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 B082DC7619A for ; Wed, 12 Apr 2023 02:58:51 +0000 (UTC) Received: by kanga.kvack.org (Postfix) id 301246B0074; Tue, 11 Apr 2023 22:58:51 -0400 (EDT) Received: by kanga.kvack.org (Postfix, from userid 40) id 2B0506B0075; Tue, 11 Apr 2023 22:58:51 -0400 (EDT) X-Delivered-To: int-list-linux-mm@kvack.org Received: by kanga.kvack.org (Postfix, from userid 63042) id 17ACF900002; Tue, 11 Apr 2023 22:58:51 -0400 (EDT) X-Delivered-To: linux-mm@kvack.org Received: from relay.hostedemail.com (smtprelay0010.hostedemail.com [216.40.44.10]) by kanga.kvack.org (Postfix) with ESMTP id 04A896B0074 for ; Tue, 11 Apr 2023 22:58:51 -0400 (EDT) Received: from smtpin30.hostedemail.com (a10.router.float.18 [10.200.18.1]) by unirelay01.hostedemail.com (Postfix) with ESMTP id C14BC1C5B25 for ; Wed, 12 Apr 2023 02:58:50 +0000 (UTC) X-FDA: 80671231620.30.6A494B2 Received: from out30-99.freemail.mail.aliyun.com (out30-99.freemail.mail.aliyun.com [115.124.30.99]) by imf18.hostedemail.com (Postfix) with ESMTP id 951CD1C000A for ; Wed, 12 Apr 2023 02:58:48 +0000 (UTC) Authentication-Results: imf18.hostedemail.com; dkim=none; spf=pass (imf18.hostedemail.com: domain of xueshuai@linux.alibaba.com designates 115.124.30.99 as permitted sender) smtp.mailfrom=xueshuai@linux.alibaba.com; dmarc=pass (policy=none) header.from=alibaba.com ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=hostedemail.com; s=arc-20220608; t=1681268329; 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=IVd9YFXd1X7keE6RdXTM/YzZdVNVtWZDFi0YdVgWhWU=; b=WuKLIZ7k4zGRygT10tohbEBz7omkwAbkUpud6nG5mg9ysIFkKOELiuwU4Qt4S5xW1Tr1B7 gH1e7e0g8twik26pxnE4TZm0lvKuoduu5Xang2GzMHULsan1L5PH9A/NTMAHtYUQ3SvePY 2UG/t1jgnq6GxOst56ikL3cEkerzqPU= ARC-Authentication-Results: i=1; imf18.hostedemail.com; dkim=none; spf=pass (imf18.hostedemail.com: domain of xueshuai@linux.alibaba.com designates 115.124.30.99 as permitted sender) smtp.mailfrom=xueshuai@linux.alibaba.com; dmarc=pass (policy=none) header.from=alibaba.com ARC-Seal: i=1; s=arc-20220608; d=hostedemail.com; t=1681268329; a=rsa-sha256; cv=none; b=RWN70oJCAQticSv5GzZfgCkHVC92kpQhC8Gl0+tUd/AXHU+4BwC9XTSeQZthxsn5W5euKq q6GJeCwRZAj/OoC/cKFSjqICQQ0ElVB88WiqipBMNRN+KsPgfYNnUKSdhRi0rkpGF1yz92 /Ru29KYhlLEsZJBiOhh5S782rUFXGBU= X-Alimail-AntiSpam:AC=PASS;BC=-1|-1;BR=01201311R101e4;CH=green;DM=||false|;DS=||;FP=0|-1|-1|-1|0|-1|-1|-1;HT=ay29a033018045192;MF=xueshuai@linux.alibaba.com;NM=1;PH=DS;RN=24;SR=0;TI=SMTPD_---0VfuNtIo_1681268322; Received: from 30.240.113.3(mailfrom:xueshuai@linux.alibaba.com fp:SMTPD_---0VfuNtIo_1681268322) by smtp.aliyun-inc.com; Wed, 12 Apr 2023 10:58:44 +0800 Message-ID: Date: Wed, 12 Apr 2023 10:58:41 +0800 MIME-Version: 1.0 User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:102.0) Gecko/20100101 Thunderbird/102.9.0 Subject: Re: [PATCH v5 2/2] ACPI: APEI: handle synchronous exceptions in task work Content-Language: en-US To: Kefeng Wang , tanxiaofei@huawei.com, mawupeng1@huawei.com, tony.luck@intel.com, naoya.horiguchi@nec.com Cc: