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 kanga.kvack.org (kanga.kvack.org [205.233.56.17]) by smtp.lore.kernel.org (Postfix) with ESMTP id C5FECC4332F for ; Tue, 8 Nov 2022 09:33:00 +0000 (UTC) Received: by kanga.kvack.org (Postfix) id 33B476B0071; Tue, 8 Nov 2022 04:33:00 -0500 (EST) Received: by kanga.kvack.org (Postfix, from userid 40) id 29D4A8E0001; Tue, 8 Nov 2022 04:33:00 -0500 (EST) X-Delivered-To: int-list-linux-mm@kvack.org Received: by kanga.kvack.org (Postfix, from userid 63042) id 165A16B0074; Tue, 8 Nov 2022 04:33:00 -0500 (EST) X-Delivered-To: linux-mm@kvack.org Received: from relay.hostedemail.com (smtprelay0014.hostedemail.com [216.40.44.14]) by kanga.kvack.org (Postfix) with ESMTP id 018586B0071 for ; Tue, 8 Nov 2022 04:32:59 -0500 (EST) Received: from smtpin23.hostedemail.com (a10.router.float.18 [10.200.18.1]) by unirelay08.hostedemail.com (Postfix) with ESMTP id C8B6C140955 for ; Tue, 8 Nov 2022 09:32:59 +0000 (UTC) X-FDA: 80109760878.23.583F84F Received: from szxga01-in.huawei.com (szxga01-in.huawei.com [45.249.212.187]) by imf12.hostedemail.com (Postfix) with ESMTP id 3FA0440008 for ; Tue, 8 Nov 2022 09:32:57 +0000 (UTC) Received: from canpemm500009.china.huawei.com (unknown [172.30.72.55]) by szxga01-in.huawei.com (SkyGuard) with ESMTP id 4N62vF2XX5zmVjP; Tue, 8 Nov 2022 17:32:41 +0800 (CST) Received: from [10.174.178.165] (10.174.178.165) by canpemm500009.china.huawei.com (7.192.105.203) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2375.31; Tue, 8 Nov 2022 17:32:52 +0800 Message-ID: Date: Tue, 8 Nov 2022 17:32:52 +0800 MIME-Version: 1.0 User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:102.0) Gecko/20100101 Thunderbird/102.4.1 Subject: Re: [PATCH v2] mm: fix unexpected changes to {failslab|fail_page_alloc}.attr To: Qi Zheng , , , , CC: , , , , References: <20221108035232.87180-1-zhengqi.arch@bytedance.com> <65863340-b32f-a2fe-67ae-f1079b19eee4@huawei.com> <70dfbac1-4c84-9567-30be-1e2594157e62@bytedance.com> From: Wei Yongjun In-Reply-To: <70dfbac1-4c84-9567-30be-1e2594157e62@bytedance.com> Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 8bit X-Originating-IP: [10.174.178.165] X-ClientProxiedBy: dggems704-chm.china.huawei.com (10.3.19.181) To canpemm500009.china.huawei.com (7.192.105.203) X-CFilter-Loop: Reflected ARC-Seal: i=1; s=arc-20220608; d=hostedemail.com; t=1667899979; a=rsa-sha256; cv=none; b=O2+/Wb6S4RljFPPJLHTqD3TbYYbD3NW/NUC37sFUSSj2TcZdDq2duHEYn3NNeshs37BRXZ mLF3gsGt2x4imQQJxu/oIYEfbLKybPzNfgjJ2E6okSxW5pGFuQumTBZpt6Z9UY4rR9p9S3 1SmDYykP1Q0G7WlWXN6QU+8T+8i5brU= ARC-Authentication-Results: i=1; imf12.hostedemail.com; dkim=none; spf=pass (imf12.hostedemail.com: domain of weiyongjun1@huawei.com designates 45.249.212.187 as permitted sender) smtp.mailfrom=weiyongjun1@huawei.com; dmarc=pass (policy=quarantine) header.from=huawei.com ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=hostedemail.com; s=arc-20220608; t=1667899979; h=from:from:sender:reply-to:subject:subject:date:date: message-id:message-id:to:to:cc:cc:mime-version:mime-version: content-type:content-type: content-transfer-encoding:content-transfer-encoding: in-reply-to:in-reply-to:references:references; bh=pp/z9V+8XK4IUpgWPTbLmcFFRzx66EC957DYWgASYzo=; b=hC9T4I6cLM7RU25/cifvmKhZmyg9ELCK/LI7/motAAjpiSbsGsxs9X4r6Y8j1f7/fzpXxA 7oQrPmO/6wswqikcJ7hLSbCv5YRlDmuQ5pwa57MnuhOCQLXlAra8EQwjU9x2g519pcO9Yq rrVSSF42OVrfufGensE6QUdhIOzdv+Q= X-Stat-Signature: 8uny8z7t3xxj645fctkf8oksbhrmu1td X-Rspamd-Queue-Id: 3FA0440008 Authentication-Results: imf12.hostedemail.com; dkim=none; spf=pass (imf12.hostedemail.com: domain of weiyongjun1@huawei.com designates 45.249.212.187 as permitted sender) smtp.mailfrom=weiyongjun1@huawei.com; dmarc=pass (policy=quarantine) header.from=huawei.com X-Rspam-User: X-Rspamd-Server: rspam11 X-HE-Tag: 1667899977-96175 X-Bogosity: Ham, tests=bogofilter, spamicity=0.000000, version=1.2.4 Sender: owner-linux-mm@kvack.org Precedence: bulk X-Loop: owner-majordomo@kvack.org List-ID: On 2022/11/8 16:58, Qi Zheng wrote: > > > On 2022/11/8 16:44, Wei Yongjun wrote: >> Hi Zheng Qi, >> >> On 2022/11/8 11:52, Qi Zheng wrote: >>> When we specify __GFP_NOWARN, we only expect that no warnings >>> will be issued for current caller. But in the __should_failslab() >>> and __should_fail_alloc_page(), the local GFP flags alter the >>> global {failslab|fail_page_alloc}.attr, which is persistent and >>> shared by all tasks. This is not what we expected, let's fix it. >>> >>> Cc: stable@vger.kernel.org >>> Fixes: 3f913fc5f974 ("mm: fix missing handler for __GFP_NOWARN") >>> Reported-by: Dmitry Vyukov >>> Signed-off-by: Qi Zheng >>> --- >>>   v1: https://lore.kernel.org/lkml/20221107033109.59709-1-zhengqi.arch@bytedance.com/ >>> >>>   Changelog in v1 -> v2: >>>    - add comment for __should_failslab() and __should_fail_alloc_page() >>>      (suggested by Jason) >>> >>>   include/linux/fault-inject.h |  7 +++++-- >>>   lib/fault-inject.c           | 14 +++++++++----- >>>   mm/failslab.c                | 12 ++++++++++-- >>>   mm/page_alloc.c              |  7 +++++-- >>>   4 files changed, 29 insertions(+), 11 deletions(-) >>> >>> diff --git a/include/linux/fault-inject.h b/include/linux/fault-inject.h >>> index 9f6e25467844..444236dadcf0 100644 >>> --- a/include/linux/fault-inject.h >>> +++ b/include/linux/fault-inject.h >>> @@ -20,7 +20,6 @@ struct fault_attr { >>>       atomic_t space; >>>       unsigned long verbose; >>>       bool task_filter; >>> -    bool no_warn; >>>       unsigned long stacktrace_depth; >>>       unsigned long require_start; >>>       unsigned long require_end; >>> @@ -32,6 +31,10 @@ struct fault_attr { >>>       struct dentry *dname; >>>   }; >>>   +enum fault_flags { >>> +    FAULT_NOWARN =    1 << 0, >>> +}; >>> + >>>   #define FAULT_ATTR_INITIALIZER {                    \ >>>           .interval = 1,                        \ >>>           .times = ATOMIC_INIT(1),                \ >>> @@ -40,11 +43,11 @@ struct fault_attr { >>>           .ratelimit_state = RATELIMIT_STATE_INIT_DISABLED,    \ >>>           .verbose = 2,                        \ >>>           .dname = NULL,                        \ >>> -        .no_warn = false,                    \ >> >> How about keep no_warn attr as it be, and export it to user? >> >> When testing with fault injection, and each fault will print an backtrace. >> but not all of the testsuit can tell us which one is fault injection >> message or other is a real warning/crash like syzkaller do. >> >> In my case, to make things simple, we usually used a regex to detect whether >> wanring/error happend. So we disabled the slab/page fault warning message by >> default, and only enable it when debug real issue. > > So you want to set/clear this no_warn attr through the procfs or sysfs > interface, so that you can easily disable/enable the slab/page fault > warning message from the user mode. Right? Yes, just like: echo 1 > /sys/kernel/debug/failslab/no_warn #disable message echo 0 > /sys/kernel/debug/failslab/no_warn #enable message Regards Wei Yongjun > > Seems reasonable to me. Anyone else has an opinion on this? If it is > really needed, I can do it later. > > Thanks, > Qi > >> >> Regards, >> >> >>>       } >>>     #define DECLARE_FAULT_ATTR(name) struct fault_attr name = FAULT_ATTR_INITIALIZER >>>   int setup_fault_attr(struct fault_attr *attr, char *str); >>> +bool should_fail_ex(struct fault_attr *attr, ssize_t size, int flags); >>>   bool