From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 912B6C6FD18 for ; Tue, 18 Apr 2023 12:41:13 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S230424AbjDRMlM (ORCPT ); Tue, 18 Apr 2023 08:41:12 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:57682 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S231847AbjDRMlL (ORCPT ); Tue, 18 Apr 2023 08:41:11 -0400 Received: from dfw.source.kernel.org (dfw.source.kernel.org [IPv6:2604:1380:4641:c500::1]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id A7F1413C3E for ; Tue, 18 Apr 2023 05:41:07 -0700 (PDT) Received: from smtp.kernel.org (relay.kernel.org [52.25.139.140]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by dfw.source.kernel.org (Postfix) with ESMTPS id 1CB7C6331F for ; Tue, 18 Apr 2023 12:41:07 +0000 (UTC) Received: by smtp.kernel.org (Postfix) with ESMTPSA id 06FBCC4339B; Tue, 18 Apr 2023 12:41:06 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=linuxfoundation.org; s=korg; t=1681821666; bh=lsK/TUZla+jC9yYFotjKe2v94tQT4NTuYAWRzpS0IkY=; h=From:To:Cc:Subject:Date:In-Reply-To:References:From; b=JBBVeRcJJssezdxGC0nHg3RHEGh9bViJFCSPDSggeRCmTyb/X8Ca8EA5+sw5jmzZJ t3BXkn9Gzzpk5At6Nyzb8LwdNO6mvlwWHwJiVcRMFk1EDi0h8ZueTqsceff31ijY4R 9gRBOgFpd9NOrqCuB9ogin99a7KNo4NfOje5O2N0= From: Greg Kroah-Hartman To: stable@vger.kernel.org Cc: Greg Kroah-Hartman , patches@lists.linux.dev, Valentin Schneider , Arnd Bergmann , Baoquan He , "Eric W . Biederman" , Juri Lelli , "Luis Claudio R. Goncalves" , Miaohe Lin , Petr Mladek , Sebastian Andrzej Siewior , Thomas Gleixner , Andrew Morton , Wen Yang Subject: [PATCH 5.15 88/91] panic, kexec: make __crash_kexec() NMI safe Date: Tue, 18 Apr 2023 14:22:32 +0200 Message-Id: <20230418120308.613894049@linuxfoundation.org> X-Mailer: git-send-email 2.40.0 In-Reply-To: <20230418120305.520719816@linuxfoundation.org> References: <20230418120305.520719816@linuxfoundation.org> User-Agent: quilt/0.67 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Precedence: bulk List-ID: X-Mailing-List: stable@vger.kernel.org From: Valentin Schneider commit 05c6257433b7212f07a7e53479a8ab038fc1666a upstream. Attempting to get a crash dump out of a debug PREEMPT_RT kernel via an NMI panic() doesn't work. The cause of that lies in the PREEMPT_RT definition of mutex_trylock(): if (IS_ENABLED(CONFIG_DEBUG_RT_MUTEXES) && WARN_ON_ONCE(!in_task())) return 0; This prevents an nmi_panic() from executing the main body of __crash_kexec() which does the actual kexec into the kdump kernel. The warning and return are explained by: 6ce47fd961fa ("rtmutex: Warn if trylock is called from hard/softirq context") [...] The reasons for this are: 1) There is a potential deadlock in the slowpath 2) Another cpu which blocks on the rtmutex will boost the task which allegedly locked the rtmutex, but that cannot work because the hard/softirq context borrows the task context. Furthermore, grabbing the lock isn't NMI safe, so do away with kexec_mutex and replace it with an atomic variable. This is somewhat overzealous as *some* callsites could keep using a mutex (e.g. the sysfs-facing ones like crash_shrink_memory()), but this has the benefit of involving a single unified lock and preventing any future NMI-related surprises. Tested by triggering NMI panics via: $ echo 1 > /proc/sys/kernel/panic_on_unrecovered_nmi $ echo 1 > /proc/sys/kernel/unknown_nmi_panic $ echo 1 > /proc/sys/kernel/panic $ ipmitool power diag Link: https://lkml.kernel.org/r/20220630223258.4144112-3-vschneid@redhat.com Fixes: 6ce47fd961fa ("rtmutex: Warn if trylock is called from hard/softirq context") Signed-off-by: Valentin Schneider Cc: Arnd Bergmann Cc: Baoquan He Cc: "Eric W . Biederman" Cc: Juri Lelli Cc: Luis Claudio R. Goncalves Cc: Miaohe Lin Cc: Petr Mladek Cc: Sebastian Andrzej Siewior Cc: Thomas Gleixner Signed-off-by: Andrew Morton Signed-off-by: Wen Yang Signed-off-by: Greg Kroah-Hartman --- kernel/kexec.c | 11 ++++------- kernel/kexec_core.c | 20 ++++++++++---------- kernel/kexec_file.c | 4 ++-- kernel/kexec_internal.h | 15 ++++++++++++++- 