linux-acpi@vger.kernel.org, linux-mm@kvack.org, linux-kernel@vger.kernel.org, justin.he@arm.com, akpm@linux-foundation.org, ardb@kernel.org, ashish.kalra@amd.com, baolin.wang@linux.alibaba.com, bp@alien8.de, cuibixuan@linux.alibaba.com, dave.hansen@linux.intel.com, james.morse@arm.com, jarkko@kernel.org, lenb@kernel.org, linmiaohe@huawei.com, lvying6@huawei.com, rafael@kernel.org, xiexiuqi@huawei.com, zhuo.song@linux.alibaba.com References: <20221027042445.60108-1-xueshuai@linux.alibaba.com> <20230411104842.37079-3-xueshuai@linux.alibaba.com> <84fd1fad-604e-0835-2813-199195aaab69@huawei.com> From: Shuai Xue In-Reply-To: <84fd1fad-604e-0835-2813-199195aaab69@huawei.com> Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit X-Rspam-User: X-Rspamd-Server: rspam03 X-Stat-Signature: ofsrfioikc8qugei3uh6djpoz8x51bqa X-Rspamd-Queue-Id: 951CD1C000A X-HE-Tag: 1681268328-223751 X-HE-Meta: U2FsdGVkX1+Ffo8oFeSp7TknHq4UxzUzNyLVrs+Z21jQPflhz3cou2l6LOj2CRyagOg7BxwZCt7su9M4OIwJzSbvQigJ9g4kjrnxZsRNUwOTlVZOM1CftdIYPCKIne5ExkCsULqXjjsTsfAcnitjrxoSxJckmoYHMz0ms014KPdSjhkWZ6NZ8QxMKE2aZKSBYd2Xit4Fakmp2DXS+HNSAO/ee+1GUh/focivuLct7oqdbzTRtl1dF85ZYyTTjwSDJzlng91yfv1UgP/uhBpayoAy1KYJcQiccOVQ2EiTBiN/FQFp30Ky5so3G0Onu+nNvTDaggPOHTlU1pdkmQiITNpmpFXIsSW53n8tFHvwx357Bl8Mc9+6VUmlj0JtaNTKTJ261kFph2var9aScCsw1e3bv8Xnk9YmiybhptsY8sc516s7en694EnM3+yghiuswlQugfGTD9NRSCseZFqRHZ+gdOiDMhFNqscGOvEtw6XsDXiSo8TYRUop5oI+bdKxc7KiQ4wXoKMu4RV5yP/Xic4uLcSBqXwoppCJHhSVj/1ec6OA7bJxvUPbFxL9pBIWr5jv6LeSYaZwTkfzqi7sZpXL4eK9fWtKPW8W6PyLBJmBfh/JQwINvbDrY4f/Xxb2vLFEogeVZT/3E92KCoCxqCsim1IdptLIfOi1AcMUjydcoGjD7EVXDt0ajTNFuBeDsXWeB/YF5ThiuSqdoLndzsoy93JvWzng4I/xhhMVOpWqlOuNlIzEgPaBjXT5DcXVK9Nakn7u21JKh8R9YU9rzdaluVz7qKRbhvHqynKZbZy6EGGALEPbP2NrKxhaIWYVY3MpSR5aIAZXtGTgZvwCtlXXJv9XUSXiMxmZVPqsaDL9UKhikvs/kwjdN7yLKFP/4g+c4B/TqNV1eQaDXN9vrvLcpmlBOYQpnlD78z5qzfjzTAl8dqvWr6pUXUsO8iEuhxlzKCeG5PQWn8G+mW6 8Uv3O7xp dSXHVmwUyleolWo7dQStYILVbymNE1J/v9PSZ+Ck9RpxM4WabNcdCmjyUkuBWsJIuRW8bKPyu6JaVS7cHxON18SgRYF6QDfE0EDc7/Prq2rbpWHo0hOkksS5oB7l4QLWuki3G9XSI1kbFSLFZYRQemo7ccQ== 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 2023/4/11 PM10:28, Kefeng Wang wrote: > > > On 2023/4/11 18:48, Shuai Xue wrote: >> Hardware errors could be signaled by synchronous interrupt, e.g.  when an >> error is detected by a background scrubber, or signaled by synchronous >> exception, e.g. when an uncorrected error is consumed. Both synchronous and >> asynchronous error are queued and handled by a dedicated kthread in >> workqueue. >> >> commit 7f17b4a121d0 ("ACPI: APEI: Kick the memory_failure() queue for >> synchronous errors") keep track of whether memory_failure() work was >> queued, and make task_work pending to flush out the workqueue so that the >> work for synchronous error is processed before returning to user-space. >> The trick ensures that the corrupted page is unmapped and poisoned. And >> after returning to user-space, the task starts at current instruction which >> triggering a page fault in which kernel will send SIGBUS to current process >> due to