should_fail(struct fault_attr *attr, ssize_t size); >>>     #ifdef CONFIG_FAULT_INJECTION_DEBUG_FS >>> diff --git a/lib/fault-inject.c b/lib/fault-inject.c >>> index 4b8fafce415c..5971f7c3e49e 100644 >>> --- a/lib/fault-inject.c >>> +++ b/lib/fault-inject.c >>> @@ -41,9 +41,6 @@ EXPORT_SYMBOL_GPL(setup_fault_attr); >>>     static void fail_dump(struct fault_attr *attr) >>>   { >>> -    if (attr->no_warn) >>> -        return; >>> - >>>       if (attr->verbose > 0 && __ratelimit(&attr->ratelimit_state)) { >>>           printk(KERN_NOTICE "FAULT_INJECTION: forcing a failure.\n" >>>                  "name %pd, interval %lu, probability %lu, " >>> @@ -103,7 +100,7 @@ static inline bool fail_stacktrace(struct fault_attr *attr) >>>    * http://www.nongnu.org/failmalloc/ >>>    */ >>>   -bool should_fail(struct fault_attr *attr, ssize_t size) >>> +bool should_fail_ex(struct fault_attr *attr, ssize_t size, int flags) >>>   { >>>       bool stack_checked = false; >>>   @@ -152,13 +149,20 @@ bool should_fail(struct fault_attr *attr, ssize_t size) >>>           return false; >>>     fail: >>> -    fail_dump(attr); >>> +    if (!(flags & FAULT_NOWARN)) >>> +        fail_dump(attr); >>>         if (atomic_read(&attr->times) != -1) >>>           atomic_dec_not_zero(&attr->times); >>>         return true; >>>   } >>> +EXPORT_SYMBOL_GPL(should_fail_ex); >>> + >>> +bool should_fail(struct fault_attr *attr, ssize_t size) >>> +{ >>> +    return should_fail_ex(attr, size, 0); >>> +} >>>   EXPORT_SYMBOL_GPL(should_fail); >>>     #ifdef CONFIG_FAULT_INJECTION_DEBUG_FS >>> diff --git a/mm/failslab.c b/mm/failslab.c >>> index 58df9789f1d2..ffc420c0e767 100644 >>> --- a/mm/failslab.c >>> +++ b/mm/failslab.c >>> @@ -16,6 +16,8 @@ static struct { >>>     bool __should_failslab(struct kmem_cache *s, gfp_t gfpflags) >>>   { >>> +    int flags = 0; >>> + >>>       /* No fault-injection for bootstrap cache */ >>>       if (unlikely(s == kmem_cache)) >>>           return false; >>> @@ -30,10 +32,16 @@ bool __should_failslab(struct kmem_cache *s, gfp_t gfpflags) >>>       if (failslab.cache_filter && !(s->flags & SLAB_FAILSLAB)) >>>           return false; >>>   +    /* >>> +     * In some cases, it expects to specify __GFP_NOWARN >>> +     * to avoid printing any information(not just a warning), >>> +     * thus avoiding deadlocks. See commit 6b9dbedbe349 for >>> +     * details. >>> +     */ >>>       if (gfpflags & __GFP_NOWARN) >>> -        failslab.attr.no_warn = true; >>> +        flags |= FAULT_NOWARN; >>>   -    return should_fail(&failslab.attr, s->object_size); >>> +    return should_fail_ex(&failslab.attr, s->object_size, flags); >>>   } >>>     static int __init setup_failslab(char *str) >>> diff --git a/mm/page_alloc.c b/mm/page_alloc.c >>> index 7192ded44ad0..cb6fe715d983 100644 >>> --- a/mm/page_alloc.c >>> +++ b/mm/page_alloc.c >>> @@ -3902,6 +3902,8 @@ __setup("fail_page_alloc=", setup_fail_page_alloc); >>>     static bool __should_fail_alloc_page(gfp_t gfp_mask, unsigned int order) >>>   { >>> +    int flags = 0; >>> + >>>       if (order < fail_page_alloc.min_order) >>>           return false; >>>       if (gfp_mask & __GFP_NOFAIL) >>> @@ -3912,10 +3914,11 @@ static bool __should_fail_alloc_page(gfp_t gfp_mask, unsigned int order) >>>               (gfp_mask & __GFP_DIRECT_RECLAIM)) >>>           return false; >>>   +    /* See comment in __should_failslab() */ >>>       if (gfp_mask & __GFP_NOWARN) >>> -        fail_page_alloc.attr.no_warn = true; >>> +        flags |= FAULT_NOWARN; >>>   -    return should_fail(&fail_page_alloc.attr, 1 << order); >>> +    return should_fail_ex(&fail_page_alloc.attr, 1 << order, flags); >>>   } >>>     #ifdef CONFIG_FAULT_INJECTION_DEBUG_FS >