_poke(). > > [ that is, vmap() can end up calling get_vm_area_caller() which in turn > calls __get_vm_area_node() with GFP_KERNEL, ergo, don't do this from > an atomic context. ] > > Now text_poke() uses local_irq_save/restore(), which conveys that it can > be called with IRQs disabled, which is exactly what happens in the trace > above, however we just established that vmap/vunmap() are not irq-safe. > > Anybody got an idea on how to fix this? Oh, I thought the consensus was not to use vmap for this? With a bit of work, we can make vunmap irq-safe with the lazy vunmapping infrastructure (vmap could also be irq-safe, but would be subject to spurious failures due to being unable to flush lazy vunmaps. I think I got a mostly working patch cobbled together sitting here somewhere. I was waiting for some _really_ good use case before spending more time on it. I would prefer if at all possible to do vmap operations in sleepable, process context.