VM_FAULT_HWPOISON. >> >> However, the memory failure recovery for hwpoison-aware mechanisms does not >> work as expected. For example, hwpoison-aware user-space processes like >> QEMU register their customized SIGBUS handler and enable early kill mode by >> seting PF_MCE_EARLY at initialization. Then the kernel will directy notify >> the process by sending a SIGBUS signal in memory failure with wrong >> si_code: the actual user-space process accessing the corrupt memory >> location, but its memory failure work is handled in a kthread context, so >> it will send SIGBUS with BUS_MCEERR_AO si_code to the actual user-space >> process instead of BUS_MCEERR_AR in kill_proc(). >> >> To this end, separate synchronous and asynchronous error handling into >> different paths like X86 platform does: >> >> - valid synchronous errors: queue a task_work to synchronously send SIGBUS >>    before ret_to_user. >> - valid asynchronous errors: queue a work into workqueue to asynchronously >>    handle memory failure. >> - abnormal branches such as invalid PA, unexpected severity, no memory >>    failure config support, invalid GUID section, OOM, etc. >> >> Then for valid synchronous errors, the current context in memory failure is >> exactly belongs to the task consuming poison data and it will send SIBBUS >> with proper si_code. >> >> Fixes: 7f17b4a121d0 ("ACPI: APEI: Kick the memory_failure() queue for synchronous errors") >> Signed-off-by: Shuai Xue >> Tested-by: Ma Wupeng >> --- >>   drivers/acpi/apei/ghes.c | 91 +++++++++++++++++++++++++++------------- >>   include/acpi/ghes.h      |  3 -- >>   mm/memory-failure.c      | 13 ------ >>   3 files changed, 61 insertions(+), 46 deletions(-) >> >> diff --git a/drivers/acpi/apei/ghes.c b/drivers/acpi/apei/ghes.c >> index c479b85899f5..4b70955e25f9 100644 >> --- a/drivers/acpi/apei/ghes.c >> +++ b/drivers/acpi/apei/ghes.c >> @@ -452,28 +452,51 @@ static void ghes_clear_estatus(struct ghes *ghes, >>   } >>     /* >> - * Called as task_work before returning to user-space. >> - * Ensure any queued work has been done before we return to the context that >> - * triggered the notification. >> + * struct sync_task_work - for synchronous RAS event >> + * >> + * @twork:                callback_head for task work >> + * @pfn:                  page frame number of corrupted page >> + * @flags:                fine tune action taken >> + * >> + * Structure to pass task work to be handled before >> + * ret_to_user via task_work_add(). >>    */ >> -static void ghes_kick_task_work(struct callback_head *head) >> +struct sync_task_work { >> +    struct callback_head twork; >> +    u64 pfn; >> +    int flags; >> +}; >> + >> +static void memory_failure_cb(struct callback_head *twork) >>   { >> -    struct acpi_hest_generic_status *estatus; >> -    struct ghes_estatus_node *estatus_node; >> -    u32 node_len; >> +    int ret; >> +    struct sync_task_work *twcb = >> +        container_of(twork, struct sync_task_work, twork); >>   -    estatus_node = container_of(head, struct