4 files changed, 30 insertions(+), 20 deletions(-) --- a/kernel/kexec.c +++ b/kernel/kexec.c @@ -93,13 +93,10 @@ static int do_kexec_load(unsigned long e /* * Because we write directly to the reserved memory region when loading - * crash kernels we need a mutex here to prevent multiple crash kernels - * from attempting to load simultaneously, and to prevent a crash kernel - * from loading over the top of a in use crash kernel. - * - * KISS: always take the mutex. + * crash kernels we need a serialization here to prevent multiple crash + * kernels from attempting to load simultaneously. */ - if (!mutex_trylock(&kexec_mutex)) + if (!kexec_trylock()) return -EBUSY; if (flags & KEXEC_ON_CRASH) { @@ -165,7 +162,7 @@ out: kimage_free(image); out_unlock: - mutex_unlock(&kexec_mutex); + kexec_unlock(); return ret; } --- a/kernel/kexec_core.c +++ b/kernel/kexec_core.c @@ -46,7 +46,7 @@ #include #include "kexec_internal.h" -DEFINE_MUTEX(kexec_mutex); +atomic_t __kexec_lock = ATOMIC_INIT(0); /* Per cpu memory for storing cpu states in case of system crash. */ note_buf_t __percpu *crash_notes; @@ -944,7 +944,7 @@ int kexec_load_disabled; */ void __noclone __crash_kexec(struct pt_regs *regs) { - /* Take the kexec_mutex here to prevent sys_kexec_load + /* Take the kexec_lock here to prevent sys_kexec_load * running on one cpu from replacing the crash kernel * we are using after a panic on a different cpu. * @@ -952,7 +952,7 @@ void __noclone __crash_kexec(struct pt_r * of memory the xchg(&kexec_crash_image) would be * sufficient. But since I reuse the memory... */ - if (mutex_trylock(&kexec_mutex)) { + if (kexec_trylock()) { if (kexec_crash_image) { struct pt_regs fixed_regs; @@ -961,7 +961,7 @@ void __noclone __crash_kexec(struct pt_r machine_crash_shutdown(&fixed_regs); machine_kexec(kexec_crash_image); } - mutex_unlock(&kexec_mutex); + kexec_unlock(); } } STACK_FRAME_NON_STANDARD(__crash_kexec); @@ -993,13 +993,13 @@ ssize_t crash_get_memory_size(void) { ssize_t size = 0; - if (!mutex_trylock(&kexec_mutex)) + if (!kexec_trylock()) return -EBUSY; if (crashk_res.end != crashk_res.start) size = resource_size(&crashk_res); - mutex_unlock(&kexec_mutex); + kexec_unlock(); return size; } @@ -1019,7 +1019,7 @@ int crash_shrink_memory(unsigned long ne unsigned long old_size; struct resource *ram_res; - if (!mutex_trylock(&kexec_mutex)) + if (!kexec_trylock()) return -EBUSY; if (kexec_crash_image) { @@ -1058,7 +1058,7 @@ int crash_shrink_memory(unsigned long ne insert_resource(&iomem_resource, ram_res); unlock: - mutex_unlock(&kexec_mutex); + kexec_unlock(); return ret; } @@ -1130,7 +1130,7 @@ int kernel_kexec(void) { int error = 0; - if (!mutex_trylock(&kexec_mutex)) + if (!kexec_trylock()) return -EBUSY; if (!kexec_image) { error = -EINVAL; @@ -1206,7 +1206,7 @@ int kernel_kexec(void) #endif Unlock: - mutex_unlock(&kexec_mutex); + kexec_unlock(); return error; } --- a/kernel/kexec_file.c +++ b/kernel/kexec_file.c @@ -343,7 +343,7 @@ SYSCALL_DEFINE5(kexec_file_load, int, ke image = NULL; - if (!mutex_trylock(&kexec_mutex)) + if (!kexec_trylock()) return -EBUSY; dest_image = &kexec_image; @@ -415,7 +415,7 @@ out: if ((flags & KEXEC_FILE_ON_CRASH) && kexec_crash_image) arch_kexec_protect_crashkres(); - mutex_unlock(&kexec_mutex); + kexec_unlock(); kimage_free(image); return ret; } --- a/kernel/kexec_internal.h +++ b/kernel/kexec_internal.h @@ -13,7 +13,20 @@ void kimage_terminate(struct kimage *ima int kimage_is_destination_range(struct kimage *image, unsigned long start, unsigned long end); -extern struct mutex kexec_mutex; +/* + * Whatever is used to serialize accesses to the kexec_crash_image needs to be + * NMI safe, as __crash_kexec() can happen during nmi_panic(), so here we use a + * "simple" atomic variable that is acquired with a cmpxchg(). + */ +extern atomic_t __kexec_lock; +static inline bool kexec_trylock(void) +{ + return atomic_cmpxchg_acquire(&__kexec_lock, 0, 1) == 0; +} +static inline void kexec_unlock(void) +{ + atomic_set_release(&__kexec_lock, 0); +} #ifdef CONFIG_KEXEC_FILE #include