Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S933409AbaLKAvt (ORCPT ); Wed, 10 Dec 2014 19:51:49 -0500 Received: from mail-la0-f51.google.com ([209.85.215.51]:59189 "EHLO mail-la0-f51.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932621AbaLKAvs (ORCPT ); Wed, 10 Dec 2014 19:51:48 -0500 MIME-Version: 1.0 In-Reply-To: <5488EA01.6020600@redhat.com> References:

<5488EA01.6020600@redhat.com> From: Andy Lutomirski Date: Wed, 10 Dec 2014 16:51:26 -0800 Message-ID: Subject: Re: [PATCH, 3.18] sleeping function called from invalid context To: Rik van Riel Cc: Daniel J Blueman , Linux Kernel , Linus Torvalds , "H. Peter Anvin" Content-Type: text/plain; charset=UTF-8 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Wed, Dec 10, 2014 at 4:49 PM, Rik van Riel wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > On 12/10/2014 07:46 PM, Daniel J Blueman wrote: >> Gah. I had some non-temporal copy changes in the wrong tree. I'll >> check with a definitely clean tree and follow up if it still >> occurs. > > The exception handlers should definitely allow sleeping, > so I suspect those changes may be related. It would be really, really nice if we could arrange for kernel_fpu_begin to be unconditionally usable in anything except NMI context. The crypto code would be much less scary, we could make non-temporal copies safe, etc. Can we have ponies, too? --Andy > >> On 10 December 2014 at 16:37, Andy Lutomirski >> wrote: >>> Adding Rik, who is probably the most recent person to have looked >>> at the disaster fondly known as "FPU". >>> >>> On Wed, Dec 10, 2014 at 4:25 PM, Daniel J Blueman >>> wrote: >>>> With 3.18, I was seeing a significant number of allocation >>>> warnings from load_elf_binary call paths [1]. >>>> >>>> Allocating FPU state atomically [2] does prevent potentially >>>> sleeping while atomic, but probably isn't the correct fix. >>>> Anyone familiar with this area? >>>> >>>> -- [1] >>>> >>>> BUG: sleeping function called from invalid context at >>>> mm/slub.c:1240 in_atomic(): 1, irqs_disabled(): 0, pid: 29016, >>>> name: dmesg INFO: lockdep is turned off. Preemption disabled >>>> at:[] handle_mm_fault+0x5ff/0x8a0 >>>> >>>> CPU: 122 PID: 29016 Comm: dmesg Tainted: G W >>>> 3.18.0-test #6 Hardware name: Supermicro AS -1042G-LTF/H8QGL, >>>> BIOS DS3.5a 11/13/2014 ffffffff8203b81b ffff886fdf62f888 >>>> ffffffff81b942e8 0000000000000007 0000000000000000 >>>> ffff886fdf62f8b8 ffffffff8109ac4c 0000000000000000 >>>> 0000000000000010 ffff8807df84bc00 00000000000000d0 >>>> ffff886fdf62f908 Call Trace: [] >>>> dump_stack+0x4f/0x7c [] >>>> __might_sleep+0x164/0x250 [] >>>> kmem_cache_alloc+0xeb/0x138 [] >>>> init_fpu+0x71/0xb0 [] >>>> math_state_restore+0xce/0x208 [] >>>> do_device_not_available+0x2b/0x60 [] >>>> device_not_available+0x15/0x20 [] ? >>>> copy_page+0x12/0x33 >>> >>> wtf? I don't see any fpu state being accessed in copy_page. >>> >>> Can you disassemble "[] ? >>> copy_page+0x12/0x33"? >>> >>> --Andy >>> >>>> [] ? get_parent_ip+0x11/0x58 >>>> [] ? preempt_count_add+0x55/0xb0 >>>> [] ? do_cow_fault+0xe9/0x258 >>>> [] handle_mm_fault+0x5ff/0x8a0 >>>> [] ? __do_page_fault+0xc8/0x498 >>>> [] ? vma_gap_callbacks_rotate+0x19/0x20 >>>> [] __do_page_fault+0x12d/0x498 >>>> [] ? up_write+0x1e/0x48 [] >>>> ? vma_link+0x80/0xc0 [] ? >>>> vma_set_page_prot+0x3a/0x60 [] ? >>>> mmap_region+0x1be/0x5e0 [] >>>> do_page_fault+0x1e/0x70 [] >>>> page_fault+0x1f/0x30 [] ? >>>> __clear_user+0x2e/0x50 [] ? >>>> __clear_user+0x12/0x50 [] >>>> clear_user+0x2a/0x30 [] padzero+0x21/0x30 >>>> [] load_elf_binary+0x8cf/0xdd0 >>>> [] search_binary_handler+0x7f/0x1f8 >>>> [] do_execve_common.isra.34+0x616/0x790 >>>> [] ? do_execve_common.isra.34+0x572/0x790 >>>> [] do_execve+0x13/0x18 [] >>>> SyS_execve+0x20/0x30 [] >>>> stub_execve+0x69/0xa0 >>>> >>>> -- [2] >>>> >>>> diff --git a/arch/x86/include/asm/fpu-internal.h >>>> b/arch/x86/include/asm/fpu-internal.h index e97622f..57029ad >>>> 100644 --- a/arch/x86/include/asm/fpu-internal.h +++ >>>> b/arch/x86/include/asm/fpu-internal.h @@ -574,7 +574,7 @@ >>>> static inline int fpu_alloc(struct fpu *fpu) { if >>>> (fpu_allocated(fpu)) return 0; - fpu->state = >>>> kmem_cache_alloc(task_xstate_cachep, GFP_KERNEL); + >>>> fpu->state = kmem_cache_alloc(task_xstate_cachep, GFP_ATOMIC); >>>> if (!fpu->state) return -ENOMEM; WARN_ON((unsigned >>>> long)fpu->state & 15); -- Daniel J Blueman >>> >>> >>> >>> -- Andy Lutomirski AMA Capital Management, LLC >> >> >> > > > - -- > All rights reversed > -----BEGIN PGP SIGNATURE----- > Version: GnuPG v1 > > iQEcBAEBAgAGBQJUiOoBAAoJEM553pKExN6DfrQH/2qDtTACDGUfQ0VgeK6hkDK0 > mMp2Xk8xHhUTKkyLQtA1NSHAjfv0RsEvrnr11yJtbr8+vfMjJyxp0/Qq0w9gqqyR > zqVBPyE9yOauKvkyzBDtLOZW8u+hb+LUY8Cv4FkZieJg7cACPwW2jSeUI+ALStI/ > nb5qM6wUNdeZGnRxd82x23J/u17ILZ8HsdNb2Pk2FpmgRnZ0/NWZdw2Kp714uF47 > fvPrV41go7OEBoGFPtOIdF9Z7XagpfwN1Jg1ThfQqlLFtVv/Yf0P6hMX6siPjmMn > YuxYXsypfdWAWPCP7PloJW+nraHz0poX5/gN4YycZD4+Dh4pzNZ0oVSADQkageI= > =Hs9Y > -----END PGP SIGNATURE----- -- Andy Lutomirski AMA Capital Management, LLC -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/