ghes_estatus_node, task_work); >> -    if (IS_ENABLED(CONFIG_ACPI_APEI_MEMORY_FAILURE)) >> -        memory_failure_queue_kick(estatus_node->task_work_cpu); >> +    ret = memory_failure(twcb->pfn, twcb->flags); >> +    kfree(twcb); >>   -    estatus = GHES_ESTATUS_FROM_NODE(estatus_node); >> -    node_len = GHES_ESTATUS_NODE_LEN(cper_estatus_len(estatus)); >> -    gen_pool_free(ghes_estatus_pool, (unsigned long)estatus_node, node_len); >> +    if (!ret) >> +        return; >> + >> +    /* >> +     * -EHWPOISON from memory_failure() means that it already sent SIGBUS >> +     * to the current process with the proper error info, > > This should be part of the comments of function memory_failure(), > >> +     * -EOPNOTSUPP means hwpoison_filter() filtered the error event, >> +     * > and this part is already there >> +     * In both cases, no further processing is required. >> +     */ > so, after that, I think we could drop this comment, also the same comment in x86's kill_me_maybe(). Ok, I will add comments on return value of memory_failure() and drop both this comment and that in kill_me_maybe() out. > >> +    if (ret == -EHWPOISON || ret == -EOPNOTSUPP) >> +        return; >> + >> +    pr_err("Memory error not recovered"); >> +    force_sig(SIGBUS); >>   } >>     static bool ghes_do_memory_failure(u64 physical_addr, int flags) >>   { >>       unsigned long pfn; >> +    struct sync_task_work *twcb; >>         if (!IS_ENABLED(CONFIG_ACPI_APEI_MEMORY_FAILURE)) >>           return false; >> @@ -486,6 +509,18 @@ static bool ghes_do_memory_failure(u64 physical_addr, int flags) >>           return false; >>       } >>   +    if (flags == MF_ACTION_REQUIRED && current->mm) { >> +        twcb = kmalloc(sizeof(*twcb), GFP_ATOMIC); >> +        if (!twcb) >> +            return false; >> + >> +        twcb->pfn = pfn; >> +        twcb->flags = flags; >> +        init_task_work(&twcb->twork, memory_failure_cb); >> +        task_work_add(current, &twcb->twork, TWA_RESUME); >> +        return true; >> +    } >> + >>       memory_failure_queue(pfn, flags); >>       return true; >>   } >> @@ -1000,9 +1035,8 @@ static void ghes_proc_in_irq(struct irq_work *irq_work) >>       struct ghes_estatus_node *estatus_node; >>       struct acpi_hest_generic *generic; >>       struct acpi_hest_generic_status *estatus; >> -    bool task_work_pending; >> +    bool queued, sync; >>       u32 len, node_len; >> -    int ret; >>         llnode = llist_del_all(&ghes_estatus_llist); >>       /* >> @@ -1015,27 +1049,25 @@ static void ghes_proc_in_irq(struct irq_work *irq_work) >>           estatus_node = llist_entry(llnode, struct ghes_estatus_node, >>                          llnode); >>           estatus = GHES_ESTATUS_FROM_NODE(estatus_node); >> +        sync = is_hest_sync_notify(estatus_node->ghes); >>           len = cper_estatus_len(estatus); >>           node_len = GHES_ESTATUS_NODE_LEN(len); >> -        task_work_pending = ghes_do_proc(estatus_node->ghes, estatus); >> + >> +        queued = ghes_do_proc(estatus_node->ghes, estatus) > +        /* >> +         * If no memory failure work is queued for abnormal synchronous >> +         * errors, do a force kill. >> +         */ >> +        if (sync && !queued) >> +            force_sig(SIGBUS); > > It's better to move this part into function ghes_do_proc(), because there is already an is_hest_sync_notify(), and no need return value, > so make ghes_do_proc() a void function, Apart from this, Good idea. I will do this and send a new version. > > Reviewed-by: Kefeng Wang Thank you. Cheers, Shuai > >> + >>           if (!ghes_estatus_cached(estatus)) { >>               generic = estatus_node->generic; >>               if (ghes_print_estatus(NULL, generic, estatus)) >>                   ghes_estatus_cache_add(generic, estatus); >>           } >> - >> -        if (task_work_pending && current->mm) { >> -            estatus_node->task_work.func = ghes_kick_task_work; >> -            estatus_node->task_work_cpu = smp_processor_id(); >> -            ret = task_work_add(current, &estatus_node->task_work, >> -                        TWA_RESUME); >> -            if (ret) >> -                estatus_node->task_work.func = NULL; >> -        } >> - >> -        if (!estatus_node->task_work.func) >> -            gen_pool_free(ghes_estatus_pool, >> -                      (unsigned long)estatus_node, node_len); >> +        gen_pool_free(ghes_estatus_pool, (unsigned long)estatus_node, >> +                  node_len); >>             llnode = next; >>       } >> @@ -1096,7 +1128,6 @@ static int ghes_in_nmi_queue_one_entry(struct ghes *ghes, >>         estatus_node->ghes = ghes; >>       estatus_node->generic = ghes->generic; >> -    estatus_node->task_work.func = NULL; >>       estatus = GHES_ESTATUS_FROM_NODE(estatus_node); >>         if (__ghes_read_estatus(estatus, buf_paddr, fixmap_idx, len)) { >> diff --git a/include/acpi/ghes.h b/include/acpi/ghes.h >> index 3c8bba9f1114..e5e0c308d27f 100644 >> --- a/include/acpi/ghes.h >> +++ b/include/acpi/ghes.h >> @@ -35,9 +35,6 @@ struct ghes_estatus_node { >>       struct llist_node llnode; >>       struct acpi_hest_generic *generic; >>       struct ghes *ghes; >> - >> -    int task_work_cpu; >> -    struct callback_head task_work; >>   }; >>     struct ghes_estatus_cache { >> diff --git a/mm/memory-failure.c b/mm/memory-failure.c >> index fae9baf3be16..6ea8c325acb3 100644 >> --- a/mm/memory-failure.c >> +++ b/mm/memory-failure.c >> @@ -2355,19 +2355,6 @@ static void memory_failure_work_func(struct work_struct *work) >>       } >>   } >>   -/* >> - * Process memory_failure work queued on the specified CPU. >> - * Used to avoid return-to-userspace racing with the memory_failure workqueue. >> - */ >> -void memory_failure_queue_kick(int cpu) >> -{ >> -    struct memory_failure_cpu *mf_cpu; >> - >> -    mf_cpu = &per_cpu(memory_failure_cpu, cpu); >> -    cancel_work_sync(&mf_cpu->work); >> -    memory_failure_work_func(&mf_cpu->work); >> -} >> - >>   static int __init memory_failure_init(void) >>   { >>       struct memory_failure_cpu *mf_cpu;